25 MB lightweight cross-platform database client for 100+ databases, including MySQL, PostgreSQL, SQLite, Redis, MongoDB, DuckDB, SQL Server, and Dameng. Built-in AI, MCP Server, CLI, desktop and Docker. | 轻量级跨平台数据库管理工具,支持 MySQL、PostgreSQL、SQLite、Redis、MongoDB、达梦等 100+ 数据库,提供桌面端、Docker、CLI、内置 AI 助手和 MCP。
TypeScript aicliclickhousedatabasedatabase-clientdatabase-managementdockerguimcpmongodbmysqlpostgresqlredisrustsql-serversqlitetaurivue
use percent_encoding::{percent_decode_str, utf8_percent_encode, AsciiSet, CONTROLS};
use regex::Regex;
use reqwest::header::{HeaderValue, AUTHORIZATION};
use reqwest::{Client as HttpClient, Method, StatusCode};
use serde::{Deserialize, Serialize};
use serde_json::Value;
use std::collections::{BTreeMap, BTreeSet, HashMap, HashSet};
use std::error::Error;
use std::net::{IpAddr, Ipv4Addr, Ipv6Addr, SocketAddr};
use std::sync::atomic::{AtomicBool, Ordering};
use std::sync::Arc;
use std::time::Duration;
use crate::db::document_result::DocumentQueryResult;
use crate::types::QueryResult;
use dbx_driver_support::db::{apply_tls_certificates, http_client_builder, with_connection_timeout};
pub mod db {
pub use dbx_driver_support::db::*;
pub use dbx_driver_support::{document_result, elasticsearch_sql};
}
pub use dbx_types::types;
const ELASTICSEARCH_PATH_SEGMENT_ENCODE_SET: &AsciiSet = &CONTROLS
.add(b' ')
.add(b'"')
.add(b'#')
.add(b'%')
.add(b'/')
.add(b'<')
.add(b'>')
.add(b'?')
.add(b'[')
.add(b'\\')
.add(b']')
.add(b'^')
.add(b'`')
.add(b'{')
.add(b'|')
.add(b'}');
const ELASTICSEARCH_QUERY_VALUE_ENCODE_SET: &AsciiSet =
&CONTROLS.add(b' ').add(b'"').add(b'#').add(b'%').add(b'&').add(b'+').add(b'/').add(b'=').add(b'?');
const KIBANA_PROXY_STATUS_HEADER: &str = "x-console-proxy-status-code";
const ELASTICSEARCH_REST_TABLE_MAX_BODY_BYTES: usize = 8 * 1024 * 1024;
const ELASTICSEARCH_REST_TABLE_MAX_ROWS: usize = 2_000;
const ELASTICSEARCH_REST_TABLE_MAX_CELLS: usize = 200_000;
const ELASTICSEARCH_MAPPING_MAX_DEPTH: u8 = 32;
const ELASTICSEARCH_MAPPING_MAX_COLUMNS: usize = 2_000;
const ELASTICSEARCH_MAPPING_THREAD_STACK: usize = 16 * 1024 * 1024;
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
enum ElasticsearchTransportMode {
Direct,
KibanaProxy,
}
pub struct EsClient {
http: HttpClient,
base_url: String,
fallback_base_urls: Vec<String>,
auth: Option<(String, String)>,
api_key: Option<HeaderValue>,
transport_mode: ElasticsearchTransportMode,
/// GET path used for connect / health / test (default "/").
connectivity_check_path: String,
/// 为 true 时完全跳过连通性检查(test_connection 直接返回 Ok)。
/// 用于账号连 `/` 或任何检查路径都无权限、且集群间权限不统一的场景。
connectivity_check_disabled: bool,
/// 正则:把易变的时间/滚动后缀折叠成 `*`,将同一前缀的滚动索引聚合成一个
/// pattern 节点。`None` 表示关闭聚合,展示原始索引名。
index_grouping: Option<Regex>,
/// 该集群能否用 PIT + `search_after` 深分页。用 `Arc` 是因为连接池里的实例会被
/// `clone()` 出去执行每次查询,探测结果必须回流到同一个连接上。
pit_search_supported: Arc<AtomicBool>,
/// 该集群的 `_sql` 端点是否只接受 PUT。ES 6.x 的 `_sql` 只注册了 GET/PUT,
/// POST 会返回 405(`allowed: [GET, PUT, HEAD, DELETE]`);ES 7/8 反过来只认
/// GET/POST。默认先按 POST 发,命中 405 后按连接改记 PUT。
sql_endpoint_uses_put: Arc<AtomicBool>,
}
impl EsClient {
pub fn new(
url: &str,
username: Option<&str>,
password: Option<&str>,
accept_invalid_certs: bool,
timeout: Duration,
) -> Self {
Self::new_with_mode(
url,
username,
password,
accept_invalid_certs,
timeout,
ElasticsearchTransportMode::Direct,
"/".to_string(),
false,
None,
None,
None,
None,
)
.expect("failed to build Elasticsearch HTTP client")
}
#[allow(clippy::too_many_arguments)]
fn new_with_mode(
url: &str,
username: Option<&str>,
password: Option<&str>,
accept_invalid_certs: bool,
timeout: Duration,
transport_mode: ElasticsearchTransportMode,
connectivity_check_path: String,
connectivity_check_disabled: bool,
index_grouping: Option<Regex>,
ca_cert_path: Option<&str>,
client_cert_path: Option<&str>,
client_key_path: Option<&str>,
) -> Result<Self, String> {
let base_url = url.trim_end_matches('/').to_string();
let auth = match (username, password) {
(Some(u), Some(p)) if !u.is_empty() => Some((u.to_string(), p.to_string())),
_ => None,
};
let mut builder = http_client_builder(timeout).danger_accept_invalid_certs(accept_invalid_certs);
builder = apply_tls_certificates(builder, ca_cert_path, client_cert_path, client_key_path)?;
if let Some(addrs) = elasticsearch_localhost_resolve_addrs(&base_url, connectivity_check_disabled) {
builder = builder.resolve_to_addrs("localhost", &addrs);
}
let http =
builder.build().map_err(|error| format!("Failed to initialize Elasticsearch HTTP client: {error}"))?;
let fallback_base_urls = elasticsearch_base_url_fallbacks(&base_url);
Ok(Self {
http,
base_url,
fallback_base_urls,
auth,
api_key: None,
transport_mode,
connectivity_check_path,
connectivity_check_disabled,
index_grouping,
pit_search_supported: Arc::new(AtomicBool::new(true)),
sql_endpoint_uses_put: Arc::new(AtomicBool::new(false)),
})
}
pub fn from_config(
url: &str,
username: Option<&str>,
password: Option<&str>,
tls_enabled: bool,
url_params: Option<&str>,
external_config: Option<&Value>,
timeout: Duration,
ca_cert_path: Option<&str>,
client_cert_path: Option<&str>,
client_key_path: Option<&str>,
) -> Result<Self, String> {
let kibana_base_path = elasticsearch_kibana_base_path(external_config);
let transport_mode = if kibana_base_path.is_some() {
ElasticsearchTransportMode::KibanaProxy
} else {
ElasticsearchTransportMode::Direct
};
let base_url = format!("{}{}", url.trim_end_matches('/'), kibana_base_path.as_deref().unwrap_or(""));
let connectivity_check_path = elasticsearch_connectivity_check_path(external_config);
let connectivity_check_disabled = elasticsearch_connectivity_check_disabled(external_config);
let index_grouping = elasticsearch_index_grouping(external_config);
let mut client = Self::new_with_mode(
&base_url,
username,
password,
elasticsearch_accept_invalid_certs(tls_enabled, url_params),
timeout,
transport_mode,
connectivity_check_path,
connectivity_check_disabled,
index_grouping,
ca_cert_path,
client_cert_path,
client_key_path,
)?;
if external_config.and_then(|config| config.get("authType")).and_then(Value::as_str) == Some("apiKey") {
// Reuse the encrypted main password slot rather than storing a secret in external_config.
let key = password
.map(str::trim)
.filter(|key| !key.is_empty())
.ok_or_else(|| "Elasticsearch API key is required".to_string())?;
let mut header = HeaderValue::from_str(&format!("ApiKey {key}"))
.map_err(|_| "Invalid Elasticsearch API key".to_string())?;
header.set_sensitive(true);
client.auth = None;
client.api_key = Some(header);
}
Ok(client)
}
fn get(&self, path: &str) -> reqwest::RequestBuilder {
self.request(Method::GET, path)
}
fn post(&self, path: &str) -> reqwest::RequestBuilder {
self.request(Method::POST, path)
}
fn put(&self, path: &str) -> reqwest::RequestBuilder {
self.request(Method::PUT, path)
}
fn delete(&self, path: &str) -> reqwest::RequestBuilder {
self.request(Method::DELETE, path)
}
fn request(&self, method: Method, path: &str) -> reqwest::RequestBuilder {
let req = match self.transport_mode {
ElasticsearchTransportMode::Direct => self.http.request(method, format!("{}{}", self.base_url, path)),
ElasticsearchTransportMode::KibanaProxy => self
.http
.post(format!("{}/api/console/proxy", self.base_url))
.query(&[("path", path), ("method", method.as_str())])
.header("kbn-xsrf", "true")
.header("osd-xsrf", "true"),
};
self.with_auth(req)
}
fn with_auth(&self, req: reqwest::RequestBuilder) -> reqwest::RequestBuilder {
if let Some(ref api_key) = self.api_key {
req.header(AUTHORIZATION, api_key.clone())
} else if let Some((ref user, ref pass)) = self.auth {
req.basic_auth(user, Some(pass))
} else {
req
}
}
fn response_status(&self, response: &reqwest::Response) -> StatusCode {
if self.transport_mode == ElasticsearchTransportMode::KibanaProxy {
if let Some(status) = response
.headers()
.get(KIBANA_PROXY_STATUS_HEADER)
.and_then(|value| value.to_str().ok())
.and_then(|value| value.parse::<u16>().ok())
.and_then(|value| StatusCode::from_u16(value).ok())
{
return status;
}
}
response.status()
}
fn supports_pit_search(&self) -> bool {
self.pit_search_supported.load(Ordering::Relaxed)
}
fn disable_pit_search(&self) {
self.pit_search_supported.store(false, Ordering::Relaxed);
}
fn sql_endpoint_uses_put(&self) -> bool {
self.sql_endpoint_uses_put.load(Ordering::Relaxed)
}
fn mark_sql_endpoint_uses_put(&self) {
self.sql_endpoint_uses_put.store(true, Ordering::Relaxed);
}
}
impl Clone for EsClient {
fn clone(&self) -> Self {
Self {
http: self.http.clone(),
base_url: self.base_url.clone(),
fallback_base_urls: self.fallback_base_urls.clone(),
auth: self.auth.clone(),
api_key: self.api_key.clone(),
transport_mode: self.transport_mode,
connectivity_check_path: self.connectivity_check_path.clone(),
connectivity_check_disabled: self.connectivity_check_disabled,
index_grouping: self.index_grouping.clone(),
// 共享标记而不是复制值:探测结果要对整个连接生效。
pit_search_supported: Arc::clone(&self.pit_search_supported),
sql_endpoint_uses_put: Arc::clone(&self.sql_endpoint_uses_put),
}
}
}
fn elasticsearch_kibana_base_path(external_config: Option<&Value>) -> Option<String> {
let config = external_config?.as_object()?;
let mode = config.get("mode").and_then(Value::as_str)?;
if !mode.eq_ignore_ascii_case("kibana") {
return None;
}
let base_path = config.get("kibanaBasePath").and_then(Value::as_str).unwrap_or("").trim().trim_matches('/');
Some(if base_path.is_empty() { String::new() } else { format!("/{base_path}") })
}
/// Path used for connectivity checks (test / open / health). Defaults to `/`.
/// Accepts bare paths (`my-index/_search`) or a single-line `GET path` paste.
pub fn elasticsearch_connectivity_check_path(external_config: Option<&Value>) -> String {
let raw = external_config
.and_then(Value::as_object)
.and_then(|config| config.get("connectivityCheckPath"))
.and_then(Value::as_str)
.unwrap_or("")
.trim();
if raw.is_empty() {
return "/".to_string();
}
// First line only — ignore accidental body lines from console paste.
let line = raw.lines().next().unwrap_or("").trim();
let without_method = line
.strip_prefix("GET ")
.or_else(|| line.strip_prefix("get "))
.or_else(|| line.strip_prefix("Get "))
.unwrap_or(line)
.trim();
if without_method.is_empty() || without_method == "/" {
return "/".to_string();
}
if without_method.starts_with('/') {
without_method.to_string()
} else {
format!("/{without_method}")
}
}
/// 解析连接配置里的索引聚合正则。**默认关闭**(命名格式因环境而异,不做全局假设)。
/// - 缺省/空/`off`/`none`/`false` → 关闭聚合,展示原始索引名。
/// - 其它 → 作为自定义正则;编译失败同样视为关闭,避免意外折叠。
///
/// 语义:用 `${1}*` 模板替换匹配区间——正则带捕获组 1 时保留其内容做前缀,
/// 不带捕获组时即把匹配到的“易变尾巴”替换成 `*`。
/// 是否完全跳过连通性检查。读取配置 `connectivityCheckDisabled`(布尔)。
/// 也兼容字符串 "true"/"1"/"yes"/"on"。用于账号无任何集群/索引探活权限的场景。
pub fn elasticsearch_connectivity_check_disabled(external_config: Option<&Value>) -> bool {
let Some(value) = external_config.and_then(Value::as_object).and_then(|c| c.get("connectivityCheckDisabled"))
else {
return false;
};
match value {
Value::Bool(b) => *b,
Value::String(s) => {
let s = s.trim();
s.eq_ignore_ascii_case("true") || s == "1" || s.eq_ignore_ascii_case("yes") || s.eq_ignore_ascii_case("on")
}
_ => false,
}
}
pub fn elasticsearch_index_grouping(external_config: Option<&Value>) -> Option<Regex> {
let raw = external_config
.and_then(Value::as_object)
.and_then(|config| config.get("indexGroupingPattern"))
.and_then(Value::as_str)
.unwrap_or("")
.trim();
if raw.is_empty()
|| raw.eq_ignore_ascii_case("off")
|| raw.eq_ignore_ascii_case("none")
|| raw.eq_ignore_ascii_case("false")
{
return None;
}
Regex::new(raw).ok()
}
/// 用分组正则把索引名聚合成 pattern。`None` 表示关闭聚合,原样返回。
fn group_index_names(names: Vec<String>, pattern: Option<&Regex>) -> Vec<String> {
let Some(re) = pattern else {
return names;
};
let mut buckets: BTreeMap<String, Vec<String>> = BTreeMap::new();
for name in names {
// `${1}*`:有捕获组则保留组1做前缀,无组则把匹配尾巴替换成 `*`。
let key = re.replace(&name, "${1}*").into_owned();
buckets.entry(key).or_default().push(name);
}
let mut out: Vec<String> = buckets.into_keys().collect();
out.sort();
out.dedup();
out
}
pub async fn test_connection(client: &mut EsClient, timeout: Duration) -> Result<(), String> {
// 用户显式关闭连通性检查:不发探活请求,直接视为可连。
if client.connectivity_check_disabled {
return Ok(());
}
let mut errors = Vec::new();
let urls = std::iter::once(client.base_url.clone()).chain(client.fallback_base_urls.clone());
let check_path = client.connectivity_check_path.clone();
for base_url in urls {
client.base_url = base_url.clone();
let path = check_path.clone();
let resp = with_connection_timeout("Elasticsearch", timeout, async {
client.get(&path).send().await.map_err(|e| {
format!(
"Elasticsearch connection failed for {} ({}): {}",
redact_elasticsearch_url(&base_url),
path,
format_reqwest_error(&e)
)
})
})
.await;
let resp = match resp {
Ok(resp) => resp,
Err(err) => {
errors.push(err);
continue;
}
};
let status = client.response_status(&resp);
if !status.is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error ({status}) for {check_path}: {body}"));
}
return Ok(());
}
if errors.is_empty() {
Err("Elasticsearch connection failed: no URL candidates".to_string())
} else {
Err(errors.join("; "))
}
}
pub fn elasticsearch_accept_invalid_certs(tls_enabled: bool, url_params: Option<&str>) -> bool {
tls_enabled
|| elasticsearch_url_params_flag(url_params, "sslmode", &["disable", "allow"])
|| elasticsearch_url_params_flag(url_params, "tlsverify", &["false", "0", "no", "off"])
|| elasticsearch_url_params_flag(url_params, "verify", &["false", "0", "no", "off"])
|| elasticsearch_url_params_flag(url_params, "insecure", &["true", "1", "yes", "on"])
|| elasticsearch_url_params_flag(url_params, "accept_invalid_certs", &["true", "1", "yes", "on"])
}
fn elasticsearch_url_params_flag(params: Option<&str>, key: &str, expected_values: &[&str]) -> bool {
params.unwrap_or("").trim().trim_start_matches('?').split('&').filter_map(|pair| pair.split_once('=')).any(
|(k, v)| {
k.trim().eq_ignore_ascii_case(key)
&& expected_values.iter().any(|expected| v.trim().eq_ignore_ascii_case(expected))
},
)
}
fn elasticsearch_base_url_fallbacks(base_url: &str) -> Vec<String> {
let Ok(parsed) = reqwest::Url::parse(base_url) else {
return Vec::new();
};
let Some(host) = parsed.host_str() else {
return Vec::new();
};
if !host.eq_ignore_ascii_case("localhost") {
return Vec::new();
}
let mut fallback = parsed;
if fallback.set_host(Some("127.0.0.1")).is_ok() {
vec![fallback.as_str().trim_end_matches('/').to_string()]
} else {
Vec::new()
}
}
fn elasticsearch_localhost_resolve_addrs(base_url: &str, connectivity_check_disabled: bool) -> Option<[SocketAddr; 2]> {
if !connectivity_check_disabled {
return None;
}
let parsed = reqwest::Url::parse(base_url).ok()?;
if !parsed.host_str()?.eq_ignore_ascii_case("localhost") {
return None;
}
Some([SocketAddr::new(IpAddr::V6(Ipv6Addr::LOCALHOST), 0), SocketAddr::new(IpAddr::V4(Ipv4Addr::LOCALHOST), 0)])
}
fn elasticsearch_index_path(index: &str, endpoint: &str) -> String {
format!("/{}/{}", elasticsearch_path_segment(index), endpoint.trim_start_matches('/'))
}
fn elasticsearch_path_segment(value: &str) -> String {
utf8_percent_encode(value, ELASTICSEARCH_PATH_SEGMENT_ENCODE_SET).to_string()
}
fn elasticsearch_query_value(value: &str) -> String {
utf8_percent_encode(value, ELASTICSEARCH_QUERY_VALUE_ENCODE_SET).to_string()
}
fn elasticsearch_document_path(index: &str, id: &str, document_type: Option<&str>, routing: Option<&str>) -> String {
let document_type = document_type.map(str::trim).filter(|value| !value.is_empty()).unwrap_or("_doc");
let base = format!(
"/{}/{}/{}",
elasticsearch_path_segment(index),
elasticsearch_path_segment(document_type),
elasticsearch_path_segment(id)
);
elasticsearch_path_with_routing_refresh(base, routing)
}
/// Auto-id index path: `POST /{index}/_doc` with optional custom routing.
fn elasticsearch_auto_id_document_path(index: &str, routing: Option<&str>) -> String {
let base = format!("/{}/_doc", elasticsearch_path_segment(index));
elasticsearch_path_with_routing_refresh(base, routing)
}
fn elasticsearch_path_with_routing_refresh(base: String, routing: Option<&str>) -> String {
if let Some(routing) = routing.map(str::trim).filter(|value| !value.is_empty()) {
format!("{base}?routing={}&refresh=true", elasticsearch_query_value(routing))
} else {
format!("{base}?refresh=true")
}
}
fn redact_elasticsearch_url(url: &str) -> String {
let Ok(mut parsed) = reqwest::Url::parse(url) else {
return url.to_string();
};
if !parsed.username().is_empty() {
let _ = parsed.set_username("user");
}
if parsed.password().is_some() {
let _ = parsed.set_password(Some("password"));
}
parsed.as_str().trim_end_matches('/').to_string()
}
fn format_reqwest_error(err: &reqwest::Error) -> String {
let mut parts = vec![err.to_string()];
let mut source = err.source();
while let Some(err) = source {
let text = err.to_string();
if !text.is_empty() && !parts.iter().any(|part| part == &text) {
parts.push(text);
}
source = err.source();
}
parts.join(": ")
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct ElasticsearchIndexEntry {
pub name: String,
pub aliases: Vec<String>,
}
struct ListedIndexNames {
indices: Vec<String>,
index_aliases: BTreeMap<String, Vec<String>>,
}
#[derive(Deserialize)]
struct CatIndex {
index: String,
}
#[derive(Deserialize)]
struct ResolveIndexResponse {
#[serde(default)]
indices: Vec<ResolveNamed>,
#[serde(default)]
aliases: Vec<ResolveAlias>,
#[serde(default)]
data_streams: Vec<ResolveNamed>,
}
#[derive(Deserialize)]
struct ResolveNamed {
name: String,
}
#[derive(Deserialize)]
struct ResolveAlias {
name: String,
#[serde(default)]
indices: Vec<String>,
}
/// 去掉 ES 内部索引(以 `.` 开头),排序并去重后返回可见索引名。
fn normalize_index_names(names: impl Iterator<Item = String>) -> Vec<String> {
let mut names: Vec<String> = names.filter(|name| !name.starts_with('.') && !name.is_empty()).collect();
names.sort();
names.dedup();
names
}
pub async fn list_indices(client: &EsClient) -> Result<Vec<String>, String> {
let names = list_raw_indices(client, false).await?.indices;
Ok(group_index_names(names, client.index_grouping.as_ref()))
}
pub async fn list_indices_with_aliases(client: &EsClient) -> Result<Vec<ElasticsearchIndexEntry>, String> {
let listed = list_raw_indices(client, true).await?;
Ok(merge_index_entries(listed.indices, &listed.index_aliases, client.index_grouping.as_ref()))
}
fn merge_index_entries(
indices: Vec<String>,
index_aliases: &BTreeMap<String, Vec<String>>,
grouping: Option<&Regex>,
) -> Vec<ElasticsearchIndexEntry> {
let Some(re) = grouping else {
return indices
.into_iter()
.map(|name| {
let aliases = index_aliases.get(&name).into_iter().flatten().cloned();
elasticsearch_index_entry(name, aliases)
})
.collect();
};
let mut buckets: BTreeMap<String, Vec<String>> = BTreeMap::new();
for name in indices {
let key = re.replace(&name, "${1}*").into_owned();
buckets.entry(key).or_default().push(name);
}
buckets
.into_iter()
.map(|(name, members)| {
let aliases = members.iter().flat_map(|member| index_aliases.get(member).into_iter().flatten().cloned());
elasticsearch_index_entry(name, aliases)
})
.collect()
}
fn elasticsearch_index_entry(name: String, aliases: impl IntoIterator<Item = String>) -> ElasticsearchIndexEntry {
let aliases = normalize_alias_names(aliases, &name);
ElasticsearchIndexEntry { name, aliases }
}
fn normalize_alias_names(aliases: impl IntoIterator<Item = String>, index_name: &str) -> Vec<String> {
let mut aliases: Vec<String> = aliases
.into_iter()
.filter(|alias| !alias.is_empty() && !alias.starts_with('.') && alias != index_name)
.collect();
aliases.sort();
aliases.dedup();
aliases
}
fn index_aliases_from_pairs(pairs: impl Iterator<Item = (String, String)>) -> BTreeMap<String, Vec<String>> {
let mut map: BTreeMap<String, BTreeSet<String>> = BTreeMap::new();
for (index, alias) in pairs {
if index.is_empty() || alias.is_empty() || alias.starts_with('.') {
continue;
}
map.entry(index).or_default().insert(alias);
}
map.into_iter().map(|(index, aliases)| (index, aliases.into_iter().collect())).collect()
}
async fn list_raw_indices(client: &EsClient, include_aliases: bool) -> Result<ListedIndexNames, String> {
// 主路径 `_cat/indices` 需要集群级 `monitor` 权限。仅有索引级权限的账号
// (例如日志采集用户)会在这里拿到 401/403,此时降级到索引级元数据端点。
let resp = client
.get("/_cat/indices?format=json&h=index")
.send()
.await
.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
let status = client.response_status(&resp);
if status.is_success() {
let indices: Vec<CatIndex> = resp.json().await.map_err(|e| format!("Elasticsearch parse error: {e}"))?;
return Ok(ListedIndexNames {
indices: normalize_index_names(indices.into_iter().map(|i| i.index)),
index_aliases: if include_aliases { alias_map_or_empty(client).await } else { BTreeMap::new() },
});
}
if status == StatusCode::FORBIDDEN || status == StatusCode::UNAUTHORIZED {
return list_indices_via_metadata(client).await;
}
let body = resp.text().await.unwrap_or_default();
Err(format!("Elasticsearch error: {body}"))
}
/// 集群 `monitor` 不可用时的降级:`_resolve/index` 与 `_alias` 属于
/// `indices:admin/*` 动作,`view_index_metadata`/`read` 索引权限即可访问,
/// 且 ES 安全层会把结果过滤为当前账号可见的索引。
async fn list_indices_via_metadata(client: &EsClient) -> Result<ListedIndexNames, String> {
// 优先 `_resolve/index`:同时覆盖普通索引、数据流(data stream)和别名。
if let Some(listed) = resolve_index_entries(client).await? {
return Ok(listed);
}
alias_endpoint_entries(client).await
}
async fn alias_map_or_empty(client: &EsClient) -> BTreeMap<String, Vec<String>> {
match alias_endpoint_entries(client).await {
Ok(listed) => listed.index_aliases,
Err(_) => BTreeMap::new(),
}
}
/// 通过 `GET /_resolve/index/*` 列举索引与别名。该端点缺权限时返回 `Ok(None)`,
/// 以便继续尝试 `_alias`;其它错误如实上抛。
async fn resolve_index_entries(client: &EsClient) -> Result<Option<ListedIndexNames>, String> {
let resp =
client.get("/_resolve/index/*").send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
let status = client.response_status(&resp);
if status == StatusCode::FORBIDDEN || status == StatusCode::UNAUTHORIZED {
return Ok(None);
}
if !status.is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
let body: ResolveIndexResponse = resp.json().await.map_err(|e| format!("Elasticsearch parse error: {e}"))?;
let indices = body.indices.into_iter().chain(body.data_streams).map(|item| item.name);
let index_aliases = index_aliases_from_pairs(
body.aliases
.into_iter()
.flat_map(|alias| alias.indices.into_iter().map(move |index| (index, alias.name.clone()))),
);
Ok(Some(ListedIndexNames { indices: normalize_index_names(indices), index_aliases }))
}
/// 通过 `GET /_alias` 列举索引(对象 key)和嵌套别名。
async fn alias_endpoint_entries(client: &EsClient) -> Result<ListedIndexNames, String> {
let resp = client.get("/_alias").send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
let body: serde_json::Map<String, Value> =
resp.json().await.map_err(|e| format!("Elasticsearch parse error: {e}"))?;
let mut indices = Vec::new();
let mut pairs = Vec::new();
for (index, value) in body {
if let Some(alias_map) = value.get("aliases").and_then(Value::as_object) {
pairs.extend(alias_map.keys().cloned().map(|alias| (index.clone(), alias)));
}
indices.push(index);
}
Ok(ListedIndexNames {
indices: normalize_index_names(indices.into_iter()),
index_aliases: index_aliases_from_pairs(pairs.into_iter()),
})
}
pub async fn get_columns(client: &EsClient, index: &str) -> Result<Vec<crate::db::ColumnInfo>, String> {
let path = elasticsearch_index_path(index, "_mapping");
let resp = client.get(&path).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
let bytes = resp.bytes().await.map_err(|e| format!("Elasticsearch read error: {e}"))?;
// serde_json parsing is recursive. Tokio worker stacks are ~2MB in debug and
// abort the process on overflow, so flatten on a dedicated larger stack.
tokio::task::spawn_blocking(move || mapping_bytes_to_columns_on_large_stack(bytes.to_vec()))
.await
.map_err(|error| format!("Elasticsearch mapping task failed: {error}"))?
}
fn mapping_properties(mapping: &Value) -> Option<&serde_json::Map<String, Value>> {
if let Some(properties) = mapping.pointer("/mappings/properties").and_then(Value::as_object) {
return Some(properties);
}
mapping
.get("mappings")
.and_then(Value::as_object)?
.values()
.find_map(|typed_mapping| typed_mapping.get("properties").and_then(Value::as_object))
}
fn mapping_bytes_to_columns_on_large_stack(bytes: Vec<u8>) -> Result<Vec<crate::db::ColumnInfo>, String> {
let (tx, rx) = std::sync::mpsc::sync_channel(1);
std::thread::Builder::new()
.name("es-mapping-columns".into())
.stack_size(ELASTICSEARCH_MAPPING_THREAD_STACK)
.spawn(move || {
let _ = tx.send(mapping_bytes_to_columns(&bytes));
})
.map_err(|error| format!("Elasticsearch mapping thread failed: {error}"))?;
rx.recv().map_err(|error| format!("Elasticsearch mapping thread failed: {error}"))?
}
fn mapping_bytes_to_columns(bytes: &[u8]) -> Result<Vec<crate::db::ColumnInfo>, String> {
let body: Value = serde_json::from_slice(bytes).map_err(|e| format!("Elasticsearch parse error: {e}"))?;
let mut seen = HashSet::new();
let mut columns = Vec::new();
if let Some(indices) = body.as_object() {
for index_mapping in indices.values() {
if let Some(properties) = mapping_properties(index_mapping) {
collect_mapping_columns("", properties, &mut seen, &mut columns);
}
}
}
columns.sort_by(|left, right| left.name.cmp(&right.name));
drop_json_iteratively(body);
Ok(columns)
}
fn collect_mapping_columns(
prefix: &str,
properties: &serde_json::Map<String, Value>,
seen: &mut HashSet<String>,
columns: &mut Vec<crate::db::ColumnInfo>,
) {
// Walk nested `properties` / multi-`fields` on the heap with hard caps.
// Recursive walks overflow tokio worker stacks on indexes like `dbx_all_types`.
let mut stack = vec![(prefix.to_string(), properties, 0_u8)];
while let Some((prefix, properties, depth)) = stack.pop() {
if columns.len() >= ELASTICSEARCH_MAPPING_MAX_COLUMNS {
break;
}
for (name, definition) in properties {
if columns.len() >= ELASTICSEARCH_MAPPING_MAX_COLUMNS {
break;
}
let field_name = if prefix.is_empty() { name.clone() } else { format!("{prefix}.{name}") };
let field_type = definition.get("type").and_then(Value::as_str);
let nested_properties = definition.get("properties").and_then(Value::as_object);
let nested_fields = definition.get("fields").and_then(Value::as_object);
if let Some(data_type) = field_type {
push_mapping_column(&field_name, data_type, seen, columns);
} else if nested_properties.is_some() {
// Elasticsearch infers `object` for a field with `properties` when
// the mapping omits an explicit type. Return the parent too because
// the document grid renders top-level object values as one column.
push_mapping_column(&field_name, "object", seen, columns);
}
if depth + 1 >= ELASTICSEARCH_MAPPING_MAX_DEPTH {
continue;
}
if let Some(properties) = nested_properties {
stack.push((field_name.clone(), properties, depth + 1));
}
if let Some(fields) = nested_fields {
stack.push((field_name, fields, depth + 1));
}
}
}
}
fn drop_json_iteratively(value: Value) {
let mut stack = vec![value];
while let Some(node) = stack.pop() {
match node {
Value::Array(items) => stack.extend(items),
Value::Object(map) => stack.extend(map.into_iter().map(|(_, child)| child)),
_ => {}
}
}
}
fn push_mapping_column(
name: &str,
data_type: &str,
seen: &mut HashSet<String>,
columns: &mut Vec<crate::db::ColumnInfo>,
) {
if !seen.insert(name.to_string()) {
return;
}
columns.push(crate::db::ColumnInfo {
name: name.to_string(),
data_type: data_type.to_string(),
is_nullable: true,
column_default: None,
is_primary_key: false,
extra: None,
comment: None,
numeric_precision: None,
numeric_scale: None,
character_maximum_length: None,
enum_values: None,
..Default::default()
});
}
#[derive(Deserialize)]
struct SearchResponse {
hits: SearchHits,
#[serde(rename = "_shards")]
shards: Option<ElasticsearchShards>,
#[serde(default)]
timed_out: bool,
#[serde(default)]
terminated_early: bool,
}
#[derive(Deserialize)]
struct SearchHits {
total: HitsTotal,
hits: Vec<SearchHit>,
}
enum HitsTotal {
Count(u64),
Value { value: u64, is_exact: bool },
}
impl HitsTotal {
fn value(&self) -> u64 {
match self {
Self::Count(value) | Self::Value { value, .. } => *value,
}
}
fn is_exact(&self) -> bool {
match self {
Self::Count(_) => true,
Self::Value { is_exact, .. } => *is_exact,
}
}
}
impl<'de> Deserialize<'de> for HitsTotal {
fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
where
D: serde::Deserializer<'de>,
{
let value = serde_json::Value::deserialize(deserializer)?;
if let Some(count) = value.as_u64() {
return Ok(Self::Count(count));
}
if let Some(count) = value.get("value").and_then(serde_json::Value::as_u64) {
// Object totals are exact only when Elasticsearch explicitly says
// so. Treat an absent or unknown relation conservatively.
let is_exact = value.get("relation").and_then(serde_json::Value::as_str) == Some("eq");
return Ok(Self::Value { value: count, is_exact });
}
Err(serde::de::Error::custom("expected hits.total as a number or an object with value"))
}
}
#[derive(Deserialize)]
struct SearchHit {
#[serde(rename = "_id")]
id: String,
#[serde(rename = "_type")]
document_type: Option<String>,
#[serde(rename = "_routing")]
routing: Option<String>,
#[serde(rename = "_source")]
source: serde_json::Value,
}
const ES_PIT_KEEP_ALIVE: &str = "1m";
const ES_SCROLL_KEEP_ALIVE: &str = "1m";
const ES_MAX_RESULT_WINDOW: usize = 10_000;
/// 一页检索的分页方式。PIT + `search_after` 是首选;集群不支持时降级为 from/size。
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
#[serde(tag = "type", rename_all = "snake_case")]
enum EsPageMode {
Pit { pit_id: String, keep_alive: String, search_after: Vec<serde_json::Value> },
Scroll { scroll_id: String, keep_alive: String },
Offset { from: u64 },
}
/// 三条链路共用的分页游标。`body` 只含 `query` / `sort`——随分页方式变化的
/// `from`/`size`/`pit`/`search_after` 由 [`es_build_page_request_body`] 现拼,
/// 因此 `body` 可以直接拿来比对「翻页途中过滤条件或排序是否被改过」。
#[derive(Debug, Clone, Serialize, Deserialize)]
struct EsPageCursor {
index: String,
body: serde_json::Value,
size: usize,
mode: EsPageMode,
}
/// Opaque cursor handed back to the caller. It is either a paged `_search`
/// cursor (PIT or offset), or a raw ES SQL cursor returned by `_sql`.
#[derive(Debug, Clone, Serialize, Deserialize)]
#[serde(tag = "kind", rename_all = "snake_case")]
enum EsSearchCursor {
Page(EsPageCursor),
Sql { cursor: String, columns_key: String, columns: Vec<serde_json::Value> },
}
fn encode_es_search_cursor(cursor: &EsSearchCursor) -> Result<String, String> {
serde_json::to_string(cursor).map_err(|e| format!("Failed to encode Elasticsearch search cursor: {e}"))
}
fn decode_es_search_cursor(cursor: &str) -> Result<EsSearchCursor, String> {
serde_json::from_str(cursor).map_err(|e| format!("Invalid Elasticsearch search cursor: {e}"))
}
pub async fn find_documents(
client: &EsClient,
index: &str,
skip: u64,
limit: i64,
filter: Option<&str>,
sort: Option<&str>,
) -> Result<DocumentQueryResult, String> {
let body = build_find_documents_body(skip, limit, filter, sort)?;
let path = elasticsearch_index_path(index, "_search");
let resp = client.post(&path).json(&body).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
let result: SearchResponse = resp.json().await.map_err(|e| format!("Elasticsearch parse error: {e}"))?;
search_response_to_document_result(result)
}
async fn open_es_pit(client: &EsClient, index: &str, keep_alive: &str) -> Result<String, String> {
let path =
format!("/{}/_pit?keep_alive={}", elasticsearch_path_segment(index), elasticsearch_query_value(keep_alive));
let resp = client.post(&path).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
let body: Value = resp.json().await.map_err(|e| format!("Elasticsearch PIT parse error: {e}"))?;
body.get("id")
.and_then(Value::as_str)
.map(str::to_string)
.ok_or_else(|| "Elasticsearch PIT response missing id".to_string())
}
async fn close_es_pit(client: &EsClient, pit_id: &str) -> Result<(), String> {
let resp = client
.delete("/_pit")
.json(&serde_json::json!({ "id": pit_id }))
.send()
.await
.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
Ok(())
}
async fn close_es_scroll(client: &EsClient, scroll_id: &str) -> Result<(), String> {
let resp = client
.delete("/_search/scroll")
.json(&serde_json::json!({ "scroll_id": [scroll_id] }))
.send()
.await
.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
Ok(())
}
async fn close_es_sql_cursor(client: &EsClient, cursor: &str) -> Result<(), String> {
let resp = send_es_sql_request(client, "/_sql/close", &serde_json::json!({ "cursor": cursor })).await?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
Ok(())
}
/// Sends a body to an Elasticsearch SQL endpoint, adapting to the cluster's
/// accepted HTTP method.
///
/// ES 7/8 register `GET`/`POST` on `_sql`, while ES 6.x only registers
/// `GET`/`PUT` and answers a POST with `405 Method Not Allowed` (issue #9484).
/// Start from POST — the method every supported version accepts the body form
/// of on 7/8 — and switch the whole connection to PUT once a 405 proves the
/// cluster rejects POST, so only the first query pays for the probe.
async fn send_es_sql_request(
client: &EsClient,
path: &str,
body: &serde_json::Value,
) -> Result<reqwest::Response, String> {
let use_put = client.sql_endpoint_uses_put();
let send = |use_put: bool| {
let builder = if use_put { client.put(path) } else { client.post(path) };
builder.json(body).send()
};
let response = send(use_put).await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if use_put || client.response_status(&response).as_u16() != 405 {
return Ok(response);
}
let retried = send(true).await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if client.response_status(&retried).as_u16() != 405 {
client.mark_sql_endpoint_uses_put();
}
Ok(retried)
}
/// Close a previously returned ES cursor. Supports DBX-wrapped paged cursors,
/// DBX-wrapped SQL cursors, and raw ES SQL cursors.
pub async fn close_cursor(client: &EsClient, cursor: &str) -> Result<(), String> {
if let Ok(search_cursor) = decode_es_search_cursor(cursor) {
return match search_cursor {
EsSearchCursor::Page(page) => match page.mode {
EsPageMode::Pit { pit_id, .. } => close_es_pit(client, &pit_id).await,
EsPageMode::Scroll { scroll_id, .. } if !scroll_id.is_empty() => {
close_es_scroll(client, &scroll_id).await
}
EsPageMode::Scroll { .. } => Ok(()),
// from/size 分页在服务端没有留下任何状态,无需释放。
EsPageMode::Offset { .. } => Ok(()),
},
EsSearchCursor::Sql { cursor, .. } => close_es_sql_cursor(client, &cursor).await,
};
}
close_es_sql_cursor(client, cursor).await
}
/// Decide whether a SELECT * query should use PIT + search_after.
///
/// The pagination plan always emits `LIMIT n OFFSET m` (even `OFFSET 0`), so a
/// zero/absent `from` plus an OFFSET clause is treated as a plan first page.
/// A user-written `OFFSET > 0` without a cursor keeps the legacy from/size path
/// so explicit offsets are not silently ignored.
fn should_use_search_cursor(body: &serde_json::Value, has_offset: bool, cursor: Option<&str>) -> bool {
if cursor.is_some() {
return true;
}
if !has_offset {
return false;
}
body.get("from").and_then(serde_json::Value::as_u64).is_none_or(|from| from == 0)
}
/// PIT + `search_after` 需要一个稳定的分片级 tiebreaker。`_shard_doc` 只在 PIT 检索下
/// 存在,所以仅在 PIT 模式追加;退回 from/size 时不需要它。
fn elasticsearch_apply_pit_sort_tiebreaker(items: &mut Vec<serde_json::Value>) {
// 无显式排序时通用文档排序给出的是 `"_doc"`——字符串形式,不是 `{"_doc": ...}`。
// 两种形式都要认,否则会拼出 `["_doc", {"_shard_doc":"asc"}]`。
let is_plain_doc_sort = items.len() == 1 && (items[0].as_str() == Some("_doc") || items[0].get("_doc").is_some());
if is_plain_doc_sort {
items.clear();
}
if items.iter().all(|item| item.get("_shard_doc").is_none()) {
items.push(serde_json::json!({ "_shard_doc": "asc" }));
}
}
/// 取出基础检索体里的 `sort`。非数组形式(对象/字符串)包成单元素数组而不是丢弃。
fn elasticsearch_sort_items(body: &serde_json::Value) -> Vec<serde_json::Value> {
match body.get("sort") {
Some(serde_json::Value::Array(items)) => items.clone(),
None | Some(serde_json::Value::Null) => Vec::new(),
Some(other) => vec![other.clone()],
}
}
struct EsPageRequest {
index: String,
body: serde_json::Value,
size: usize,
}
struct EsPageOutcome {
status: u16,
response: serde_json::Value,
next_cursor: Option<String>,
/// 指向当前页的游标,供调用方在翻完后释放 PIT。
active_cursor: Option<String>,
}
/// `es_send_page` 的失败原因。只有「ES 收到并拒绝了请求」才值得换一种分页方式重试;
/// 传输层失败(超时、连接断开)重发一次只会把用户的等待时间翻倍。
struct EsPageError {
message: String,
rejected_by_server: bool,
}
impl From<EsPageError> for String {
fn from(error: EsPageError) -> Self {
error.message
}
}
/// 集群接受了 PIT,却不认 PIT 专属的 `_shard_doc` 排序字段。这是能力缺失,
/// 对整个连接永久生效;其余错误只回退本次请求。
fn elasticsearch_error_lacks_pit_sort(error: &str) -> bool {
error.contains("_shard_doc")
}
/// 按分页方式拼出真正发给 ES 的检索体;游标里只存不变的部分。
fn es_build_page_request_body(page: &EsPageCursor) -> serde_json::Value {
let mut body = page.body.clone();
let Some(map) = body.as_object_mut() else {
return body;
};
map.insert("size".to_string(), serde_json::json!(page.size));
match &page.mode {
EsPageMode::Pit { pit_id, keep_alive, search_after } => {
// PIT 检索自带快照范围,不能再带 `from`。
map.remove("from");
map.insert("track_total_hits".to_string(), serde_json::json!(true));
let mut sort = elasticsearch_sort_items(&page.body);
elasticsearch_apply_pit_sort_tiebreaker(&mut sort);
map.insert("sort".to_string(), serde_json::Value::Array(sort));
map.insert("pit".to_string(), serde_json::json!({ "id": pit_id, "keep_alive": keep_alive }));
if !search_after.is_empty() {
map.insert("search_after".to_string(), serde_json::Value::Array(search_after.clone()));
}
}
EsPageMode::Scroll { .. } => {
map.remove("from");
}
EsPageMode::Offset { from } => {
map.insert("from".to_string(), serde_json::json!(from));
// from/size 各页之间没有快照,排序有并列值时页边界会重复或漏行。
// `_doc` 在所有 ES 版本都可用,作为次级键把并列打散。
let mut sort = elasticsearch_sort_items(&page.body);
let has_doc_key = sort.iter().any(|item| item.as_str() == Some("_doc") || item.get("_doc").is_some());
if !sort.is_empty() && !has_doc_key {
sort.push(serde_json::json!("_doc"));
map.insert("sort".to_string(), serde_json::Value::Array(sort));
}
}
}
body
}
/// 发出一页检索并算出下一页游标。`is_continuation` 为 false 表示这是首页,
/// 出错或翻到底时由这里关闭刚开的 PIT;续页的 PIT 归调用方持有,不能关。
async fn es_send_page(
client: &EsClient,
page: &EsPageCursor,
is_continuation: bool,
) -> Result<EsPageOutcome, EsPageError> {
let request_body = es_build_page_request_body(page);
// PIT 检索必须打全局 `/_search`——PIT 自带索引信息,路径里再带索引名会被拒。
let path = match &page.mode {
EsPageMode::Pit { .. } => "/_search".to_string(),
EsPageMode::Scroll { scroll_id, keep_alive } if !scroll_id.is_empty() => {
let _ = keep_alive;
"/_search/scroll".to_string()
}
EsPageMode::Scroll { keep_alive, .. } => {
format!(
"/{}/_search?scroll={}",
elasticsearch_path_segment(&page.index),
elasticsearch_query_value(keep_alive)
)
}
EsPageMode::Offset { .. } => elasticsearch_index_path(&page.index, "_search"),
};
let request_body = match &page.mode {
EsPageMode::Scroll { scroll_id, keep_alive } if !scroll_id.is_empty() => {
serde_json::json!({ "scroll": keep_alive, "scroll_id": scroll_id })
}
_ => request_body,
};
let opened_pit = match &page.mode {
EsPageMode::Pit { pit_id, .. } if !is_continuation => Some(pit_id.clone()),
_ => None,
};
let resp = client.post(&path).json(&request_body).send().await.map_err(|e| EsPageError {
message: format!("Elasticsearch request failed: {e}"),
rejected_by_server: false,
})?;
let status = client.response_status(&resp);
if !status.is_success() {
if let Some(pit_id) = &opened_pit {
let _ = close_es_pit(client, pit_id).await;
}
let error_body = resp.text().await.unwrap_or_default();
return Err(EsPageError { message: format!("Elasticsearch error: {error_body}"), rejected_by_server: true });
}
let response: serde_json::Value = resp.json().await.unwrap_or(serde_json::Value::Null);
let hits = response.pointer("/hits/hits").and_then(serde_json::Value::as_array).cloned().unwrap_or_default();
// `!hits.is_empty()` 不能省:`LIMIT 0` 时 `0 == 0` 会让空页被当成满页,
// offset 模式下就会不断吐出停在同一个 from 上的游标,翻页永远结束不了。
let is_full_page = !hits.is_empty() && hits.len() == page.size;
let last_sort = hits.last().and_then(|hit| hit.get("sort").and_then(serde_json::Value::as_array).cloned());
let (current_mode, next_mode) = match &page.mode {
EsPageMode::Pit { pit_id, keep_alive, search_after } => {
// 每次响应都可能给出新的 pit_id,必须跟着换,否则续页会用到过期的。
let pit_id = response
.get("pit_id")
.and_then(serde_json::Value::as_str)
.map(str::to_string)
.unwrap_or_else(|| pit_id.clone());
let current = EsPageMode::Pit {
pit_id: pit_id.clone(),
keep_alive: keep_alive.clone(),
search_after: last_sort.clone().unwrap_or_else(|| search_after.clone()),
};
let next = match last_sort {
Some(search_after) if is_full_page => {
Some(EsPageMode::Pit { pit_id, keep_alive: keep_alive.clone(), search_after })
}
_ => None,
};
(current, next)
}
EsPageMode::Scroll { scroll_id, keep_alive } => {
let scroll_id = response
.get("_scroll_id")
.and_then(serde_json::Value::as_str)
.map(str::to_string)
.unwrap_or_else(|| scroll_id.clone());
let current = EsPageMode::Scroll { scroll_id: scroll_id.clone(), keep_alive: keep_alive.clone() };
let next = (!scroll_id.is_empty() && is_full_page)
.then_some(EsPageMode::Scroll { scroll_id, keep_alive: keep_alive.clone() });
(current, next)
}
EsPageMode::Offset { from } => {
let next_from = from.saturating_add(page.size as u64);
// 下一页整页都要落在 max_result_window 内,否则那次请求必被 ES 拒绝。
let next = if is_full_page && next_from.saturating_add(page.size as u64) <= ES_MAX_RESULT_WINDOW as u64 {
Some(EsPageMode::Offset { from: next_from })
} else {
None
};
(EsPageMode::Offset { from: *from }, next)
}
};
let encode = |mode| {
encode_es_search_cursor(&EsSearchCursor::Page(EsPageCursor { mode, ..page.clone() }))
.map_err(|message| EsPageError { message, rejected_by_server: false })
};
let next_cursor = match next_mode {
Some(mode) => Some(encode(mode)?),
None => None,
};
// 首页就翻到底时没有「上一页」可回,直接关掉 PIT 而不是挂着等它过期。
if next_cursor.is_none() && !is_continuation {
match ¤t_mode {
EsPageMode::Pit { pit_id, .. } => {
let _ = close_es_pit(client, pit_id).await;
}
EsPageMode::Scroll { scroll_id, .. } if !scroll_id.is_empty() => {
let _ = close_es_scroll(client, scroll_id).await;
}
_ => {}
}
}
let active_cursor = if next_cursor.is_some() || is_continuation { Some(encode(current_mode)?) } else { None };
Ok(EsPageOutcome { status: status.as_u16(), response, next_cursor, active_cursor })
}
/// 三条链路共用的分页入口。首页优先 PIT + `search_after`,被拒就退回 from/size。
/// 判据是集群的实际响应而非版本号——`GET /` 可能没权限或被改写,兼容实现的版本号也不可信。
async fn es_execute_paged_search(
client: &EsClient,
request: EsPageRequest,
cursor: Option<EsPageCursor>,
) -> Result<EsPageOutcome, String> {
if let Some(page) = cursor {
let mismatch = if page.index != request.index {
Some("Elasticsearch cursor was created for a different index".to_string())
} else if page.body != request.body {
Some("Elasticsearch cursor no longer matches the current filter/sort; rerun the query".to_string())
} else {
None
};
if let Some(error) = mismatch {
match &page.mode {
EsPageMode::Pit { pit_id, .. } => {
let _ = close_es_pit(client, pit_id).await;
}
EsPageMode::Scroll { scroll_id, .. } if !scroll_id.is_empty() => {
let _ = close_es_scroll(client, scroll_id).await;
}
_ => {}
}
return Err(error);
}
return es_send_page(client, &page, true).await.map_err(String::from);
}
if client.supports_pit_search() {
let keep_alive = ES_PIT_KEEP_ALIVE.to_string();
// PIT 开不起来(版本过老、无权限、代理拦截)时静默退回 from/size。
if let Ok(pit_id) = open_es_pit(client, &request.index, &keep_alive).await {
let page = EsPageCursor {
index: request.index.clone(),
body: request.body.clone(),
size: request.size,
mode: EsPageMode::Pit { pit_id, keep_alive, search_after: Vec::new() },
};
match es_send_page(client, &page, false).await {
Ok(outcome) => return Ok(outcome),
// 请求压根没送到(超时、连接断开)时不重试:换个分页方式也送不到,
// 只会让用户多等一倍。
Err(error) if !error.rejected_by_server => return Err(error.message),
Err(error) => {
// 只有能力缺失才记到连接上;偶发 5xx 仅回退本次请求。
if elasticsearch_error_lacks_pit_sort(&error.message) {
client.disable_pit_search();
}
}
}
}
}
let page = EsPageCursor {
index: request.index.clone(),
body: request.body.clone(),
size: request.size,
mode: EsPageMode::Scroll { scroll_id: String::new(), keep_alive: ES_SCROLL_KEEP_ALIVE.to_string() },
};
match es_send_page(client, &page, false).await {
Ok(outcome) => Ok(outcome),
Err(error) if !error.rejected_by_server => Err(error.message),
Err(_) => {
let page = EsPageCursor {
index: request.index,
body: request.body,
size: request.size,
mode: EsPageMode::Offset { from: 0 },
};
es_send_page(client, &page, false).await.map_err(String::from)
}
}
}
/// 首页解析失败时游标不会交到调用方手上,这里释放它可能持有的 PIT。
/// 续页的 PIT 归调用方所有,不能碰。
async fn es_release_unreturned_cursor(client: &EsClient, outcome_cursor: Option<&String>, is_first_page: bool) {
if !is_first_page {
return;
}
if let Some(cursor) = outcome_cursor {
let _ = close_cursor(client, cursor).await;
}
}
/// 把一个不透明游标串解出分页游标;SQL 游标不能用来续 `_search`。
async fn es_page_cursor_from_str(
client: &EsClient,
cursor: Option<&str>,
misuse_error: &str,
) -> Result<Option<EsPageCursor>, String> {
let Some(cursor) = cursor else {
return Ok(None);
};
match decode_es_search_cursor(cursor)? {
EsSearchCursor::Page(page) => Ok(Some(page)),
EsSearchCursor::Sql { cursor, .. } => {
let _ = close_es_sql_cursor(client, &cursor).await;
Err(misuse_error.to_string())
}
}
}
/// 分页浏览一个 ES 索引。`cursor` 是上一次调用返回的不透明串;`None` 表示第一页。
pub async fn find_documents_with_cursor(
client: &EsClient,
index: &str,
limit: i64,
filter: Option<&str>,
sort: Option<&str>,
cursor: Option<&str>,
) -> Result<DocumentQueryResult, String> {
let size = limit.max(1).min(ES_MAX_RESULT_WINDOW as i64) as usize;
// 基础检索体只留 query/sort,from 与 size 交给分页执行器。
let mut body = build_find_documents_body(0, size as i64, filter, sort)?;
if let Some(map) = body.as_object_mut() {
map.remove("from");
map.remove("size");
}
let page_cursor =
es_page_cursor_from_str(client, cursor, "Elasticsearch SQL cursor cannot be used to continue a document query")
.await?;
let outcome =
es_execute_paged_search(client, EsPageRequest { index: index.to_string(), body, size }, page_cursor).await?;
let parsed: Result<SearchResponse, String> =
serde_json::from_value(outcome.response).map_err(|e| format!("Elasticsearch parse error: {e}"));
let mut document_result = match parsed.and_then(search_response_to_document_result) {
Ok(result) => result,
Err(error) => {
es_release_unreturned_cursor(client, outcome.next_cursor.as_ref(), cursor.is_none()).await;
return Err(error);
}
};
document_result.next_cursor = outcome.next_cursor;
Ok(document_result)
}
#[derive(Deserialize)]
struct CountResponse {
count: u64,
#[serde(rename = "_shards")]
shards: ElasticsearchShards,
}
#[derive(Deserialize)]
struct ElasticsearchShards {
total: u64,
successful: u64,
#[serde(default)]
skipped: u64,
failed: u64,
}
impl ElasticsearchShards {
fn is_complete(&self) -> bool {
// Elasticsearch only skips shards that cannot match, so successful and
// skipped shards together must account for every requested shard.
self.failed == 0 && self.successful.checked_add(self.skipped) == Some(self.total)
}
}
pub async fn count_documents(client: &EsClient, index: &str, filter: Option<&str>) -> Result<u64, String> {
let body = build_count_documents_body(filter)?;
let path = elasticsearch_index_path(index, "_count");
let resp = client.post(&path).json(&body).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
let result: CountResponse = resp.json().await.map_err(|e| format!("Elasticsearch count parse error: {e}"))?;
if !result.shards.is_complete() {
return Err(format!(
"Elasticsearch count returned an incomplete shard response: {} successful, {} skipped, {} failed of {} shards",
result.shards.successful, result.shards.skipped, result.shards.failed, result.shards.total,
));
}
Ok(result.count)
}
/// `GET /{index}/_mapping` —— 字段映射。
pub async fn get_index_mapping(client: &EsClient, index: &str) -> Result<Value, String> {
get_index_metadata(client, index, "_mapping").await
}
/// `GET /{index}/_settings` —— 索引配置(分片数、副本数、analysis 等)。
pub async fn get_index_settings(client: &EsClient, index: &str) -> Result<Value, String> {
get_index_metadata(client, index, "_settings").await
}
/// `GET /{index}/_stats` —— 索引统计(文档数、存储大小、各类操作计数)。
pub async fn get_index_stats(client: &EsClient, index: &str) -> Result<Value, String> {
get_index_metadata(client, index, "_stats").await
}
/// 侧边栏的索引节点可能是 `index_grouping` 折叠出的通配模式(如 `logs-*`)。
/// 这三个端点都原生接受通配,`elasticsearch_path_segment` 也不转义 `*`,
/// 因此聚合节点会如实返回该模式命中的全部索引。
async fn get_index_metadata(client: &EsClient, index: &str, endpoint: &str) -> Result<Value, String> {
let path = elasticsearch_index_path(index, endpoint);
let resp = client.get(&path).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
resp.json().await.map_err(|e| format!("Elasticsearch parse error: {e}"))
}
/// 清空索引数据的结果。文档删除可能部分失败(分片错误、并发写入),
/// 所以返回计数与失败明细,让调用方能如实告知用户而不是笼统报成功。
#[derive(Debug, Clone, Default, Serialize)]
#[serde(rename_all = "camelCase")]
pub struct ElasticsearchDeleteByQueryResult {
/// 匹配到的文档数。
pub total: u64,
/// 实际删除的文档数。
pub deleted: u64,
/// 因并发写入导致的版本冲突数(`conflicts=proceed` 下会被跳过而非中断)。
pub version_conflicts: u64,
/// 请求是否在删完前超时。
pub timed_out: bool,
/// 分片级失败的原始 JSON,逐条保留以便排查。
pub failures: Vec<String>,
}
#[derive(Deserialize)]
struct DeleteByQueryResponse {
#[serde(default)]
total: u64,
#[serde(default)]
deleted: u64,
#[serde(default)]
version_conflicts: u64,
#[serde(default)]
timed_out: bool,
#[serde(default)]
failures: Vec<Value>,
}
/// 清空索引数据:`POST /{index}/_delete_by_query` + `match_all`,只删文档,
/// 保留 mapping、settings、别名。
///
/// `conflicts=proceed` 让并发写入引起的版本冲突被跳过而不是中断整批删除;
/// `refresh=true` 让删除立即对随后的搜索与计数可见,否则清空后刷新数据页
/// 仍会看到旧文档。
pub async fn delete_all_documents(client: &EsClient, index: &str) -> Result<ElasticsearchDeleteByQueryResult, String> {
let path = format!("{}?conflicts=proceed&refresh=true", elasticsearch_index_path(index, "_delete_by_query"));
let body = serde_json::json!({ "query": { "match_all": {} } });
let resp = client.post(&path).json(&body).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
let result: DeleteByQueryResponse =
resp.json().await.map_err(|e| format!("Elasticsearch delete-by-query parse error: {e}"))?;
Ok(ElasticsearchDeleteByQueryResult {
total: result.total,
deleted: result.deleted,
version_conflicts: result.version_conflicts,
timed_out: result.timed_out,
failures: result.failures.iter().map(Value::to_string).collect(),
})
}
fn search_response_to_document_result(result: SearchResponse) -> Result<DocumentQueryResult, String> {
// A 200 search response can still omit failed shards. Only expose an
// exact total when both the total relation and shard metadata agree.
let total_is_exact = !result.timed_out
&& !result.terminated_early
&& result.hits.total.is_exact()
&& result.shards.as_ref().is_some_and(ElasticsearchShards::is_complete);
let total = result.hits.total.value();
let documents: Vec<serde_json::Value> = result
.hits
.hits
.into_iter()
.map(|hit| {
let mut doc = match hit.source {
serde_json::Value::Object(map) => map,
_ => serde_json::Map::new(),
};
doc.insert("_id".to_string(), serde_json::Value::String(hit.id));
if let Some(document_type) = hit.document_type.filter(|value| value != "_doc") {
doc.insert("_type".to_string(), serde_json::Value::String(document_type));
}
if let Some(routing) = hit.routing {
doc.insert("_routing".to_string(), serde_json::Value::String(routing));
}
serde_json::Value::Object(doc)
})
.collect();
// Keep the exact JSON numeric tokens alongside the compatibility value tree.
// Desktop IPC and browser JSON parsing otherwise coerce ES long values through
// JavaScript Number before the document editor can serialize them again.
let raw_documents = documents
.iter()
.map(serde_json::to_string)
.collect::<Result<Vec<_>, _>>()
.map_err(|e| format!("Elasticsearch document serialization failed: {e}"))?;
Ok(DocumentQueryResult {
documents,
raw_documents: Some(raw_documents),
extended_documents: None,
total,
total_is_exact,
next_cursor: None,
})
}
fn build_find_documents_body(
skip: u64,
limit: i64,
filter: Option<&str>,
sort: Option<&str>,
) -> Result<serde_json::Value, String> {
let mut body = serde_json::Map::new();
body.insert("from".to_string(), serde_json::json!(skip));
body.insert("size".to_string(), serde_json::json!(limit));
if let Some(query) = elasticsearch_query_from_document_filter(filter)? {
body.insert("query".to_string(), query);
}
body.insert("sort".to_string(), elasticsearch_sort_from_document_sort(sort)?);
Ok(serde_json::Value::Object(body))
}
fn build_count_documents_body(filter: Option<&str>) -> Result<serde_json::Value, String> {
let mut body = serde_json::Map::new();
if let Some(query) = elasticsearch_query_from_document_filter(filter)? {
body.insert("query".to_string(), query);
}
Ok(serde_json::Value::Object(body))
}
fn elasticsearch_query_from_document_filter(filter: Option<&str>) -> Result<Option<serde_json::Value>, String> {
let Some(filter) = filter.map(str::trim).filter(|value| !value.is_empty()) else {
return Ok(None);
};
let value: serde_json::Value = serde_json::from_str(filter).map_err(|e| format!("Invalid filter JSON: {e}"))?;
let query = translate_document_filter_value(&value)?;
Ok(query)
}
fn translate_document_filter_value(value: &serde_json::Value) -> Result<Option<serde_json::Value>, String> {
let Some(object) = value.as_object() else {
return Err("Elasticsearch filter must be a JSON object".to_string());
};
if object.is_empty() {
return Ok(None);
}
let mut must = Vec::new();
for (key, value) in object {
match key.as_str() {
"$and" => must.extend(translate_logical_filter_array("$and", value)?),
"$or" => {
let should = translate_logical_filter_array("$or", value)?;
if !should.is_empty() {
must.push(serde_json::json!({ "bool": { "should": should, "minimum_should_match": 1 } }));
}
}
"$esQuery" => {
if !value.is_object() {
return Err("$esQuery must be an object".to_string());
}
must.push(value.clone());
}
key if key.starts_with('$') => {
return Err(format!("Unsupported Elasticsearch filter operator: {key}"));
}
field => must.push(translate_field_filter(field, value)?),
}
}
Ok(single_or_bool_filter(must))
}
fn translate_logical_filter_array(operator: &str, value: &serde_json::Value) -> Result<Vec<serde_json::Value>, String> {
let items = value.as_array().ok_or_else(|| format!("{operator} must be an array"))?;
let mut queries = Vec::new();
for item in items {
if let Some(query) = translate_document_filter_value(item)? {
queries.push(query);
}
}
Ok(queries)
}
fn translate_field_filter(field: &str, value: &serde_json::Value) -> Result<serde_json::Value, String> {
let Some(object) = value.as_object() else {
return Ok(term_or_null_query(field, value));
};
if object.keys().any(|key| key.starts_with('$')) {
return translate_field_operator_filter(field, object);
}
Ok(term_or_null_query(field, value))
}
fn translate_field_operator_filter(
field: &str,
object: &serde_json::Map<String, serde_json::Value>,
) -> Result<serde_json::Value, String> {
let mut must = Vec::new();
let mut must_not = Vec::new();
let mut range = serde_json::Map::new();
for (operator, value) in object {
match operator.as_str() {
"$options" => {}
"$ne" => {
if value.is_null() {
must.push(serde_json::json!({ "exists": { "field": field } }));
} else {
must_not.push(serde_json::json!({ "term": { field: value.clone() } }));
}
}
"$gt" => {
range.insert("gt".to_string(), value.clone());
}
"$gte" => {
range.insert("gte".to_string(), value.clone());
}
"$lt" => {
range.insert("lt".to_string(), value.clone());
}
"$lte" => {
range.insert("lte".to_string(), value.clone());
}
"$regex" => {
must.push(regex_like_query(field, value, object.get("$options"))?);
}
"$not" => {
let Some(inner) = value.as_object() else {
return Err("$not must be a JSON object".to_string());
};
if let Some(regex) = inner.get("$regex") {
must_not.push(regex_like_query(
field,
regex,
inner.get("$options").or_else(|| object.get("$options")),
)?);
} else {
return Err("Unsupported Elasticsearch $not filter".to_string());
}
}
other => return Err(format!("Unsupported Elasticsearch field filter operator: {other}")),
}
}
if !range.is_empty() {
must.push(serde_json::json!({ "range": { field: serde_json::Value::Object(range) } }));
}
match (must.len(), must_not.is_empty()) {
(1, true) => Ok(must.remove(0)),
(0, false) => Ok(serde_json::json!({ "bool": { "must_not": must_not } })),
_ => {
let mut bool_query = serde_json::Map::new();
if !must.is_empty() {
bool_query.insert("must".to_string(), serde_json::Value::Array(must));
}
if !must_not.is_empty() {
bool_query.insert("must_not".to_string(), serde_json::Value::Array(must_not));
}
Ok(serde_json::json!({ "bool": bool_query }))
}
}
}
fn term_or_null_query(field: &str, value: &serde_json::Value) -> serde_json::Value {
if value.is_null() {
serde_json::json!({ "bool": { "must_not": [{ "exists": { "field": field } }] } })
} else {
serde_json::json!({ "term": { field: value.clone() } })
}
}
fn regex_like_query(
field: &str,
value: &serde_json::Value,
options: Option<&serde_json::Value>,
) -> Result<serde_json::Value, String> {
let pattern = value.as_str().ok_or_else(|| "$regex must be a string for Elasticsearch filters".to_string())?;
let case_insensitive = options
.and_then(serde_json::Value::as_str)
.is_some_and(|value| value.chars().any(|ch| ch.eq_ignore_ascii_case(&'i')));
Ok(serde_json::json!({
"wildcard": {
field: {
"value": wildcard_contains_pattern(pattern),
"case_insensitive": case_insensitive
}
}
}))
}
fn wildcard_contains_pattern(pattern: &str) -> String {
if pattern.starts_with('*') || pattern.ends_with('*') {
pattern.to_string()
} else {
format!("*{}*", pattern)
}
}
fn single_or_bool_filter(mut queries: Vec<serde_json::Value>) -> Option<serde_json::Value> {
match queries.len() {
0 => None,
1 => queries.pop(),
_ => Some(serde_json::json!({ "bool": { "filter": queries } })),
}
}
fn elasticsearch_sort_from_document_sort(sort: Option<&str>) -> Result<serde_json::Value, String> {
let Some(sort) = sort.map(str::trim).filter(|value| !value.is_empty()) else {
return Ok(serde_json::json!(["_doc"]));
};
let value: serde_json::Value = serde_json::from_str(sort).map_err(|e| format!("Invalid sort JSON: {e}"))?;
let object = value.as_object().ok_or_else(|| "Elasticsearch sort must be a JSON object".to_string())?;
if object.is_empty() {
return Ok(serde_json::json!(["_doc"]));
}
let items = object
.iter()
.map(|(field, direction)| {
let order = match direction {
serde_json::Value::Number(number) if number.as_i64().unwrap_or(1) < 0 => "desc",
serde_json::Value::String(value) if value.eq_ignore_ascii_case("desc") => "desc",
_ => "asc",
};
serde_json::json!({ field: { "order": order } })
})
.collect::<Vec<_>>();
Ok(serde_json::Value::Array(items))
}
pub async fn insert_document(
client: &EsClient,
index: &str,
doc_json: &str,
routing: Option<&str>,
) -> Result<String, String> {
// Prefer explicit routing arg; fall back to body metadata for backward compatibility.
let (doc, routing) = elasticsearch_document_body_and_routing_from_json(doc_json, routing)?;
let path = elasticsearch_auto_id_document_path(index, routing.as_deref());
let resp = client.post(&path).json(&doc).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
let result: serde_json::Value = resp.json().await.map_err(|e| format!("Elasticsearch parse error: {e}"))?;
Ok(result["_id"].as_str().unwrap_or("").to_string())
}
pub async fn update_document(
client: &EsClient,
index: &str,
id: &str,
doc_json: &str,
routing: Option<&str>,
) -> Result<u64, String> {
let (doc, routing, document_type) = elasticsearch_update_document_body_and_metadata(doc_json, routing)?;
let path = elasticsearch_document_path(index, id, document_type.as_deref(), routing.as_deref());
let resp = client.put(&path).json(&doc).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
Ok(1)
}
fn elasticsearch_update_document_body_and_metadata(
doc_json: &str,
routing: Option<&str>,
) -> Result<(serde_json::Value, Option<String>, Option<String>), String> {
let (mut doc, routing) = elasticsearch_document_body_and_routing_from_json(doc_json, routing)?;
let document_type = match &mut doc {
serde_json::Value::Object(map) => map.remove("_type").and_then(|value| match value {
serde_json::Value::String(value) => {
let trimmed = value.trim();
(!trimmed.is_empty()).then(|| trimmed.to_string())
}
_ => None,
}),
_ => None,
};
Ok((doc, routing, document_type))
}
fn elasticsearch_document_body_and_routing_from_json(
doc_json: &str,
routing: Option<&str>,
) -> Result<(serde_json::Value, Option<String>), String> {
let mut doc: serde_json::Value = serde_json::from_str(doc_json).map_err(|e| format!("Invalid JSON: {e}"))?;
let mut routing = routing.map(str::trim).filter(|value| !value.is_empty()).map(str::to_string);
if let serde_json::Value::Object(map) = &mut doc {
map.remove("_id");
if routing.is_none() {
routing = map.get("_routing").and_then(elasticsearch_routing_from_value);
}
map.remove("_routing");
}
Ok((doc, routing))
}
fn elasticsearch_routing_from_value(value: &serde_json::Value) -> Option<String> {
match value {
serde_json::Value::String(value) => {
let trimmed = value.trim();
(!trimmed.is_empty()).then(|| trimmed.to_string())
}
_ => None,
}
}
pub async fn delete_document(
client: &EsClient,
index: &str,
id: &str,
document_type: Option<&str>,
routing: Option<&str>,
) -> Result<u64, String> {
let path = elasticsearch_document_path(index, id, document_type, routing);
let resp = client.delete(&path).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
if !client.response_status(&resp).is_success() {
let body = resp.text().await.unwrap_or_default();
return Err(format!("Elasticsearch error: {body}"));
}
Ok(1)
}
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
enum ElasticsearchRestBodyKind {
Json,
Ndjson,
}
#[derive(Debug, Clone, PartialEq, Eq)]
struct ElasticsearchRestRequest {
method: Method,
path: String,
body: Option<String>,
body_kind: ElasticsearchRestBodyKind,
}
fn strip_leading_elasticsearch_comments(input: &str) -> &str {
let mut rest = input;
loop {
rest = rest.trim_start();
if let Some(comment) = rest.strip_prefix('#').or_else(|| rest.strip_prefix("//")) {
rest = comment.split_once('\n').map_or("", |(_, remaining)| remaining);
continue;
}
if let Some(comment) = rest.strip_prefix("/*") {
rest = comment.split_once("*/").map_or("", |(_, remaining)| remaining);
continue;
}
return rest.trim();
}
}
fn is_elasticsearch_ndjson_path(path: &str) -> bool {
let path = path.split('?').next().unwrap_or(path).trim_end_matches('/');
path == "/_bulk"
|| path.ends_with("/_bulk")
|| path == "/_msearch"
|| path.ends_with("/_msearch")
|| path == "/_msearch/template"
|| path.ends_with("/_msearch/template")
}
fn is_elasticsearch_cat_path(path: &str) -> bool {
let path = path.split('?').next().unwrap_or(path).trim_end_matches('/');
path == "/_cat" || path.starts_with("/_cat/")
}
fn elasticsearch_query_parameter(path: &str, name: &str) -> Option<String> {
let query = path.split_once('?')?.1;
query.split('&').find_map(|part| {
let (key, value) = part.split_once('=').unwrap_or((part, ""));
let key = percent_decode_str(key).decode_utf8_lossy();
if key.eq_ignore_ascii_case(name) {
Some(percent_decode_str(value).decode_utf8_lossy().into_owned())
} else {
None
}
})
}
fn add_default_cat_json_format(mut request: ElasticsearchRestRequest) -> ElasticsearchRestRequest {
if is_elasticsearch_cat_path(&request.path) && elasticsearch_query_parameter(&request.path, "format").is_none() {
request.path.push(if request.path.contains('?') { '&' } else { '?' });
request.path.push_str("format=json");
}
request
}
fn normalize_elasticsearch_rest_path(path: &str) -> String {
let (path_part, query) = path.split_once('?').map_or((path, None), |(path, query)| (path, Some(query)));
let mut normalized = String::with_capacity(path.len());
let mut chars = path_part.chars().peekable();
let mut in_date_math = false;
while let Some(ch) = chars.next() {
if ch == '%' {
let mut lookahead = chars.clone();
if let (Some(first), Some(second)) = (lookahead.next(), lookahead.next()) {
if first.is_ascii_hexdigit() && second.is_ascii_hexdigit() {
normalized.push(ch);
normalized.push(chars.next().unwrap());
normalized.push(chars.next().unwrap());
continue;
}
}
}
if ch == '<' {
in_date_math = true;
}
let encoded = if in_date_math {
match ch {
'<' => Some("%3C"),
'>' => Some("%3E"),
'/' => Some("%2F"),
'{' => Some("%7B"),
'}' => Some("%7D"),
'|' => Some("%7C"),
'+' => Some("%2B"),
':' => Some("%3A"),
',' => Some("%2C"),
_ => None,
}
} else {
None
};
if let Some(encoded) = encoded {
normalized.push_str(encoded);
} else {
normalized.push(ch);
}
if ch == '>' {
in_date_math = false;
}
}
if let Some(query) = query {
normalized.push('?');
normalized.push_str(query);
}
normalized
}
fn parse_elasticsearch_rest_request(input: &str) -> Result<ElasticsearchRestRequest, String> {
let input = strip_leading_elasticsearch_comments(input);
if input.is_empty() {
return Err("Invalid query: expected METHOD /path".to_string());
}
let (request_line, body) = input.split_once('\n').map_or((input, None), |(line, body)| {
let body = body.trim();
(line, (!body.is_empty()).then(|| body.to_string()))
});
let (method, path) =
request_line.trim().split_once(char::is_whitespace).ok_or("Invalid query: expected METHOD /path")?;
let method = method.to_ascii_uppercase();
let method = Method::from_bytes(method.as_bytes())
.map_err(|_| format!("Unsupported HTTP method: {method}. Use GET, POST, PUT, DELETE, or HEAD."))?;
if !matches!(method, Method::GET | Method::POST | Method::PUT | Method::DELETE | Method::HEAD) {
return Err(format!("Unsupported HTTP method: {}. Use GET, POST, PUT, DELETE, or HEAD.", method.as_str()));
}
let path = path.trim();
if path.is_empty() {
return Err("Invalid query: expected METHOD /path".to_string());
}
let path = if path.starts_with('/') { path.to_string() } else { format!("/{path}") };
let path = normalize_elasticsearch_rest_path(&path);
let body_kind = if is_elasticsearch_ndjson_path(&path) {
ElasticsearchRestBodyKind::Ndjson
} else {
ElasticsearchRestBodyKind::Json
};
Ok(ElasticsearchRestRequest { method, path, body, body_kind })
}
fn validate_elasticsearch_ndjson(body: &str) -> Result<String, String> {
for (index, line) in body.lines().enumerate() {
let line = line.trim();
if line.is_empty() {
continue;
}
serde_json::from_str::<serde_json::Value>(line)
.map_err(|error| format!("Invalid NDJSON body at line {}: {error}", index + 1))?;
}
let mut normalized = body.trim_end().to_string();
normalized.push('\n');
Ok(normalized)
}
pub type SqlResponseParser = fn(&serde_json::Value, std::time::Instant) -> Option<QueryResult>;
pub async fn execute_rest_query(client: &EsClient, input: &str) -> Result<QueryResult, String> {
execute_rest_query_with_sql_parser(client, input, parse_sql_response, None).await
}
pub async fn execute_rest_query_with_cursor(
client: &EsClient,
input: &str,
cursor: Option<&str>,
) -> Result<QueryResult, String> {
execute_rest_query_with_sql_parser(client, input, parse_sql_response, cursor).await
}
pub async fn execute_rest_query_with_sql_parser(
client: &EsClient,
input: &str,
sql_response_parser: SqlResponseParser,
cursor: Option<&str>,
) -> Result<QueryResult, String> {
let start = std::time::Instant::now();
let input = strip_leading_elasticsearch_comments(input);
if let Some(search_query) = parse_select_star_search_query(input) {
return execute_search_query(client, search_query, start, sql_response_parser, cursor).await;
}
if is_elasticsearch_sql_query(input) {
// `SELECT *` with clauses our simple parser doesn't cover (WHERE, IN,
// BETWEEN, LIKE, ...). Translate the SQL to an ES `_search` body
// ourselves rather than going through ES's `_sql` endpoint — `_sql`
// refuses several common shapes (LIKE on a `text` field with no
// `.keyword`, `SELECT *` returning an array field like `host.ip`, ...)
// that translate cleanly to raw DSL. Adapt first so that hyphenated
// index names (`aifanfan-python-bot-logs-*`) and `@timestamp`-style
// identifiers come out as double-quoted identifiers sqlparser will
// accept.
let adapted_for_translator = adapt_elasticsearch_sql_query(input);
match crate::db::elasticsearch_sql::translate_select_star(&adapted_for_translator) {
Ok(Some(translated)) => {
return execute_translated_select_star(client, translated, start, sql_response_parser, cursor).await;
}
Ok(None) => {}
Err(message) => return Err(format!("Elasticsearch SQL error: {message}")),
}
return execute_sql_query(client, input, start, sql_response_parser, cursor).await;
}
// CAT APIs default to text, so request JSON for an unformatted CAT call.
// The frontend renders the returned HTTP body in its JSON response panel.
let request = add_default_cat_json_format(parse_elasticsearch_rest_request(input)?);
let mut builder = client.request(request.method, &request.path);
if let Some(body) = request.body {
builder = match request.body_kind {
ElasticsearchRestBodyKind::Json => {
let json: serde_json::Value =
serde_json::from_str(&body).map_err(|e| format!("Invalid JSON body: {e}"))?;
builder.json(&json)
}
ElasticsearchRestBodyKind::Ndjson => builder
.header(reqwest::header::CONTENT_TYPE, "application/x-ndjson")
.body(validate_elasticsearch_ndjson(&body)?),
};
}
let resp = builder.send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
let status = client.response_status(&resp).as_u16();
let body = resp.text().await.map_err(|e| format!("Elasticsearch response read failed: {e}"))?;
parse_elasticsearch_rest_response_with_sql_parser(status, &body, start, sql_response_parser)
}
// Size to use when `SELECT *` is run without an explicit LIMIT — large enough
// to be useful, small enough that the user doesn't accidentally pull millions
// of documents. The result-grid surfaces the index's true total separately so
// the user can see how much was actually held back.
const AUTO_PAGED_SELECT_STAR_SIZE: usize = 100;
const ES_SQL_FETCH_SIZE: usize = 10_000;
struct ElasticsearchSearchQuery {
index: String,
body: serde_json::Value,
// True when the SQL came through the pagination plan (it carries OFFSET).
// In that case the result-grid total must reflect the index's true match
// count so the front-end can compute the total page count. A bare
// user-written `LIMIT N` (no OFFSET) is the "give me exactly N rows" case
// and reports affected_rows = N for client-side paging.
from_plan_pagination: bool,
}
/// `SELECT *` 快路径与 SQL 翻译路径只有字段来源不同,检索与分页逻辑收敛到这里。
struct EsIndexedSearch {
index: String,
body: serde_json::Value,
/// SQL 里带了 OFFSET,说明分页由前端的分页计划驱动(而不是用户手写的 LIMIT)。
from_plan_pagination: bool,
/// 用索引的真实命中总数覆盖 `affected_rows`,让表格能算出总页数。
report_index_total: bool,
}
async fn execute_indexed_search(
client: &EsClient,
search: EsIndexedSearch,
start: std::time::Instant,
sql_response_parser: SqlResponseParser,
cursor: Option<&str>,
) -> Result<QueryResult, String> {
let EsIndexedSearch { index, mut body, from_plan_pagination, report_index_total } = search;
// Only pagination-plan first pages (OFFSET 0) and explicit cursor
// continuations use PIT + search_after. A bare user `LIMIT n` or a
// user-written `OFFSET > 0` stays on the legacy from/size path.
if !should_use_search_cursor(&body, from_plan_pagination, cursor) {
let path = elasticsearch_index_path(&index, "_search");
let resp =
client.post(&path).json(&body).send().await.map_err(|e| format!("Elasticsearch request failed: {e}"))?;
let status = client.response_status(&resp).as_u16();
let response: serde_json::Value = resp.json().await.unwrap_or(serde_json::Value::Null);
let index_total = response.pointer("/hits/total/value").and_then(|v| v.as_u64());
let mut result = parse_elasticsearch_response_with_sql_parser(status, response, start, sql_response_parser)?;
if report_index_total {
if let Some(total) = index_total {
result.affected_rows = total;
}
}
return Ok(result);
}
let page_cursor =
es_page_cursor_from_str(client, cursor, "Elasticsearch SQL cursor cannot be used to continue a _search query")
.await?;
// 基础检索体只留 query/sort,from 与 size 交给分页执行器。
let size =
body.get("size").and_then(serde_json::Value::as_u64).unwrap_or(AUTO_PAGED_SELECT_STAR_SIZE as u64) as usize;
if let Some(map) = body.as_object_mut() {
map.remove("from");
map.remove("size");
}
let outcome = es_execute_paged_search(client, EsPageRequest { index, body, size }, page_cursor).await?;
let index_total = outcome.response.pointer("/hits/total/value").and_then(|v| v.as_u64());
let has_more = outcome.next_cursor.is_some();
let mut result = match parse_elasticsearch_response_with_sql_parser(
outcome.status,
outcome.response,
start,
sql_response_parser,
) {
Ok(result) => result,
Err(error) => {
es_release_unreturned_cursor(client, outcome.next_cursor.as_ref(), cursor.is_none()).await;
return Err(error);
}
};
if report_index_total {
if let Some(total) = index_total {
result.affected_rows = total;
}
}
// 翻到底后仍要把最后一个游标交回去,调用方据此释放 PIT。
result.session_id = outcome.next_cursor.or(outcome.active_cursor).or_else(|| cursor.map(str::to_string));
result.has_more = has_more;
Ok(result)
}
async fn execute_search_query(
client: &EsClient,
query: ElasticsearchSearchQuery,
start: std::time::Instant,
sql_response_parser: SqlResponseParser,
cursor: Option<&str>,
) -> Result<QueryResult, String> {
let search = EsIndexedSearch {
index: query.index,
body: query.body,
from_plan_pagination: query.from_plan_pagination,
report_index_total: query.from_plan_pagination,
};
execute_indexed_search(client, search, start, sql_response_parser, cursor).await
}
#[cfg(test)]
fn parse_elasticsearch_response(
status: u16,
body: serde_json::Value,
start: std::time::Instant,
) -> Result<QueryResult, String> {
parse_elasticsearch_response_with_sql_parser(status, body, start, parse_sql_response)
}
fn parse_elasticsearch_response_with_sql_parser(
status: u16,
mut body: serde_json::Value,
start: std::time::Instant,
sql_response_parser: SqlResponseParser,
) -> Result<QueryResult, String> {
if let Some(result) = sql_response_parser(&body, start) {
Ok(result)
} else if let Some(aggs) = body.get("aggregations").or_else(|| body.get("aggs")).and_then(|v| v.as_object()) {
let (columns, rows) = parse_aggregations(aggs);
if !columns.is_empty() {
let row_count = rows.len() as u64;
Ok(crate::types::QueryResult {
columns,
column_types: Vec::new(),
deferred_large_value_cells: Vec::new(),
column_sortables: vec![],
spatial_columns: vec![],
raw_key_cells: Vec::new(),
spatial_values: vec![],
rows,
affected_rows: row_count,
execution_time_ms: start.elapsed().as_millis(),
server_execute_time_us: None,
query_timings_ms: None,
truncated: false,
session_id: None,
has_more: false,
elasticsearch_raw_body: None,
messages: Vec::new(),
})
} else {
Ok(json_response_result(status, &body, start))
}
} else if let Some(hits) = body.pointer_mut("/hits/hits").and_then(serde_json::Value::as_array_mut) {
// Treat any `_search`-shaped body as the hits result, even when empty —
// a 0-row match is a valid empty result, not a reason to fall back to
// the raw-JSON status/response view.
let hits = std::mem::take(hits);
let (columns, column_types, rows) = parse_elasticsearch_search_hits(hits);
let row_count = rows.len() as u64;
Ok(crate::types::QueryResult {
columns,
column_types,
deferred_large_value_cells: Vec::new(),
column_sortables: vec![],
spatial_columns: vec![],
raw_key_cells: Vec::new(),
spatial_values: vec![],
rows,
affected_rows: row_count,
execution_time_ms: start.elapsed().as_millis(),
server_execute_time_us: None,
query_timings_ms: None,
truncated: false,
session_id: None,
has_more: false,
elasticsearch_raw_body: None,
messages: Vec::new(),
})
} else {
Ok(json_response_result(status, &body, start))
}
}
fn parse_elasticsearch_search_hits(
hits: Vec<serde_json::Value>,
) -> (Vec<String>, Vec<String>, Vec<Vec<serde_json::Value>>) {
let mut columns = Vec::<String>::new();
let mut column_indexes = HashMap::<String, usize>::new();
let mut json_column_indexes = HashSet::<usize>::new();
let mut rows = Vec::<Vec<serde_json::Value>>::with_capacity(hits.len());
for mut hit in hits {
let mut row = vec![serde_json::Value::Null; columns.len()];
if let Some(source) = hit.get_mut("_source").and_then(serde_json::Value::as_object_mut) {
for (key, value) in std::mem::take(source) {
append_elasticsearch_json_cell(
&mut columns,
&mut column_indexes,
&mut json_column_indexes,
&mut rows,
&mut row,
key,
value,
);
}
}
let id = hit.get_mut("_id").map(serde_json::Value::take).unwrap_or(serde_json::Value::Null);
append_elasticsearch_json_cell(
&mut columns,
&mut column_indexes,
&mut json_column_indexes,
&mut rows,
&mut row,
"_id".to_string(),
id,
);
if let Some(routing) = hit.get_mut("_routing") {
append_elasticsearch_json_cell(
&mut columns,
&mut column_indexes,
&mut json_column_indexes,
&mut rows,
&mut row,
"_routing".to_string(),
routing.take(),
);
}
rows.push(row);
}
if columns.is_empty() {
columns.push("_id".to_string());
}
let column_types = infer_elasticsearch_json_column_types(&rows, columns.len(), &json_column_indexes);
(columns, column_types, rows)
}
fn append_elasticsearch_json_cell(
columns: &mut Vec<String>,
column_indexes: &mut HashMap<String, usize>,
json_column_indexes: &mut HashSet<usize>,
previous_rows: &mut [Vec<serde_json::Value>],
row: &mut Vec<serde_json::Value>,
column: String,
value: serde_json::Value,
) {
let is_json_cell = matches!(value, serde_json::Value::Array(_) | serde_json::Value::Object(_));
let value = if is_json_cell { serde_json::Value::String(value.to_string()) } else { value };
if let Some(index) = column_indexes.get(&column).copied() {
if is_json_cell {
json_column_indexes.insert(index);
}
row[index] = value;
return;
}
let index = columns.len();
if is_json_cell {
json_column_indexes.insert(index);
}
column_indexes.insert(column.clone(), index);
columns.push(column);
for previous_row in previous_rows {
previous_row.push(serde_json::Value::Null);
}
row.push(value);
}
fn infer_elasticsearch_json_column_types(
rows: &[Vec<serde_json::Value>],
column_count: usize,
json_column_indexes: &HashSet<usize>,
) -> Vec<String> {
(0..column_count)
.map(|column_index| {
if json_column_indexes.contains(&column_index) {
return "json".to_string();
}
let mut inferred = None;
for value in rows.iter().filter_map(|row| row.get(column_index)) {
let value_type = match value {
serde_json::Value::Null => continue,
serde_json::Value::Bool(_) => "boolean",
serde_json::Value::Number(_) => "number",
serde_json::Value::String(_) => "text",
serde_json::Value::Array(_) | serde_json::Value::Object(_) => "json",
};
inferred = match inferred {
None => Some(value_type),
Some(existing) if existing == value_type => Some(existing),
Some(_) => Some("json"),
};
if inferred == Some("json") {
break;
}
}
inferred.unwrap_or("unknown").to_string()
})
.collect()
}
fn elasticsearch_rest_search_exceeds_table_limits(body: &serde_json::Value) -> bool {
let Some(hits) = body.pointer("/hits/hits").and_then(serde_json::Value::as_array) else {
return false;
};
if hits.len() > ELASTICSEARCH_REST_TABLE_MAX_ROWS {
return true;
}
let mut columns = HashSet::<&str>::new();
columns.insert("_id");
for hit in hits {
if hit.get("_routing").is_some() {
columns.insert("_routing");
}
if let Some(source) = hit.get("_source").and_then(serde_json::Value::as_object) {
columns.extend(source.keys().map(String::as_str));
}
if hits.len().saturating_mul(columns.len()) > ELASTICSEARCH_REST_TABLE_MAX_CELLS {
return true;
}
}
false
}
fn json_response_result(status: u16, body: &serde_json::Value, start: std::time::Instant) -> crate::types::QueryResult {
let body_text = serde_json::to_string_pretty(body).unwrap_or_else(|_| body.to_string());
raw_json_response_result(status, body_text, start)
}
fn raw_json_response_result(
status: u16,
body_text: impl Into<String>,
start: std::time::Instant,
) -> crate::types::QueryResult {
crate::types::QueryResult {
columns: vec!["status".to_string(), "response".to_string()],
column_types: Vec::new(),
deferred_large_value_cells: Vec::new(),
column_sortables: vec![],
spatial_columns: vec![],
raw_key_cells: Vec::new(),
spatial_values: vec![],
rows: vec![vec![serde_json::Value::Number(status.into()), serde_json::Value::String(body_text.into())]],
affected_rows: 0,
execution_time_ms: start.elapsed().as_millis(),
server_execute_time_us: None,
query_timings_ms: None,
truncated: false,
session_id: None,
has_more: false,
elasticsearch_raw_body: None,
messages: Vec::new(),
}
}
#[cfg(test)]
fn parse_elasticsearch_rest_response(
status: u16,
body_text: &str,
start: std::time::Instant,
) -> Result<QueryResult, String> {
parse_elasticsearch_rest_response_with_sql_parser(status, body_text, start, parse_sql_response)
}
fn parse_elasticsearch_rest_response_with_sql_parser(
status: u16,
body_text: &str,
start: std::time::Instant,
sql_response_parser: SqlResponseParser,
) -> Result<QueryResult, String> {
if body_text.trim().is_empty() {
return Ok(json_response_result(status, &serde_json::Value::Null, start));
}
if status >= 400 {
return Ok(raw_json_response_result(status, body_text, start));
}
if body_text.len() > ELASTICSEARCH_REST_TABLE_MAX_BODY_BYTES {
return Ok(raw_json_response_result(status, body_text, start));
}
if let Ok(body) = serde_json::from_str::<serde_json::Value>(body_text) {
if elasticsearch_rest_search_exceeds_table_limits(&body) {
return Ok(raw_json_response_result(status, body_text, start));
}
// Prefer a tabular view for search hits (_source columns), SQL API
// responses, and aggregations so the desktop data grid can display and
// copy rows like relational results. Other JSON (mapping, cluster
// info, …) stays as a lossless status/response panel with the raw body.
let mut result = parse_elasticsearch_response_with_sql_parser(status, body, start, sql_response_parser)?;
if result.columns == ["status".to_string(), "response".to_string()] {
return Ok(raw_json_response_result(status, body_text, start));
}
// Attach the raw response body so the UI can toggle between the
// table and the original JSON for Elasticsearch REST results.
result.elasticsearch_raw_body = Some(body_text.to_string());
return Ok(result);
}
// CAT APIs default to text/plain for human-readable output. Keep those
// responses visible instead of dropping them when JSON parsing is not valid.
let rows: Vec<Vec<serde_json::Value>> =
body_text.lines().map(|line| vec![serde_json::Value::String(line.to_string())]).collect();
let affected_rows = rows.len() as u64;
Ok(crate::types::QueryResult {
columns: vec!["response".to_string()],
column_types: Vec::new(),
deferred_large_value_cells: Vec::new(),
column_sortables: vec![],
spatial_columns: vec![],
raw_key_cells: Vec::new(),
spatial_values: vec![],
rows,
affected_rows,
execution_time_ms: start.elapsed().as_millis(),
server_execute_time_us: None,
query_timings_ms: None,
truncated: false,
session_id: None,
has_more: false,
elasticsearch_raw_body: None,
messages: Vec::new(),
})
}
fn parse_select_star_search_query(input: &str) -> Option<ElasticsearchSearchQuery> {
let mut cursor = skip_sql_whitespace(input, 0);
cursor = consume_sql_keyword(input, cursor, "select")?;
cursor = skip_sql_whitespace(input, cursor);
if next_char_at(input, cursor)? != '*' {
return None;
}
cursor += '*'.len_utf8();
cursor = skip_sql_whitespace(input, cursor);
cursor = consume_sql_keyword(input, cursor, "from")?;
cursor = skip_sql_whitespace(input, cursor);
let (index, next_cursor) = read_sql_token(input, cursor)?;
cursor = next_cursor;
let mut sort_field = None;
let mut sort_order = "asc";
let mut limit = None;
let mut offset: Option<usize> = None;
loop {
cursor = skip_sql_whitespace(input, cursor);
if cursor >= input.len() {
break;
}
if next_char_at(input, cursor) == Some(';') {
cursor += ';'.len_utf8();
cursor = skip_sql_whitespace(input, cursor);
if cursor == input.len() {
break;
}
return None;
}
if is_keyword_at(input, cursor, "order") {
cursor = consume_sql_keyword(input, cursor, "order")?;
cursor = skip_sql_whitespace(input, cursor);
cursor = consume_sql_keyword(input, cursor, "by")?;
cursor = skip_sql_whitespace(input, cursor);
let (field, next_cursor) = read_sql_token(input, cursor)?;
sort_field = Some(field);
cursor = skip_sql_whitespace(input, next_cursor);
if is_keyword_at(input, cursor, "asc") {
sort_order = "asc";
cursor = consume_sql_keyword(input, cursor, "asc")?;
} else if is_keyword_at(input, cursor, "desc") {
sort_order = "desc";
cursor = consume_sql_keyword(input, cursor, "desc")?;
}
} else if is_keyword_at(input, cursor, "limit") {
cursor = consume_sql_keyword(input, cursor, "limit")?;
cursor = skip_sql_whitespace(input, cursor);
let (value, next_cursor) = read_while(input, cursor, |ch| ch.is_ascii_digit());
limit = value.parse::<usize>().ok();
cursor = next_cursor;
} else if is_keyword_at(input, cursor, "offset") {
cursor = consume_sql_keyword(input, cursor, "offset")?;
cursor = skip_sql_whitespace(input, cursor);
let (value, next_cursor) = read_while(input, cursor, |ch| ch.is_ascii_digit());
offset = value.parse::<usize>().ok();
cursor = next_cursor;
} else {
return None;
}
}
// The pagination plan emits `LIMIT N OFFSET M` (always with OFFSET, even
// when 0) for ES; a user-written SQL that only has `LIMIT N` leaves
// OFFSET absent. We use that as the signal for whether the front-end is
// driving server-side pagination — in that case affected_rows must reflect
// the index's true total so the grid can compute the total page count.
let from_plan_pagination = offset.is_some();
let effective_size = limit.unwrap_or(AUTO_PAGED_SELECT_STAR_SIZE);
let effective_from = offset.unwrap_or(0);
let mut body = serde_json::Map::new();
body.insert("size".to_string(), serde_json::json!(effective_size));
if effective_from > 0 {
body.insert("from".to_string(), serde_json::json!(effective_from));
}
if let Some(field) = sort_field {
let mut sort_item = serde_json::Map::new();
sort_item.insert(field, serde_json::json!({ "order": sort_order }));
body.insert("sort".to_string(), serde_json::Value::Array(vec![serde_json::Value::Object(sort_item)]));
}
Some(ElasticsearchSearchQuery { index, body: serde_json::Value::Object(body), from_plan_pagination })
}
fn is_elasticsearch_sql_query(input: &str) -> bool {
input
.trim_start()
.split_once(char::is_whitespace)
.map(|(keyword, _)| keyword.eq_ignore_ascii_case("select"))
.unwrap_or_else(|| input.trim_start().eq_ignore_ascii_case("select"))
}
async fn execute_translated_select_star(
client: &EsClient,
translated: crate::db::elasticsearch_sql::TranslatedSelectStar,
start: std::time::Instant,
sql_response_parser: SqlResponseParser,
cursor: Option<&str>,
) -> Result<QueryResult, String> {
let search = EsIndexedSearch {
index: translated.index,
body: translated.body,
from_plan_pagination: translated.from_plan_pagination,
// 用户自己写了 LIMIT 时不覆盖行数,否则会把「取 10 条」显示成索引总量。
report_index_total: !translated.user_limited,
};
execute_indexed_search(client, search, start, sql_response_parser, cursor).await
}
/// Split a trailing `LIMIT n OFFSET m` from an ES SQL statement. The OFFSET
/// form is produced by the DBX pagination plan; it must be removed before
/// sending the query to `_sql` so ES SQL cursor pagination can drive paging.
/// A bare user `LIMIT n` (no OFFSET) is preserved as an explicit row cap.
fn es_sql_pagination(query: &str) -> (String, Option<usize>) {
let trimmed = query.trim().trim_end_matches(';').trim();
let re =
Regex::new(r"(?i)^(.*?)\s+limit\s+(\d+)(?:\s+offset\s+(\d+))?\s*$").expect("valid ES SQL pagination regex");
if let Some(caps) = re.captures(trimmed) {
let limit = caps.get(2).and_then(|value| value.as_str().parse::<usize>().ok());
let offset = caps.get(3).and_then(|value| value.as_str().parse::<usize>().ok());
// Only the plan's `OFFSET 0` first page is safe to strip. A
// user-written `OFFSET > 0` must keep its explicit offset semantics.
if offset == Some(0) {
let base = caps.get(1).map(|value| value.as_str().trim().to_string()).unwrap_or_default();
(base, limit)
} else {
(trimmed.to_string(), limit)
}
} else {
(trimmed.to_string(), None)
}
}
async fn execute_sql_query(
client: &EsClient,
query: &str,
start: std::time::Instant,
sql_response_parser: SqlResponseParser,
cursor: Option<&str>,
) -> Result<QueryResult, String> {
let (body, mut column_metadata) = if let Some(cursor) = cursor {
if let Ok(search_cursor) = decode_es_search_cursor(cursor) {
match search_cursor {
EsSearchCursor::Sql { cursor, columns_key, columns } => {
(serde_json::json!({ "cursor": cursor }), Some((columns_key, columns)))
}
EsSearchCursor::Page(page) => {
if let EsPageMode::Pit { pit_id, .. } = &page.mode {
let _ = close_es_pit(client, pit_id).await;
}
return Err("Elasticsearch PIT cursor cannot be used to continue a _sql query".to_string());
}
}
} else {
(serde_json::json!({ "cursor": cursor }), None)
}
} else {
// The pagination plan rewrites ES SQL to `LIMIT n OFFSET m`. Strip the
// OFFSET form so ES SQL cursor pagination can return exactly `n` rows
// per page and continue beyond index.max_result_window.
let (base_query, limit) = es_sql_pagination(query);
let query = adapt_elasticsearch_sql_query(&base_query);
let fetch_size = limit.unwrap_or(ES_SQL_FETCH_SIZE);
(serde_json::json!({ "query": query, "fetch_size": fetch_size }), None)
};
let resp = send_es_sql_request(client, "/_sql", &body).await?;
let status = client.response_status(&resp);
let mut response_body: serde_json::Value = resp.json().await.unwrap_or_else(|_| serde_json::Value::Null);
if !status.is_success() {
return Err(format_sql_error(status, &response_body));
}
if response_body.get("schema").is_none() && response_body.get("columns").is_none() {
if let (Some(body), Some((columns_key, columns))) = (response_body.as_object_mut(), column_metadata.as_ref()) {
body.insert(columns_key.clone(), serde_json::Value::Array(columns.clone()));
}
}
if column_metadata.is_none() {
column_metadata = sql_cursor_column_metadata(&response_body);
}
let mut result = match sql_response_parser(&response_body, start) {
Some(result) => result,
None => {
if let Some(raw_cursor) = response_body.get("cursor").and_then(serde_json::Value::as_str) {
let _ = close_es_sql_cursor(client, raw_cursor).await;
}
let pretty = serde_json::to_string_pretty(&response_body).unwrap_or_else(|_| response_body.to_string());
return Err(format!("Unexpected Elasticsearch SQL response: {pretty}"));
}
};
// Wrap the raw ES SQL cursor so close_cursor can distinguish it from a PIT
// cursor and close it with POST /_sql/close.
if let Some(raw_cursor) = result.session_id.take() {
let (columns_key, columns) =
column_metadata.ok_or_else(|| "Elasticsearch SQL response missing column metadata".to_string())?;
result.session_id =
Some(encode_es_search_cursor(&EsSearchCursor::Sql { cursor: raw_cursor, columns_key, columns })?);
} else if let Some(cursor) = cursor {
// On an exhausted continuation, keep returning the cursor that was used
// so the frontend can still close the SQL cursor during cleanup.
result.session_id = Some(cursor.to_string());
}
Ok(result)
}
fn sql_cursor_column_metadata(body: &serde_json::Value) -> Option<(String, Vec<serde_json::Value>)> {
["schema", "columns"].into_iter().find_map(|key| {
body.get(key).and_then(serde_json::Value::as_array).map(|columns| (key.to_string(), columns.clone()))
})
}
fn adapt_elasticsearch_sql_query(query: &str) -> String {
let mut output = String::with_capacity(query.len());
let mut index = 0;
let mut state = SqlScanState::Normal;
while let Some(ch) = next_char_at(query, index) {
match state {
SqlScanState::Normal => match ch {
'\'' => {
output.push(ch);
index += ch.len_utf8();
state = SqlScanState::SingleQuoted;
}
'"' => {
output.push(ch);
index += ch.len_utf8();
state = SqlScanState::DoubleQuoted;
}
'`' => {
output.push(ch);
index += ch.len_utf8();
state = SqlScanState::BacktickQuoted;
}
'-' if query[index..].starts_with("--") => {
output.push_str("--");
index += 2;
state = SqlScanState::LineComment;
}
'/' if query[index..].starts_with("/*") => {
output.push_str("/*");
index += 2;
state = SqlScanState::BlockComment;
}
'@' if is_at_identifier_boundary(&output) => {
let (identifier, next_index) = read_while(query, index, is_elasticsearch_identifier_part);
output.push('"');
output.push_str(identifier);
output.push('"');
index = next_index;
}
_ => {
if let Some(keyword) = relation_keyword_at(query, index) {
index = quote_relation_after_keyword(query, index, keyword, &mut output);
} else {
output.push(ch);
index += ch.len_utf8();
}
}
},
SqlScanState::SingleQuoted => {
if copy_quoted_char(query, &mut index, ch, '\'', &mut output) {
state = SqlScanState::Normal;
}
}
SqlScanState::DoubleQuoted => {
if copy_quoted_char(query, &mut index, ch, '"', &mut output) {
state = SqlScanState::Normal;
}
}
SqlScanState::BacktickQuoted => {
if copy_quoted_char(query, &mut index, ch, '`', &mut output) {
state = SqlScanState::Normal;
}
}
SqlScanState::LineComment => {
output.push(ch);
index += ch.len_utf8();
if ch == '\n' {
state = SqlScanState::Normal;
}
}
SqlScanState::BlockComment => {
if query[index..].starts_with("*/") {
output.push_str("*/");
index += 2;
state = SqlScanState::Normal;
} else {
output.push(ch);
index += ch.len_utf8();
}
}
}
}
output
}
fn quote_relation_after_keyword(query: &str, index: usize, keyword: &str, output: &mut String) -> usize {
let mut cursor = index + keyword.len();
output.push_str(&query[index..cursor]);
while let Some(ch) = next_char_at(query, cursor) {
if !ch.is_whitespace() {
break;
}
output.push(ch);
cursor += ch.len_utf8();
}
if matches!(next_char_at(query, cursor), Some('"' | '`' | '\'' | '(')) {
return cursor;
}
let relation_start = cursor;
while let Some(ch) = next_char_at(query, cursor) {
if !is_relation_name_char(ch) {
break;
}
cursor += ch.len_utf8();
}
let relation = &query[relation_start..cursor];
if relation_name_needs_quotes(relation) {
output.push('"');
output.push_str(relation);
output.push('"');
} else {
output.push_str(relation);
}
cursor
}
fn copy_quoted_char(query: &str, index: &mut usize, ch: char, quote: char, output: &mut String) -> bool {
output.push(ch);
*index += ch.len_utf8();
if ch != quote {
return false;
}
if next_char_at(query, *index).is_some_and(|next| next == quote) {
output.push(quote);
*index += quote.len_utf8();
false
} else {
true
}
}
fn read_while(query: &str, start: usize, predicate: fn(char) -> bool) -> (&str, usize) {
let mut cursor = start;
while let Some(ch) = next_char_at(query, cursor) {
if !predicate(ch) {
break;
}
cursor += ch.len_utf8();
}
(&query[start..cursor], cursor)
}
fn skip_sql_whitespace(query: &str, mut cursor: usize) -> usize {
while let Some(ch) = next_char_at(query, cursor) {
if !ch.is_whitespace() {
break;
}
cursor += ch.len_utf8();
}
cursor
}
fn consume_sql_keyword(query: &str, cursor: usize, keyword: &str) -> Option<usize> {
is_keyword_at(query, cursor, keyword).then_some(cursor + keyword.len())
}
fn read_sql_token(query: &str, cursor: usize) -> Option<(String, usize)> {
let quote = match next_char_at(query, cursor)? {
'"' => Some('"'),
'`' => Some('`'),
_ => None,
};
if let Some(quote) = quote {
let mut output = String::new();
let mut next_cursor = cursor + quote.len_utf8();
while let Some(ch) = next_char_at(query, next_cursor) {
next_cursor += ch.len_utf8();
if ch == quote {
if next_char_at(query, next_cursor).is_some_and(|next| next == quote) {
output.push(quote);
next_cursor += quote.len_utf8();
} else {
return Some((output, next_cursor));
}
} else {
output.push(ch);
}
}
return None;
}
let (token, next_cursor) = read_while(query, cursor, is_relation_name_char);
(!token.is_empty()).then(|| (token.to_string(), next_cursor))
}
fn relation_keyword_at(query: &str, index: usize) -> Option<&'static str> {
["from", "join"].into_iter().find(|keyword| is_keyword_at(query, index, keyword))
}
#[derive(Clone, Copy)]
enum SqlScanState {
Normal,
SingleQuoted,
DoubleQuoted,
BacktickQuoted,
LineComment,
BlockComment,
}
fn is_at_identifier_boundary(output: &str) -> bool {
output.chars().next_back().is_none_or(|ch| !is_sql_identifier_part(ch))
}
fn is_sql_identifier_part(ch: char) -> bool {
ch.is_ascii_alphanumeric() || matches!(ch, '_' | '.')
}
fn is_elasticsearch_identifier_part(ch: char) -> bool {
ch.is_ascii_alphanumeric() || matches!(ch, '_' | '.' | '-' | '@')
}
fn is_relation_name_char(ch: char) -> bool {
!ch.is_whitespace() && !matches!(ch, ',' | ';' | '(' | ')')
}
fn relation_name_needs_quotes(relation: &str) -> bool {
relation.chars().any(|ch| matches!(ch, '-' | '*' | '@'))
}
fn is_keyword_at(query: &str, index: usize, keyword: &str) -> bool {
query.get(index..index + keyword.len()).is_some_and(|candidate| candidate.eq_ignore_ascii_case(keyword))
&& query[..index].chars().next_back().is_none_or(|ch| !is_keyword_boundary_char(ch))
&& query[index + keyword.len()..].chars().next().is_none_or(|ch| !is_keyword_boundary_char(ch))
}
fn is_keyword_boundary_char(ch: char) -> bool {
ch.is_ascii_alphanumeric() || ch == '_'
}
fn next_char_at(query: &str, index: usize) -> Option<char> {
query.get(index..)?.chars().next()
}
fn parse_sql_response(body: &serde_json::Value, start: std::time::Instant) -> Option<QueryResult> {
parse_tabular_sql_response(body, start, "columns", "rows", None)
}
pub fn parse_tabular_sql_response(
body: &serde_json::Value,
start: std::time::Instant,
columns_key: &str,
rows_key: &str,
total_key: Option<&str>,
) -> Option<QueryResult> {
let columns = body.get(columns_key)?.as_array()?;
let rows = body.get(rows_key)?.as_array()?;
let column_names: Vec<String> = columns
.iter()
.filter_map(|column| column.get("name").and_then(|name| name.as_str()).map(str::to_string))
.collect();
if column_names.is_empty() && !columns.is_empty() {
return None;
}
let result_rows: Vec<Vec<serde_json::Value>> =
rows.iter().filter_map(|row| row.as_array().map(|values| values.to_vec())).collect();
Some(QueryResult {
columns: column_names,
column_types: Vec::new(),
deferred_large_value_cells: Vec::new(),
column_sortables: vec![],
spatial_columns: vec![],
raw_key_cells: Vec::new(),
spatial_values: vec![],
rows: result_rows,
affected_rows: total_key
.and_then(|key| body.get(key))
.and_then(serde_json::Value::as_u64)
.unwrap_or(rows.len() as u64),
execution_time_ms: start.elapsed().as_millis(),
server_execute_time_us: None,
query_timings_ms: None,
truncated: false,
session_id: body.get("cursor").and_then(|cursor| cursor.as_str()).map(str::to_string),
has_more: body.get("cursor").and_then(|cursor| cursor.as_str()).is_some(),
elasticsearch_raw_body: None,
messages: Vec::new(),
})
}
fn format_sql_error(status: reqwest::StatusCode, body: &serde_json::Value) -> String {
let detail = body
.pointer("/error/reason")
.and_then(|reason| reason.as_str())
.map(str::to_string)
.unwrap_or_else(|| serde_json::to_string_pretty(body).unwrap_or_else(|_| body.to_string()));
if status == reqwest::StatusCode::NOT_FOUND {
format!("Elasticsearch SQL API is not available ({status}): {detail}")
} else {
format!("Elasticsearch SQL error ({status}): {detail}")
}
}
fn parse_aggregations(aggs: &serde_json::Map<String, serde_json::Value>) -> (Vec<String>, Vec<Vec<serde_json::Value>>) {
for (_name, agg_value) in aggs {
if let Some(buckets) = agg_value.get("buckets").and_then(|b| b.as_array()) {
if buckets.is_empty() {
continue;
}
let mut all_keys = Vec::<String>::new();
let mut bucket_rows = Vec::new();
for bucket in buckets {
if let Some(obj) = bucket.as_object() {
let mut row = serde_json::Map::new();
for (k, v) in obj {
if let Some(sub) = v.as_object() {
if let Some(val) = sub.get("value") {
row.insert(k.clone(), val.clone());
} else {
row.insert(k.clone(), serde_json::Value::String(v.to_string()));
}
} else {
row.insert(k.clone(), v.clone());
}
}
for key in row.keys() {
if !all_keys.contains(key) {
all_keys.push(key.clone());
}
}
bucket_rows.push(row);
}
}
let rows = bucket_rows
.iter()
.map(|br| {
all_keys
.iter()
.map(|k| {
br.get(k)
.map(|v| match v {
serde_json::Value::String(s) => serde_json::Value::String(s.clone()),
other => serde_json::Value::String(other.to_string()),
})
.unwrap_or(serde_json::Value::Null)
})
.collect()
})
.collect();
return (all_keys, rows);
}
}
let mut columns = Vec::new();
let mut values = Vec::new();
for (name, agg_value) in aggs {
if let Some(obj) = agg_value.as_object() {
if let Some(val) = obj.get("value") {
columns.push(name.clone());
values.push(match val {
serde_json::Value::String(s) => serde_json::Value::String(s.clone()),
other => serde_json::Value::String(other.to_string()),
});
}
}
}
if !columns.is_empty() {
return (columns, vec![values]);
}
(Vec::new(), Vec::new())
}
#[cfg(test)]
mod tests {
use super::{
build_count_documents_body, build_find_documents_body, elasticsearch_accept_invalid_certs,
elasticsearch_base_url_fallbacks, elasticsearch_index_grouping, group_index_names, merge_index_entries,
normalize_index_names, redact_elasticsearch_url, ElasticsearchIndexEntry, EsClient, SearchResponse,
};
use serde_json::json;
use std::time::Duration;
async fn read_http_request(socket: &mut tokio::net::TcpStream) -> String {
use tokio::io::AsyncReadExt;
let mut bytes = Vec::new();
let mut buffer = [0_u8; 1024];
loop {
let read = socket.read(&mut buffer).await.unwrap();
assert!(read > 0, "HTTP request ended before its body was received");
bytes.extend_from_slice(&buffer[..read]);
let Some(headers_end) = bytes.windows(4).position(|window| window == b"\r\n\r\n") else {
continue;
};
let content_length = std::str::from_utf8(&bytes[..headers_end])
.unwrap()
.lines()
.find_map(|line| {
let (name, value) = line.split_once(':')?;
name.eq_ignore_ascii_case("content-length").then(|| value.trim().parse::<usize>().unwrap())
})
.unwrap_or(0);
let request_end = headers_end + 4 + content_length;
if bytes.len() >= request_end {
bytes.truncate(request_end);
return String::from_utf8(bytes).unwrap();
}
}
}
#[test]
fn parses_rest_request_after_comments_and_preserves_query_parameters() {
let request = super::parse_elasticsearch_rest_request(
"# JVM statistics\n// available on Elasticsearch 7+\nGET /_nodes/stats/jvm?pretty",
)
.unwrap();
assert_eq!(request.method, reqwest::Method::GET);
assert_eq!(request.path, "/_nodes/stats/jvm?pretty");
assert_eq!(request.body, None);
assert_eq!(request.body_kind, super::ElasticsearchRestBodyKind::Json);
}
#[test]
fn normalize_index_names_filters_sorts_and_dedups() {
let names = normalize_index_names(
[".kibana", "ngx-log-2", "ngx-log-1", "ngx-log-1", ".security-7"].into_iter().map(String::from),
);
// 去掉点前缀内部索引、排序、去重。
assert_eq!(names, vec!["ngx-log-1".to_string(), "ngx-log-2".to_string()]);
}
#[test]
fn mapping_columns_include_implicit_object_parents() {
let mapping = json!({
"profile": {
"properties": {
"name": { "type": "keyword" }
}
}
});
let properties = mapping.as_object().unwrap();
let mut seen = std::collections::HashSet::new();
let mut columns = Vec::new();
super::collect_mapping_columns("", properties, &mut seen, &mut columns);
assert!(columns.iter().any(|column| column.name == "profile" && column.data_type == "object"));
assert!(columns.iter().any(|column| column.name == "profile.name" && column.data_type == "keyword"));
}
#[test]
fn mapping_columns_do_not_overflow_on_deep_nested_objects() {
let mut current = json!({ "type": "keyword" });
for depth in 0..512 {
current = json!({ "properties": { format!("n{depth}"): current } });
}
let properties = current.get("properties").and_then(|value| value.as_object()).unwrap();
let mut seen = std::collections::HashSet::new();
let mut columns = Vec::new();
super::collect_mapping_columns("", properties, &mut seen, &mut columns);
assert!(columns.iter().any(|column| column.name == "n511" && column.data_type == "object"));
assert_eq!(columns.len(), super::ELASTICSEARCH_MAPPING_MAX_DEPTH as usize);
super::drop_json_iteratively(current);
}
#[test]
fn connectivity_check_disabled_parses_bool_and_string() {
use super::elasticsearch_connectivity_check_disabled as disabled;
assert!(!disabled(None));
assert!(!disabled(Some(&serde_json::json!({}))));
assert!(disabled(Some(&serde_json::json!({ "connectivityCheckDisabled": true }))));
assert!(disabled(Some(&serde_json::json!({ "connectivityCheckDisabled": "on" }))));
assert!(disabled(Some(&serde_json::json!({ "connectivityCheckDisabled": "TRUE" }))));
assert!(!disabled(Some(&serde_json::json!({ "connectivityCheckDisabled": false }))));
assert!(!disabled(Some(&serde_json::json!({ "connectivityCheckDisabled": "off" }))));
}
#[test]
fn group_index_names_off_by_default() {
// 缺省配置 → 关闭聚合,原样返回。
assert!(elasticsearch_index_grouping(None).is_none());
let raw = vec!["a-2026.08.04".to_string(), "a-2026.08.05".to_string()];
assert_eq!(group_index_names(raw.clone(), None), raw);
}
#[test]
fn group_index_names_tenant_level_with_capture_group() {
// 保留 `@<第一段>`(到第一个下划线),其后全部折叠成 `*`。用中性占位名。
let cfg = serde_json::json!({ "indexGroupingPattern": r"^([^@]*@[^_]+)_.*$" });
let re = elasticsearch_index_grouping(Some(&cfg));
let out = group_index_names(
vec![
"svc@alpha_r1-2026.08.06@0-000001".to_string(),
"svc@alpha_r1-2026.08.07@0-000001".to_string(),
"svc@beta_r1-2026.08.06@0-000001".to_string(),
"svc_err-2026.08.06@0-000001".to_string(), // 无 @ → 不匹配 → 原样
],
re.as_ref(),
);
assert_eq!(
out,
vec!["svc@alpha*".to_string(), "svc@beta*".to_string(), "svc_err-2026.08.06@0-000001".to_string(),]
);
}
#[test]
fn group_index_names_tail_strip_without_capture_group() {
// 无捕获组:按 ES 惯例剥掉“日期+滚动号”尾巴,`${1}` 为空即替换成 `*`。
let cfg = serde_json::json!({ "indexGroupingPattern": r"[-_.@]\d{4}[-_.]?\d{2}[-_.]?\d{2}.*$" });
let re = elasticsearch_index_grouping(Some(&cfg));
let out = group_index_names(
vec!["logs-2026.08.06".to_string(), "logs-2026.08.07".to_string(), "orders-2026.08.01".to_string()],
re.as_ref(),
);
assert_eq!(out, vec!["logs*".to_string(), "orders*".to_string()]);
}
#[test]
fn group_index_names_mixed_tenant_and_plain_scheme() {
// 同一条正则同时处理:真租户(@后跟字母)折到 @租户;无租户的按名字剥日期尾巴;
// 普通非时间序列索引保持原样。区分点:真租户 @ 后是字母,滚动号 @ 后是数字。用中性名。
let cfg = serde_json::json!({ "indexGroupingPattern": r"^([^@]*@[a-zA-Z][a-zA-Z0-9]*|[^-@]*)[-_.@].*$" });
let re = elasticsearch_index_grouping(Some(&cfg));
let out = group_index_names(
vec![
"svc_err-2026.08.10@0-000001".to_string(), // 无租户 → svc_err*
"svc_err-2026.08.11@0-000001".to_string(),
"svc@alpha_r1-2026.08.10@0-000001".to_string(), // 带区域租户 → svc@alpha*
"[email protected]@0-000001".to_string(), // 无区域租户 → svc@beta*
"catalog".to_string(), // 普通索引无尾巴 → 原样
],
re.as_ref(),
);
assert_eq!(
out,
vec!["catalog".to_string(), "svc@alpha*".to_string(), "svc@beta*".to_string(), "svc_err*".to_string(),]
);
}
fn entry(name: &str, aliases: &[&str]) -> ElasticsearchIndexEntry {
ElasticsearchIndexEntry {
name: name.to_string(),
aliases: aliases.iter().map(|alias| alias.to_string()).collect(),
}
}
fn alias_map(pairs: &[(&str, &[&str])]) -> std::collections::BTreeMap<String, Vec<String>> {
pairs
.iter()
.map(|(index, aliases)| (index.to_string(), aliases.iter().map(|alias| alias.to_string()).collect()))
.collect()
}
#[test]
fn merge_index_entries_attaches_aliases_to_the_same_index_row() {
let entries = merge_index_entries(
vec!["orders".to_string(), "users".to_string()],
&alias_map(&[("orders", &["orders-write", "orders"]), ("users", &[".hidden"])]),
None,
);
assert_eq!(entries, vec![entry("orders", &["orders-write"]), entry("users", &[])]);
}
#[test]
fn merge_index_entries_collects_grouped_index_aliases_onto_the_pattern_row() {
let cfg = serde_json::json!({ "indexGroupingPattern": r"[-_.@]\d{4}[-_.]?\d{2}[-_.]?\d{2}.*$" });
let re = elasticsearch_index_grouping(Some(&cfg));
let entries = merge_index_entries(
vec!["logs-2026.08.06".to_string(), "logs-2026.08.07".to_string()],
&alias_map(&[("logs-2026.08.06", &["logs"]), ("logs-2026.08.07", &["logs", "logs-write"])]),
re.as_ref(),
);
assert_eq!(entries, vec![entry("logs*", &["logs", "logs-write"])]);
}
async fn write_json_http_response(socket: &mut tokio::net::TcpStream, status: u16, body: &str) {
use tokio::io::AsyncWriteExt;
let reason = match status {
200 => "OK",
401 => "Unauthorized",
403 => "Forbidden",
_ => "Error",
};
let response = format!(
"HTTP/1.1 {status} {reason}\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{body}",
body.len()
);
socket.write_all(response.as_bytes()).await.unwrap();
}
async fn serve_elasticsearch_json_routes(
listener: tokio::net::TcpListener,
routes: Vec<(&'static str, u16, &'static str)>,
) {
for _ in 0..routes.len() {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
let (status, body) = routes
.iter()
.find(|(prefix, _, _)| {
let encoded = prefix.replace('*', "%2A");
request.starts_with(&format!("GET {prefix}")) || request.starts_with(&format!("GET {encoded}"))
})
.map(|(_, status, body)| (*status, *body))
.unwrap_or((404, "{}"));
write_json_http_response(&mut socket, status, body).await;
}
}
#[tokio::test]
async fn list_indices_lists_aliases_from_existing_alias_endpoint() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(serve_elasticsearch_json_routes(
listener,
vec![
("/_cat/indices", 200, r#"[{"index":"orders"},{"index":".security"}]"#),
(
"/_alias",
200,
r#"{"orders":{"aliases":{"orders-write":{},".kibana":{}}},"hidden":{"aliases":{"orders-write":{}}}}"#,
),
],
));
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(2));
let entries = super::list_indices_with_aliases(&client).await.unwrap();
server.await.unwrap();
assert_eq!(entries, vec![entry("orders", &["orders-write"])]);
}
#[tokio::test]
async fn list_indices_keeps_indices_when_alias_endpoint_is_forbidden() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(serve_elasticsearch_json_routes(
listener,
vec![("/_cat/indices", 200, r#"[{"index":"orders"}]"#), ("/_alias", 403, r#"{"error":"forbidden"}"#)],
));
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(2));
let entries = super::list_indices_with_aliases(&client).await.unwrap();
server.await.unwrap();
assert_eq!(entries, vec![entry("orders", &[])]);
}
#[tokio::test]
async fn list_indices_reads_aliases_from_resolve_index_fallback() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(serve_elasticsearch_json_routes(
listener,
vec![
("/_cat/indices", 403, r#"{"error":"forbidden"}"#),
(
"/_resolve/index/*",
200,
r#"{"indices":[{"name":"orders"}],"aliases":[{"name":"orders-write","indices":["orders"]}],"data_streams":[{"name":"logs"}]}"#,
),
],
));
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(2));
let entries = super::list_indices_with_aliases(&client).await.unwrap();
server.await.unwrap();
assert_eq!(entries, vec![entry("logs", &[]), entry("orders", &["orders-write"])]);
}
#[tokio::test]
async fn list_indices_reads_nested_aliases_from_alias_endpoint_fallback() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(serve_elasticsearch_json_routes(
listener,
vec![
("/_cat/indices", 403, r#"{"error":"forbidden"}"#),
("/_resolve/index/*", 403, r#"{"error":"forbidden"}"#),
("/_alias", 200, r#"{"orders":{"aliases":{"orders-write":{}}},"users":{"aliases":{}}}"#),
],
));
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(2));
let entries = super::list_indices_with_aliases(&client).await.unwrap();
server.await.unwrap();
assert_eq!(entries, vec![entry("orders", &["orders-write"]), entry("users", &[])]);
}
#[test]
fn parses_rest_request_after_multiline_block_comment() {
let request = super::parse_elasticsearch_rest_request(
"/* node statistics\n safe on supported clusters */\nGET /_nodes/stats/jvm?pretty",
)
.unwrap();
assert_eq!(request.method, reqwest::Method::GET);
assert_eq!(request.path, "/_nodes/stats/jvm?pretty");
}
#[test]
fn parses_lowercase_rest_method() {
let request = super::parse_elasticsearch_rest_request("get /_cluster/health").unwrap();
assert_eq!(request.method, reqwest::Method::GET);
assert_eq!(request.path, "/_cluster/health");
}
#[test]
fn encodes_raw_date_math_paths_without_double_encoding() {
let raw = super::parse_elasticsearch_rest_request("GET /<logs-{now/d}>/_search?pretty").unwrap();
assert_eq!(raw.path, "/%3Clogs-%7Bnow%2Fd%7D%3E/_search?pretty");
let encoded = super::parse_elasticsearch_rest_request("GET /%3Clogs-%7Bnow%2Fd%7D%3E/_search?pretty").unwrap();
assert_eq!(encoded.path, "/%3Clogs-%7Bnow%2Fd%7D%3E/_search?pretty");
}
#[test]
fn detects_ndjson_endpoints_with_index_and_query_parameters() {
let request = super::parse_elasticsearch_rest_request(
"POST /orders/_bulk?refresh=true\n{\"index\":{\"_id\":\"1\"}}\n{\"name\":\"Notebook\"}",
)
.unwrap();
assert_eq!(request.method, reqwest::Method::POST);
assert_eq!(request.body_kind, super::ElasticsearchRestBodyKind::Ndjson);
assert!(request.body.is_some());
}
#[test]
fn normalizes_ndjson_with_a_required_trailing_newline() {
let body = "{\"index\":{}}\n{\"name\":\"Notebook\"}";
assert_eq!(super::validate_elasticsearch_ndjson(body).unwrap(), format!("{body}\n"));
assert!(super::validate_elasticsearch_ndjson("{\"index\":{}}\nnot-json").is_err());
}
#[test]
fn url_params_can_disable_elasticsearch_tls_verification() {
assert!(elasticsearch_accept_invalid_certs(false, Some("sslmode=disable")));
assert!(elasticsearch_accept_invalid_certs(false, Some("?tlsVerify=false")));
assert!(elasticsearch_accept_invalid_certs(false, Some("verify=0")));
assert!(elasticsearch_accept_invalid_certs(false, Some("insecure=true")));
assert!(elasticsearch_accept_invalid_certs(false, Some("accept_invalid_certs=on")));
assert!(!elasticsearch_accept_invalid_certs(false, Some("sslmode=require&verify=true")));
}
#[test]
fn tls_checkbox_keeps_legacy_insecure_elasticsearch_behavior() {
assert!(elasticsearch_accept_invalid_certs(true, None));
}
#[test]
fn localhost_elasticsearch_url_falls_back_to_ipv4_loopback() {
assert_eq!(
elasticsearch_base_url_fallbacks("https://localhost:9200"),
vec!["https://127.0.0.1:9200".to_string()]
);
assert_eq!(elasticsearch_base_url_fallbacks("https://search.example.com:9200"), Vec::<String>::new());
}
#[test]
fn disabled_connectivity_check_keeps_localhost_request_fallback() {
assert_eq!(
super::elasticsearch_localhost_resolve_addrs("https://localhost:9200", true),
Some(["[::1]:0".parse().unwrap(), "127.0.0.1:0".parse().unwrap()])
);
assert_eq!(super::elasticsearch_localhost_resolve_addrs("https://localhost:9200", false), None);
assert_eq!(super::elasticsearch_localhost_resolve_addrs("https://search.example.com:9200", true), None);
}
#[tokio::test]
async fn disabled_connectivity_check_can_request_ipv4_localhost() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("GET /_cluster/health "), "unexpected request: {request}");
let body = r#"{"status":"green"}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let mut client = EsClient::from_config(
&format!("http://localhost:{}", addr.port()),
None,
None,
false,
None,
Some(&json!({ "connectivityCheckDisabled": "yes" })),
Duration::from_secs(2),
None,
None,
None,
)
.expect("failed to build Elasticsearch test client");
super::test_connection(&mut client, Duration::from_secs(2)).await.unwrap();
let response = client.get("/_cluster/health").send().await.unwrap();
assert!(response.status().is_success());
server.await.unwrap();
}
#[tokio::test]
async fn api_key_authenticates_connectivity_and_query_requests_in_both_modes() {
for mode in ["direct", "kibana"] {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
for _ in 0..2 {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
let auth = request
.lines()
.filter_map(|line| line.split_once(':'))
.find(|(name, _)| name.eq_ignore_ascii_case("authorization"))
.map(|(_, value)| value.trim());
assert_eq!(auth, Some("ApiKey c3ludGhldGljOmtleQ=="));
if mode == "kibana" {
assert!(request.starts_with("POST /api/console/proxy?"));
assert!(request.contains("kbn-xsrf: true"));
}
write_json_http_response(&mut socket, 200, "{}").await;
}
});
let config = json!({ "mode": mode, "authType": "apiKey" });
let mut client = EsClient::from_config(
&format!("http://{addr}"),
Some("ignored-user"),
Some(" c3ludGhldGljOmtleQ== "),
false,
None,
Some(&config),
Duration::from_secs(2),
None,
None,
None,
)
.unwrap();
super::test_connection(&mut client, Duration::from_secs(2)).await.unwrap();
let cloned = client.clone();
super::execute_rest_query(&cloned, "GET /_cluster/health").await.unwrap();
server.await.unwrap();
}
}
#[test]
fn api_key_mode_rejects_missing_or_invalid_keys_without_echoing_the_credential() {
for key in [None, Some(" "), Some("synthetic-secret\r\nX-Injected: yes")] {
let result = EsClient::from_config(
"http://localhost:9200",
None,
key,
false,
None,
Some(&json!({ "authType": "apiKey" })),
Duration::from_secs(1),
None,
None,
None,
);
let error = result.err().expect("invalid API key should be rejected");
assert!(!error.contains("synthetic-secret"));
}
}
#[test]
fn existing_connections_keep_basic_or_anonymous_authentication() {
for (username, expected) in [(Some("elastic"), Some("Basic ZWxhc3RpYzpzZWNyZXQ=")), (None, None)] {
let client = EsClient::from_config(
"http://localhost:9200",
username,
Some("secret"),
false,
None,
None,
Duration::from_secs(1),
None,
None,
None,
)
.unwrap();
let request = client.get("/").build().unwrap();
assert_eq!(
request.headers().get(reqwest::header::AUTHORIZATION).map(|value| value.to_str().unwrap()),
expected
);
}
}
#[test]
fn elasticsearch_client_from_config_uses_url_params_for_tls_verification() {
let client = EsClient::from_config(
"https://localhost:9200/",
Some("elastic"),
Some("secret"),
false,
Some("sslmode=disable"),
None,
Duration::from_secs(1),
None,
None,
None,
)
.expect("failed to build Elasticsearch test client");
assert_eq!(client.base_url, "https://localhost:9200");
assert_eq!(client.fallback_base_urls, vec!["https://127.0.0.1:9200"]);
assert_eq!(client.connectivity_check_path, "/");
}
#[test]
fn connectivity_check_path_normalizes_get_path_and_defaults() {
assert_eq!(super::elasticsearch_connectivity_check_path(None), "/");
assert_eq!(super::elasticsearch_connectivity_check_path(Some(&json!({ "connectivityCheckPath": "" }))), "/");
assert_eq!(
super::elasticsearch_connectivity_check_path(Some(&json!({
"connectivityCheckPath": "GET pro-jmsau-nwm-applog-*/_search"
}))),
"/pro-jmsau-nwm-applog-*/_search"
);
assert_eq!(
super::elasticsearch_connectivity_check_path(Some(&json!({
"connectivityCheckPath": "my-index/_search\n{\"query\":{\"match_all\":{}}}"
}))),
"/my-index/_search"
);
let client = EsClient::from_config(
"https://localhost:5601/",
None,
None,
false,
None,
Some(&json!({
"mode": "kibana",
"connectivityCheckPath": "GET pro-logs-*/_search"
})),
Duration::from_secs(1),
None,
None,
None,
)
.expect("failed to build Elasticsearch test client");
assert_eq!(client.connectivity_check_path, "/pro-logs-*/_search");
}
#[tokio::test]
async fn test_connection_uses_configured_connectivity_check_path() {
use tokio::io::{AsyncReadExt, AsyncWriteExt};
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let mut request = [0_u8; 2048];
let read = socket.read(&mut request).await.unwrap();
let request = String::from_utf8_lossy(&request[..read]);
assert!(request.starts_with("GET /pro-logs-*/_search "), "unexpected request: {request}");
let body = r#"{"hits":{"total":{"value":0,"relation":"eq"},"hits":[]}}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let mut client = EsClient::from_config(
&format!("http://{addr}"),
None,
None,
false,
None,
Some(&json!({ "connectivityCheckPath": "/pro-logs-*/_search" })),
Duration::from_secs(2),
None,
None,
None,
)
.expect("failed to build Elasticsearch test client");
super::test_connection(&mut client, Duration::from_secs(2)).await.unwrap();
server.await.unwrap();
}
#[test]
fn elasticsearch_client_from_config_enables_kibana_proxy_with_base_path() {
let external_config = json!({ "mode": "kibana", "kibanaBasePath": "/kibana/s/analytics/" });
let client = EsClient::from_config(
"https://localhost:5601/",
Some("elastic"),
Some("secret"),
false,
None,
Some(&external_config),
Duration::from_secs(1),
None,
None,
None,
)
.expect("failed to build Elasticsearch test client");
assert_eq!(client.base_url, "https://localhost:5601/kibana/s/analytics");
assert_eq!(client.fallback_base_urls, vec!["https://127.0.0.1:5601/kibana/s/analytics"]);
assert_eq!(client.transport_mode, super::ElasticsearchTransportMode::KibanaProxy);
}
#[test]
fn redacts_elasticsearch_url_credentials_in_errors() {
assert_eq!(
redact_elasticsearch_url("https://elastic:secret@localhost:9200"),
"https://user:password@localhost:9200"
);
}
#[test]
fn encodes_elasticsearch_index_path_segments() {
assert_eq!(super::elasticsearch_index_path("%kuzzle.users", "_search"), "/%25kuzzle.users/_search");
assert_eq!(super::elasticsearch_index_path("logs-*", "_search"), "/logs-*/_search");
assert_eq!(super::elasticsearch_index_path("logs/2026", "_mapping"), "/logs%2F2026/_mapping");
}
#[test]
fn encodes_elasticsearch_document_id_path_segment() {
assert_eq!(super::elasticsearch_path_segment("a%b/c"), "a%25b%2Fc");
}
#[test]
fn builds_elasticsearch_document_path_with_type_and_routing() {
assert_eq!(
super::elasticsearch_document_path("orders/2026", "a%b/c", Some("legacy/order"), Some("tenant/a&b")),
"/orders%2F2026/legacy%2Forder/a%25b%2Fc?routing=tenant%2Fa%26b&refresh=true"
);
assert_eq!(
super::elasticsearch_document_path("orders", "1", Some("_doc"), None),
"/orders/_doc/1?refresh=true"
);
assert_eq!(super::elasticsearch_document_path("orders", "1", None, None), "/orders/_doc/1?refresh=true");
}
#[test]
fn builds_elasticsearch_auto_id_document_path_with_routing() {
assert_eq!(
super::elasticsearch_auto_id_document_path("orders/2026", Some("tenant/a&b")),
"/orders%2F2026/_doc?routing=tenant%2Fa%26b&refresh=true"
);
assert_eq!(super::elasticsearch_auto_id_document_path("orders", None), "/orders/_doc?refresh=true");
}
#[test]
fn insert_document_body_extracts_routing_without_embedding_it() {
let (doc, routing) =
super::elasticsearch_document_body_and_routing_from_json(r#"{"_routing":"tenant-1","name":"Alice"}"#, None)
.expect("parse insert body");
assert_eq!(routing.as_deref(), Some("tenant-1"));
assert_eq!(doc, serde_json::json!({"name":"Alice"}));
assert_eq!(
super::elasticsearch_auto_id_document_path("orders", routing.as_deref()),
"/orders/_doc?routing=tenant-1&refresh=true"
);
}
#[test]
fn elasticsearch_sql_detection_does_not_treat_rest_methods_as_sql() {
assert!(super::is_elasticsearch_sql_query("SELECT * FROM index_task_v1"));
assert!(super::is_elasticsearch_sql_query(" select count(*) from index_task_v1"));
assert!(!super::is_elasticsearch_sql_query("GET /index_task_v1/_mapping"));
assert!(!super::is_elasticsearch_sql_query("POST /index_task_v1/_search\n{}"));
assert!(!super::is_elasticsearch_sql_query("DELETE /index_task_v1/_doc/1"));
}
#[test]
fn builds_elasticsearch_find_body_with_filter_and_sort() {
let body = build_find_documents_body(20, 10, Some(r#"{"city":"长治"}"#), Some(r#"{"created_at":-1}"#)).unwrap();
assert_eq!(
body,
json!({
"from": 20,
"size": 10,
"query": { "term": { "city": "长治" } },
"sort": [{ "created_at": { "order": "desc" } }]
})
);
assert!(body.get("track_total_hits").is_none());
}
#[test]
fn pit_sort_tiebreaker_replaces_plain_doc_sort() {
// 无排序时文档链路给出的是字符串形式的 "_doc",必须被替换而不是叠加。
let mut items = vec![serde_json::json!("_doc")];
super::elasticsearch_apply_pit_sort_tiebreaker(&mut items);
assert_eq!(items, vec![serde_json::json!({ "_shard_doc": "asc" })]);
let mut items = Vec::new();
super::elasticsearch_apply_pit_sort_tiebreaker(&mut items);
assert_eq!(items, vec![serde_json::json!({ "_shard_doc": "asc" })]);
let mut items = vec![serde_json::json!({ "created_at": { "order": "desc" } })];
super::elasticsearch_apply_pit_sort_tiebreaker(&mut items);
assert_eq!(
items,
vec![serde_json::json!({ "created_at": { "order": "desc" } }), serde_json::json!({ "_shard_doc": "asc" })]
);
}
#[test]
fn detects_only_missing_pit_sort_errors() {
assert!(super::elasticsearch_error_lacks_pit_sort(
r#"Elasticsearch error: {"error":{"root_cause":[{"type":"query_shard_exception","reason":"No mapping found for [_shard_doc] in order to sort on","index":"products"}]},"status":400}"#
));
// 偶发故障不该被当成能力缺失。
assert!(!super::elasticsearch_error_lacks_pit_sort("Elasticsearch request failed: operation timed out"));
assert!(!super::elasticsearch_error_lacks_pit_sort(
r#"Elasticsearch error: {"error":{"type":"search_phase_execution_exception"},"status":503}"#
));
}
#[test]
fn pit_page_body_includes_pit_and_search_after() {
let page = super::EsPageCursor {
index: "logs".to_string(),
body: json!({
"query": { "term": { "city": "长治" } },
"sort": [{ "created_at": { "order": "desc" } }],
}),
size: 25,
mode: super::EsPageMode::Pit {
pit_id: "pit-1".to_string(),
keep_alive: "1m".to_string(),
search_after: vec![json!("abc"), json!(123)],
},
};
assert_eq!(
super::es_build_page_request_body(&page),
json!({
"size": 25,
"track_total_hits": true,
"query": { "term": { "city": "长治" } },
"sort": [{ "created_at": { "order": "desc" } }, { "_shard_doc": "asc" }],
"pit": { "id": "pit-1", "keep_alive": "1m" },
"search_after": ["abc", 123]
})
);
}
#[test]
fn offset_page_body_uses_from_and_never_adds_shard_doc() {
let page = super::EsPageCursor {
index: "logs".to_string(),
body: json!({ "sort": ["_doc"], "query": { "match_all": {} } }),
size: 25,
mode: super::EsPageMode::Offset { from: 50 },
};
// 退回 from/size 时绝不能出现 `_shard_doc`。
assert_eq!(
super::es_build_page_request_body(&page),
// 回退路径刻意不带 track_total_hits:那会让每页都对整个索引做精确计数。
json!({
"size": 25,
"from": 50,
"sort": ["_doc"],
"query": { "match_all": {} }
})
);
}
#[test]
fn scroll_page_body_removes_offset_for_legacy_clusters() {
let page = super::EsPageCursor {
index: "events".to_string(),
body: json!({ "from": 9_500, "query": { "match_all": {} }, "sort": ["created_at"] }),
size: 500,
mode: super::EsPageMode::Scroll { scroll_id: String::new(), keep_alive: "1m".to_string() },
};
assert_eq!(
super::es_build_page_request_body(&page),
json!({ "query": { "match_all": {} }, "size": 500, "sort": ["created_at"] })
);
}
#[test]
fn es_sql_pagination_strips_plan_offset_but_keeps_user_limit() {
let (base, limit) = super::es_sql_pagination("SELECT field FROM idx LIMIT 500 OFFSET 0");
assert_eq!(base, "SELECT field FROM idx");
assert_eq!(limit, Some(500));
let (base, limit) = super::es_sql_pagination("SELECT field FROM idx LIMIT 10");
assert_eq!(base, "SELECT field FROM idx LIMIT 10");
assert_eq!(limit, Some(10));
let (base, limit) = super::es_sql_pagination("SELECT field FROM idx LIMIT 500 OFFSET 100");
assert_eq!(base, "SELECT field FROM idx LIMIT 500 OFFSET 100");
assert_eq!(limit, Some(500));
let (base, limit) = super::es_sql_pagination("SELECT field FROM idx");
assert_eq!(base, "SELECT field FROM idx");
assert_eq!(limit, None);
}
#[test]
fn should_use_search_cursor_distinguishes_plan_from_user_offset() {
let plan_first = serde_json::json!({ "size": 100 });
let user_offset = serde_json::json!({ "size": 100, "from": 5 });
assert!(super::should_use_search_cursor(&plan_first, true, None));
assert!(!super::should_use_search_cursor(&user_offset, true, None));
assert!(super::should_use_search_cursor(&user_offset, true, Some("cursor")));
assert!(!super::should_use_search_cursor(&plan_first, false, None));
}
#[test]
fn search_cursor_encode_decode_round_trips() {
let cursor = super::EsSearchCursor::Page(super::EsPageCursor {
index: "logs".to_string(),
body: serde_json::json!({ "query": { "match_all": {} } }),
size: 100,
mode: super::EsPageMode::Pit {
pit_id: "pit-1".to_string(),
keep_alive: "1m".to_string(),
search_after: vec![serde_json::json!("abc")],
},
});
let encoded = super::encode_es_search_cursor(&cursor).unwrap();
let decoded = super::decode_es_search_cursor(&encoded).unwrap();
match decoded {
super::EsSearchCursor::Page(page) => match page.mode {
super::EsPageMode::Pit { pit_id, search_after, .. } => {
assert_eq!(pit_id, "pit-1");
assert_eq!(search_after, vec![serde_json::json!("abc")]);
}
super::EsPageMode::Scroll { .. } => panic!("expected PIT page cursor"),
super::EsPageMode::Offset { .. } => panic!("expected PIT page cursor"),
},
super::EsSearchCursor::Sql { .. } => panic!("expected PIT search cursor"),
}
let sql = super::EsSearchCursor::Sql {
cursor: "raw-sql-cursor".to_string(),
columns_key: "columns".to_string(),
columns: vec![serde_json::json!({ "name": "message", "type": "keyword" })],
};
let encoded = super::encode_es_search_cursor(&sql).unwrap();
match super::decode_es_search_cursor(&encoded).unwrap() {
super::EsSearchCursor::Sql { cursor, columns_key, columns } => {
assert_eq!(cursor, "raw-sql-cursor");
assert_eq!(columns_key, "columns");
assert_eq!(columns, vec![serde_json::json!({ "name": "message", "type": "keyword" })]);
}
super::EsSearchCursor::Page(_) => panic!("expected SQL cursor"),
}
}
#[tokio::test]
async fn pit_search_uses_global_endpoint_and_latest_pit_id() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("POST /products/_pit?keep_alive=1m "), "{request}");
let response_body = r#"{"id":"pit-1"}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("POST /_search "), "{request}");
assert!(request.contains(r#""pit":{"id":"pit-1","keep_alive":"1m"}"#), "{request}");
let response_body = r#"{"pit_id":"pit-2","hits":{"total":{"value":2,"relation":"eq"},"hits":[{"_id":"1","_source":{"name":"one"},"sort":[1]}]}}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::find_documents_with_cursor(&client, "products", 1, None, None, None).await.unwrap();
server.await.unwrap();
let page = match super::decode_es_search_cursor(result.next_cursor.as_deref().unwrap()).unwrap() {
super::EsSearchCursor::Page(page) => page,
super::EsSearchCursor::Sql { .. } => panic!("expected page cursor"),
};
match page.mode {
super::EsPageMode::Pit { pit_id, .. } => assert_eq!(pit_id, "pit-2"),
super::EsPageMode::Scroll { .. } => panic!("expected PIT mode"),
super::EsPageMode::Offset { .. } => panic!("expected PIT mode"),
}
}
/// 依次应答固定的一串响应,并记下收到的请求。
async fn serve_responses(
listener: tokio::net::TcpListener,
responses: Vec<(u16, String)>,
) -> tokio::task::JoinHandle<Vec<String>> {
use tokio::io::AsyncWriteExt;
tokio::spawn(async move {
let mut requests = Vec::new();
for (status, body) in responses {
let (mut socket, _) = listener.accept().await.unwrap();
requests.push(read_http_request(&mut socket).await);
let reason = if status >= 400 { "Bad Request" } else { "OK" };
let response = format!(
"HTTP/1.1 {status} {reason}\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
}
requests
})
}
/// PIT 开得成功、但不认 `_shard_doc` 的集群返回的报错体。
fn missing_shard_doc_error_body() -> String {
r#"{"error":{"root_cause":[{"type":"query_shard_exception","reason":"No mapping found for [_shard_doc] in order to sort on","index":"products"}],"type":"search_phase_execution_exception","reason":"all shards failed"},"status":400}"#
.to_string()
}
fn one_hit_response() -> String {
r#"{"hits":{"total":{"value":42,"relation":"eq"},"hits":[{"_id":"1","_source":{"name":"one"}}]},"_shards":{"total":1,"successful":1,"skipped":0,"failed":0}}"#.to_string()
}
#[tokio::test]
async fn unsupported_pit_uses_scroll_cursor_before_offset_fallback() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = serve_responses(
listener,
vec![
(400, r#"{"error":"PIT is not supported"}"#.to_string()),
(
200,
r#"{"_scroll_id":"scroll-1","hits":{"total":{"value":2,"relation":"eq"},"hits":[{"_id":"1","_source":{"name":"one"}}]},"_shards":{"total":1,"successful":1,"skipped":0,"failed":0}}"#.to_string(),
),
(
200,
r#"{"_scroll_id":"scroll-2","hits":{"total":{"value":2,"relation":"eq"},"hits":[]},"_shards":{"total":1,"successful":1,"skipped":0,"failed":0}}"#.to_string(),
),
],
)
.await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let first = super::find_documents_with_cursor(&client, "products", 1, None, None, None).await.unwrap();
let second =
super::find_documents_with_cursor(&client, "products", 1, None, None, first.next_cursor.as_deref())
.await
.unwrap();
let requests = server.await.unwrap();
assert!(requests[0].starts_with("POST /products/_pit?keep_alive=1m "), "{}", requests[0]);
assert!(requests[1].starts_with("POST /products/_search?scroll=1m "), "{}", requests[1]);
assert!(requests[1].contains(r#""size":1"#), "{}", requests[1]);
assert!(requests[2].starts_with("POST /_search/scroll "), "{}", requests[2]);
assert!(requests[2].contains(r#""scroll_id":"scroll-1""#), "{}", requests[2]);
assert_eq!(first.documents.len(), 1);
assert!(first.next_cursor.is_some());
assert!(second.documents.is_empty());
assert!(second.next_cursor.is_none());
}
#[tokio::test]
async fn missing_shard_doc_support_falls_back_to_scroll_paging() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = serve_responses(
listener,
vec![
(200, r#"{"id":"pit-1"}"#.to_string()),
(400, missing_shard_doc_error_body()),
(200, "{}".to_string()),
(
200,
r#"{"_scroll_id":"scroll-1","hits":{"total":{"value":42,"relation":"eq"},"hits":[{"_id":"1","_source":{"name":"one"}}]},"_shards":{"total":1,"successful":1,"skipped":0,"failed":0}}"#.to_string(),
),
],
)
.await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::find_documents_with_cursor(&client, "products", 1, None, None, None).await.unwrap();
let requests = server.await.unwrap();
assert!(requests[0].starts_with("POST /products/_pit?keep_alive=1m "), "{}", requests[0]);
assert!(requests[1].starts_with("POST /_search "), "{}", requests[1]);
assert!(requests[1].contains("_shard_doc"), "{}", requests[1]);
assert!(requests[2].starts_with("DELETE /_pit "), "{}", requests[2]);
assert!(requests[3].starts_with("POST /products/_search?scroll=1m "), "{}", requests[3]);
assert!(!requests[3].contains("_shard_doc"), "fallback must not sort on _shard_doc: {}", requests[3]);
assert!(!requests[3].contains(r#""from":0"#), "{}", requests[3]);
assert_eq!(result.documents.len(), 1);
assert_eq!(result.total, 42);
// 回退后仍要给出游标,否则数据浏览器只能停在第一页。
let page = match super::decode_es_search_cursor(result.next_cursor.as_deref().unwrap()).unwrap() {
super::EsSearchCursor::Page(page) => page,
super::EsSearchCursor::Sql { .. } => panic!("expected page cursor"),
};
assert_eq!(
page.mode,
super::EsPageMode::Scroll { scroll_id: "scroll-1".to_string(), keep_alive: "1m".to_string() }
);
}
#[tokio::test]
async fn scroll_fallback_is_remembered_for_the_connection() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = serve_responses(
listener,
vec![
(200, r#"{"id":"pit-1"}"#.to_string()),
(400, missing_shard_doc_error_body()),
(200, "{}".to_string()),
(200, one_hit_response()),
(200, one_hit_response()),
],
)
.await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let _first = super::find_documents_with_cursor(&client, "products", 1, None, None, None).await.unwrap();
let second = super::find_documents_with_cursor(&client, "products", 1, None, None, None).await.unwrap();
let requests = server.await.unwrap();
// 探测结果记在连接上:第二次查询不再尝试开 PIT。
assert_eq!(requests.len(), 5);
assert!(requests[4].starts_with("POST /products/_search?scroll=1m "), "{}", requests[4]);
assert_eq!(second.documents.len(), 1);
}
#[tokio::test]
async fn transient_search_failure_does_not_disable_pit_for_the_connection() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = serve_responses(
listener,
vec![
(200, r#"{"id":"pit-1"}"#.to_string()),
(503, r#"{"error":{"type":"unavailable_shards_exception"},"status":503}"#.to_string()),
(200, "{}".to_string()),
(200, one_hit_response()),
(200, r#"{"id":"pit-2"}"#.to_string()),
(200, one_hit_response()),
(200, "{}".to_string()),
],
)
.await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let first = super::find_documents_with_cursor(&client, "products", 1, None, None, None).await.unwrap();
assert_eq!(first.documents.len(), 1);
let _ = super::find_documents_with_cursor(&client, "products", 1, None, None, None).await.unwrap();
let requests = server.await.unwrap();
// 偶发 5xx 只回退这一次请求,下一次查询仍旧先试 PIT。
assert!(requests[3].starts_with("POST /products/_search?scroll=1m "), "{}", requests[3]);
assert!(requests[4].starts_with("POST /products/_pit?keep_alive=1m "), "{}", requests[4]);
}
#[tokio::test]
async fn offset_paging_stops_at_the_max_result_window() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let hits = (0..5_000).map(|id| format!(r#"{{"_id":"{id}","_source":{{}}}}"#)).collect::<Vec<_>>().join(",");
let full_page = format!(
r#"{{"hits":{{"total":{{"value":100000,"relation":"eq"}},"hits":[{hits}]}},"_shards":{{"total":1,"successful":1,"skipped":0,"failed":0}}}}"#
);
let server = serve_responses(listener, vec![(200, full_page.clone()), (200, full_page)]).await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let cursor_at = |from: u64| {
super::encode_es_search_cursor(&super::EsSearchCursor::Page(super::EsPageCursor {
index: "products".to_string(),
body: serde_json::json!({ "sort": ["_doc"] }),
size: 5_000,
mode: super::EsPageMode::Offset { from },
}))
.unwrap()
};
// from=0:下一页是 5000..10000,正好落在窗口内,可以继续翻。
let first = super::find_documents_with_cursor(&client, "products", 5_000, None, None, Some(&cursor_at(0)))
.await
.unwrap();
let page = match super::decode_es_search_cursor(first.next_cursor.as_deref().unwrap()).unwrap() {
super::EsSearchCursor::Page(page) => page,
super::EsSearchCursor::Sql { .. } => panic!("expected page cursor"),
};
assert_eq!(page.mode, super::EsPageMode::Offset { from: 5_000 });
// from=5000:再下一页要到 10000..15000,越过 max_result_window,必须停在这里,
// 否则那次请求一定会被 ES 拒绝。
let second = super::find_documents_with_cursor(&client, "products", 5_000, None, None, Some(&cursor_at(5_000)))
.await
.unwrap();
server.await.unwrap();
assert_eq!(second.documents.len(), 5_000);
assert!(second.next_cursor.is_none(), "must not offer a page past max_result_window");
}
#[tokio::test]
async fn offset_paging_never_loops_on_an_empty_page() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let empty = r#"{"hits":{"total":{"value":0,"relation":"eq"},"hits":[]},"_shards":{"total":1,"successful":1,"skipped":0,"failed":0}}"#;
let server = serve_responses(listener, vec![(200, empty.to_string())]).await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
// `LIMIT 0` 会让 size 为 0;空页若被当成「满页」就会不断吐出停在同一个
// from 上的游标,翻页永远结束不了。
let cursor = super::encode_es_search_cursor(&super::EsSearchCursor::Page(super::EsPageCursor {
index: "products".to_string(),
body: serde_json::json!({}),
size: 0,
mode: super::EsPageMode::Offset { from: 0 },
}))
.unwrap();
let result =
super::execute_rest_query_with_cursor(&client, "SELECT * FROM products LIMIT 0 OFFSET 0", Some(&cursor))
.await
.unwrap();
server.await.unwrap();
assert!(result.rows.is_empty());
assert!(!result.has_more, "an empty page must not advertise a next page");
}
#[tokio::test]
async fn scroll_fallback_keeps_track_total_hits_off_and_preserves_requested_sort() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = serve_responses(
listener,
vec![
(200, r#"{"id":"pit-1"}"#.to_string()),
(400, missing_shard_doc_error_body()),
(200, "{}".to_string()),
(200, one_hit_response()),
],
)
.await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
super::find_documents_with_cursor(&client, "products", 1, None, Some(r#"{"created_at":-1}"#), None)
.await
.unwrap();
let requests = server.await.unwrap();
let fallback = &requests[3];
// Scroll 持有搜索快照,保留用户的排序即可,无需追加 `_doc`。
assert!(fallback.starts_with("POST /products/_search?scroll=1m "), "{fallback}");
assert!(fallback.contains(r#""sort":[{"created_at":{"order":"desc"}}]"#), "{fallback}");
// 回退路径不加 track_total_hits:那会让每页都对整个索引做一次精确计数,
// 而这条路径针对的正是老的大集群。
assert!(!fallback.contains("track_total_hits"), "{fallback}");
assert!(requests[1].contains(r#""track_total_hits":true"#), "{}", requests[1]);
}
#[tokio::test]
async fn transport_failure_is_not_retried_through_the_offset_path() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
// 开 PIT 成功后直接断连,模拟检索请求根本没送到。
let server = tokio::spawn(async move {
use tokio::io::AsyncWriteExt;
let (mut socket, _) = listener.accept().await.unwrap();
read_http_request(&mut socket).await;
let body = r#"{"id":"pit-1"}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
drop(socket);
let (socket, _) = listener.accept().await.unwrap();
drop(socket);
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let error = super::find_documents_with_cursor(&client, "products", 1, None, None, None).await.unwrap_err();
server.abort();
// 传输层失败换个分页方式也送不到,只会让用户多等一倍。
assert!(error.starts_with("Elasticsearch request failed:"), "{error}");
assert!(client.supports_pit_search(), "transport failures must not downgrade the connection");
}
#[tokio::test]
async fn select_star_shares_the_same_offset_fallback() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = serve_responses(
listener,
vec![
(200, r#"{"id":"pit-1"}"#.to_string()),
(400, missing_shard_doc_error_body()),
(200, "{}".to_string()),
(200, one_hit_response()),
],
)
.await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
// 查询编辑器与文档浏览器共用同一个分页执行器,兜底行为必须一致。
let result = super::execute_rest_query_with_cursor(&client, "SELECT * FROM products LIMIT 2 OFFSET 0", None)
.await
.unwrap();
let requests = server.await.unwrap();
assert!(requests[0].starts_with("POST /products/_pit?keep_alive=1m "), "{}", requests[0]);
assert!(requests[1].contains("_shard_doc"), "{}", requests[1]);
assert!(requests[3].starts_with("POST /products/_search?scroll=1m "), "{}", requests[3]);
assert!(!requests[3].contains("_shard_doc"), "fallback must not sort on _shard_doc: {}", requests[3]);
assert_eq!(result.rows.len(), 1);
// 分页计划要拿索引真实总数来算总页数。
assert_eq!(result.affected_rows, 42);
}
#[tokio::test]
async fn sql_cursor_continuation_reuses_first_page_columns() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("POST /_sql "), "{request}");
let response_body = r#"{"columns":[{"name":"name","type":"keyword"}],"rows":[["first"]],"cursor":"raw-1"}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("POST /_sql "), "{request}");
assert!(request.ends_with(r#"{"cursor":"raw-1"}"#), "{request}");
let response_body = r#"{"rows":[["second"]]}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let first = super::execute_rest_query(&client, "SELECT name FROM products").await.unwrap();
let second =
super::execute_rest_query_with_cursor(&client, "SELECT name FROM products", first.session_id.as_deref())
.await
.unwrap();
server.await.unwrap();
assert_eq!(second.columns, vec!["name"]);
assert_eq!(second.rows, vec![vec![json!("second")]]);
assert!(!second.has_more);
}
/// ES 6.x 的 `_sql` 只注册了 GET/PUT,POST 会返回 405(#9484)。
fn es6_method_not_allowed_body() -> String {
r#"{"error":"Incorrect HTTP method for uri [/_sql] and method [POST], allowed: [GET, PUT, HEAD, DELETE]","status":405}"#
.to_string()
}
fn es_sql_rows_body(row: &str) -> String {
format!(r#"{{"columns":[{{"name":"name","type":"keyword"}}],"rows":[["{row}"]]}}"#)
}
#[tokio::test]
async fn es6_sql_endpoint_retries_with_put_after_method_not_allowed() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server =
serve_responses(listener, vec![(405, es6_method_not_allowed_body()), (200, es_sql_rows_body("es6-row"))])
.await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "SELECT name FROM products").await.unwrap();
let requests = server.await.unwrap();
assert!(requests[0].starts_with("POST /_sql "), "{}", requests[0]);
assert!(requests[1].starts_with("PUT /_sql "), "{}", requests[1]);
assert!(
requests[1].ends_with(r#"{"query":"SELECT name FROM products","fetch_size":10000}"#),
"{}",
requests[1]
);
assert_eq!(result.rows, vec![vec![json!("es6-row")]]);
// 探测结果按连接保存:后续查询直接走 PUT,不再多打一次 POST。
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = serve_responses(listener, vec![(200, es_sql_rows_body("es6-row-2"))]).await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
client.mark_sql_endpoint_uses_put();
super::execute_rest_query(&client, "SELECT name FROM products").await.unwrap();
let requests = server.await.unwrap();
assert!(requests[0].starts_with("PUT /_sql "), "{}", requests[0]);
}
#[tokio::test]
async fn es6_sql_cursor_close_uses_the_learned_put_method() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = serve_responses(
listener,
vec![
(405, es6_method_not_allowed_body()),
(
200,
r#"{"columns":[{"name":"name","type":"keyword"}],"rows":[["es6-row"]],"cursor":"raw-1"}"#
.to_string(),
),
(200, r#"{"succeeded":true}"#.to_string()),
],
)
.await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "SELECT name FROM products").await.unwrap();
super::close_cursor(&client, &result.session_id.expect("cursor returned")).await.unwrap();
let requests = server.await.unwrap();
assert!(requests[1].starts_with("PUT /_sql "), "{}", requests[1]);
assert!(requests[2].starts_with("PUT /_sql/close "), "{}", requests[2]);
assert!(requests[2].ends_with(r#"{"cursor":"raw-1"}"#), "{}", requests[2]);
}
#[tokio::test]
async fn es78_sql_endpoint_keeps_posting_when_put_is_rejected() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = serve_responses(
listener,
vec![
(405, es6_method_not_allowed_body()),
(405, r#"{"error":"Incorrect HTTP method for uri [/_sql] and method [PUT], allowed: [GET, POST, HEAD, DELETE]","status":405}"#.to_string()),
],
)
.await;
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let error = super::execute_rest_query(&client, "SELECT name FROM products").await.unwrap_err();
let requests = server.await.unwrap();
assert!(requests[1].starts_with("PUT /_sql "), "{}", requests[1]);
assert!(error.contains("405"), "{error}");
// 两次都是 405 时不锁定 PUT,避免把只认 POST 的集群永久改坏。
assert!(!client.sql_endpoint_uses_put());
}
#[tokio::test]
async fn closes_pit_and_sql_cursors_with_elasticsearch_protocols() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("DELETE /_pit "), "{request}");
assert!(request.ends_with(r#"{"id":"pit-1"}"#), "{request}");
socket
.write_all(b"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: 2\r\nConnection: close\r\n\r\n{}")
.await
.unwrap();
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("POST /_sql/close "), "{request}");
assert!(request.ends_with(r#"{"cursor":"sql-1"}"#), "{request}");
socket
.write_all(b"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: 2\r\nConnection: close\r\n\r\n{}")
.await
.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
super::close_es_pit(&client, "pit-1").await.unwrap();
super::close_es_sql_cursor(&client, "sql-1").await.unwrap();
server.await.unwrap();
}
#[test]
fn offset_cursor_encode_decode_round_trips() {
let cursor = super::EsSearchCursor::Page(super::EsPageCursor {
index: "logs".to_string(),
body: serde_json::json!({ "sort": ["_doc"], "query": { "term": { "city": "长治" } } }),
size: 100,
mode: super::EsPageMode::Offset { from: 200 },
});
let encoded = super::encode_es_search_cursor(&cursor).unwrap();
match super::decode_es_search_cursor(&encoded).unwrap() {
super::EsSearchCursor::Page(page) => {
assert_eq!(page.index, "logs");
assert_eq!(page.size, 100);
assert_eq!(page.mode, super::EsPageMode::Offset { from: 200 });
}
super::EsSearchCursor::Sql { .. } => panic!("expected page cursor"),
}
}
#[test]
fn builds_elasticsearch_find_body_with_native_query_builder_filter() {
let body = build_find_documents_body(
0,
25,
Some(
r#"{
"$esQuery": {
"bool": {
"must": [{"match": {"customer_name": "Customer"}}],
"filter": [{"range": {"amount": {"gte": 500}}}],
"must_not": [{"term": {"status": "cancelled"}}]
}
}
}"#,
),
None,
)
.unwrap();
assert_eq!(
body["query"],
json!({
"bool": {
"must": [{"match": {"customer_name": "Customer"}}],
"filter": [{"range": {"amount": {"gte": 500}}}],
"must_not": [{"term": {"status": "cancelled"}}]
}
})
);
}
#[test]
fn builds_elasticsearch_count_body_with_the_same_native_query_filter() {
let body = build_count_documents_body(Some(r#"{"$esQuery":{"term":{"status":"active"}}}"#)).unwrap();
assert_eq!(body, json!({ "query": { "term": { "status": "active" } } }));
assert!(body.get("from").is_none());
assert!(body.get("size").is_none());
assert!(body.get("sort").is_none());
}
#[test]
fn builds_elasticsearch_count_body_with_structured_filter_operators() {
let filter = r#"{"$and":[{"city":{"$ne":"上海"}},{"age":{"$gte":18}}]}"#;
assert_eq!(
build_count_documents_body(Some(filter)).unwrap(),
json!({
"query": {
"bool": {
"filter": [
{ "bool": { "must_not": [{ "term": { "city": "上海" } }] } },
{ "range": { "age": { "gte": 18 } } }
]
}
}
})
);
}
#[test]
fn accepts_only_complete_elasticsearch_count_shards() {
assert!(super::ElasticsearchShards { total: 3, successful: 2, skipped: 1, failed: 0 }.is_complete());
assert!(!super::ElasticsearchShards { total: 3, successful: 2, skipped: 0, failed: 0 }.is_complete());
assert!(!super::ElasticsearchShards { total: 3, successful: 2, skipped: 0, failed: 1 }.is_complete());
}
#[tokio::test]
async fn counts_documents_with_the_translated_filter() {
use tokio::io::AsyncWriteExt;
let response_body = r#"{"count":552033,"_shards":{"total":3,"successful":3,"skipped":0,"failed":0}}"#;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("POST /orders/_count "));
let body = request.split_once("\r\n\r\n").unwrap().1;
assert_eq!(
serde_json::from_str::<serde_json::Value>(body).unwrap(),
json!({ "query": { "term": { "status": "active" } } })
);
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body,
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
assert_eq!(super::count_documents(&client, "orders", Some(r#"{"status":"active"}"#)).await.unwrap(), 552_033);
server.await.unwrap();
}
#[tokio::test]
async fn rejects_partial_elasticsearch_document_count() {
use tokio::io::AsyncWriteExt;
let response_body = r#"{"count":4,"_shards":{"total":3,"successful":2,"skipped":0,"failed":1}}"#;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("POST /orders/_count "));
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body,
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let error = super::count_documents(&client, "orders", None).await.unwrap_err();
assert!(error.contains("incomplete shard response: 2 successful, 0 skipped, 1 failed of 3 shards"));
server.await.unwrap();
}
#[test]
fn builds_elasticsearch_find_body_with_structured_filter_operators() {
let body = build_find_documents_body(
0,
100,
Some(
r#"{
"$and": [
{"city": {"$ne": "上海"}},
{"age": {"$gt": 18, "$lte": 60}},
{"name": {"$not": {"$regex": "test", "$options": "i"}}}
]
}"#,
),
None,
)
.unwrap();
assert_eq!(
body,
json!({
"from": 0,
"size": 100,
"query": {
"bool": {
"filter": [
{ "bool": { "must_not": [{ "term": { "city": "上海" } }] } },
{ "range": { "age": { "gt": 18, "lte": 60 } } },
{
"bool": {
"must_not": [
{
"wildcard": {
"name": {
"value": "*test*",
"case_insensitive": true
}
}
}
]
}
}
]
}
},
"sort": ["_doc"]
})
);
}
#[test]
fn builds_elasticsearch_find_body_with_or_filter() {
let body =
build_find_documents_body(0, 50, Some(r#"{"$or":[{"city":"长治"},{"city":"上海"}]}"#), None).unwrap();
assert_eq!(
body["query"],
json!({
"bool": {
"should": [
{ "term": { "city": "长治" } },
{ "term": { "city": "上海" } }
],
"minimum_should_match": 1
}
})
);
}
#[test]
fn parses_search_total_from_elasticsearch_6_number_shape() {
let response: SearchResponse = serde_json::from_value(json!({
"_shards": { "total": 1, "successful": 1, "skipped": 0, "failed": 0 },
"hits": {
"total": 5,
"hits": []
}
}))
.unwrap();
assert_eq!(response.hits.total.value(), 5);
assert!(response.hits.total.is_exact());
}
#[test]
fn parses_search_total_from_elasticsearch_7_object_shape() {
let response: SearchResponse = serde_json::from_value(json!({
"_shards": { "total": 1, "successful": 1, "skipped": 0, "failed": 0 },
"hits": {
"total": { "value": 5, "relation": "eq" },
"hits": []
}
}))
.unwrap();
assert_eq!(response.hits.total.value(), 5);
assert!(response.hits.total.is_exact());
let result = super::search_response_to_document_result(response).unwrap();
assert!(result.total_is_exact);
}
#[test]
fn preserves_elasticsearch_lower_bound_total_relation() {
let response: SearchResponse = serde_json::from_value(json!({
"hits": {
"total": { "value": 10_000, "relation": "gte" },
"hits": []
}
}))
.unwrap();
assert_eq!(response.hits.total.value(), 10_000);
assert!(!response.hits.total.is_exact());
let result = super::search_response_to_document_result(response).unwrap();
assert_eq!(result.total, 10_000);
assert!(!result.total_is_exact);
}
#[test]
fn treats_search_total_as_a_lower_bound_when_a_shard_failed() {
let response: SearchResponse = serde_json::from_value(json!({
"_shards": { "total": 3, "successful": 2, "skipped": 0, "failed": 1 },
"hits": {
"total": { "value": 5, "relation": "eq" },
"hits": []
}
}))
.unwrap();
let result = super::search_response_to_document_result(response).unwrap();
assert_eq!(result.total, 5);
assert!(!result.total_is_exact);
}
#[test]
fn treats_timed_out_or_terminated_searches_as_lower_bounds() {
for response in [
json!({
"timed_out": true,
"_shards": { "total": 1, "successful": 1, "skipped": 0, "failed": 0 },
"hits": { "total": { "value": 5, "relation": "eq" }, "hits": [] }
}),
json!({
"terminated_early": true,
"_shards": { "total": 1, "successful": 1, "skipped": 0, "failed": 0 },
"hits": { "total": { "value": 5, "relation": "eq" }, "hits": [] }
}),
] {
let response: SearchResponse = serde_json::from_value(response).unwrap();
let result = super::search_response_to_document_result(response).unwrap();
assert_eq!(result.total, 5);
assert!(!result.total_is_exact);
}
}
#[test]
fn treats_search_total_without_shard_metadata_as_a_lower_bound() {
let response: SearchResponse = serde_json::from_value(json!({
"hits": {
"total": { "value": 5, "relation": "eq" },
"hits": []
}
}))
.unwrap();
let result = super::search_response_to_document_result(response).unwrap();
assert_eq!(result.total, 5);
assert!(!result.total_is_exact);
}
#[test]
fn parses_elasticsearch_hit_routing_metadata() {
let response: SearchResponse = serde_json::from_value(json!({
"hits": {
"total": { "value": 1, "relation": "eq" },
"hits": [
{ "_id": "doc-1", "_routing": "tenant-1", "_source": { "name": "Alice" } }
]
}
}))
.unwrap();
assert_eq!(response.hits.hits[0].routing.as_deref(), Some("tenant-1"));
}
#[test]
fn preserves_legacy_elasticsearch_document_type_metadata() {
let response: SearchResponse = serde_json::from_value(json!({
"hits": {
"total": { "value": 2, "relation": "eq" },
"hits": [
{ "_id": "legacy-1", "_type": "order", "_source": { "name": "Legacy" } },
{ "_id": "modern-1", "_type": "_doc", "_source": { "name": "Modern" } }
]
}
}))
.unwrap();
let result = super::search_response_to_document_result(response).unwrap();
assert_eq!(result.documents[0]["_type"], json!("order"));
assert!(result.documents[1].get("_type").is_none());
}
#[test]
fn preserves_long_literals_in_document_transport_json() {
let response: SearchResponse = serde_json::from_str(
r#"{"hits":{"total":{"value":1,"relation":"eq"},"hits":[{"_id":"doc-1","_source":{"id":2018551659033767937,"string_id":"2018551659033767937","legacy":{"$numberLong":"2018551659033767937"}}}]}}"#,
)
.unwrap();
let result = super::search_response_to_document_result(response).unwrap();
assert_eq!(
result.raw_documents.as_deref(),
Some(&[r#"{"id":2018551659033767937,"string_id":"2018551659033767937","legacy":{"$numberLong":"2018551659033767937"},"_id":"doc-1"}"#.to_string()][..])
);
}
#[test]
fn parses_search_response_rows_with_routing_metadata() {
let result = super::parse_elasticsearch_response(
200,
json!({
"hits": {
"hits": [
{ "_id": "doc-1", "_routing": "tenant-1", "_source": { "name": "Alice" } }
]
}
}),
std::time::Instant::now(),
)
.unwrap();
assert_ne!(result.columns, vec!["status", "response"]);
let name_idx = result.columns.iter().position(|column| column == "name").unwrap();
assert_eq!(result.rows[0][name_idx], json!("Alice"));
let routing_idx = result.columns.iter().position(|column| column == "_routing").unwrap();
assert_eq!(result.rows[0][routing_idx], json!("tenant-1"));
}
#[test]
fn keeps_sql_api_response_tabular() {
let result = super::parse_elasticsearch_response(
200,
json!({
"columns": [{ "name": "name" }],
"rows": [["Alice"]]
}),
std::time::Instant::now(),
)
.unwrap();
assert_eq!(result.columns, vec!["name"]);
assert_eq!(result.rows, vec![vec![json!("Alice")]]);
}
#[test]
fn parses_aggregation_response_before_empty_hits() {
let result = super::parse_elasticsearch_response(
200,
json!({
"hits": {
"total": { "value": 5, "relation": "eq" },
"hits": []
},
"aggregations": {
"by_status": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{ "key": "paid", "doc_count": 3 },
{ "key": "cancelled", "doc_count": 1 },
{ "key": "pending", "doc_count": 1 }
]
}
}
}),
std::time::Instant::now(),
)
.unwrap();
let key_idx = result.columns.iter().position(|column| column == "key").unwrap();
let count_idx = result.columns.iter().position(|column| column == "doc_count").unwrap();
assert_eq!(result.rows.len(), 3);
assert_eq!(result.rows[0][key_idx], json!("paid"));
assert_eq!(result.rows[0][count_idx], json!("3"));
assert_eq!(result.affected_rows, 3);
}
#[test]
fn parses_plain_text_rest_response_without_dropping_body() {
let body =
"health status index docs.count store.size\ngreen open app-log-2026-07 42 10mb\n";
let result = super::parse_elasticsearch_rest_response(200, body, std::time::Instant::now()).unwrap();
assert_eq!(result.columns, vec!["response"]);
assert_eq!(result.rows.len(), 2);
assert_eq!(result.rows[0][0], json!("health status index docs.count store.size"));
assert_eq!(result.rows[1][0], json!("green open app-log-2026-07 42 10mb"));
assert_eq!(result.affected_rows, 2);
}
#[test]
fn adds_json_format_only_when_cat_format_is_not_explicit() {
let plain =
super::add_default_cat_json_format(super::parse_elasticsearch_rest_request("GET /_cat/indices").unwrap());
assert_eq!(plain.path, "/_cat/indices?format=json");
let defaulted =
super::add_default_cat_json_format(super::parse_elasticsearch_rest_request("GET /_cat/indices?v").unwrap());
assert_eq!(defaulted.path, "/_cat/indices?v&format=json");
let explicit_text = super::add_default_cat_json_format(
super::parse_elasticsearch_rest_request("GET /_cat/indices?format=txt").unwrap(),
);
assert_eq!(explicit_text.path, "/_cat/indices?format=txt");
}
#[test]
fn preserves_http_status_for_plain_text_rest_errors() {
let result =
super::parse_elasticsearch_rest_response(503, "service temporarily unavailable", std::time::Instant::now())
.unwrap();
assert_eq!(result.columns, vec!["status", "response"]);
assert_eq!(result.rows, vec![vec![json!(503), json!("service temporarily unavailable")]]);
}
#[test]
fn keeps_mapping_rest_response_numeric_literals_lossless() {
let body = r#"{
"products": {
"mappings": {
"_meta": {
"largest_id": 123456789012345678901234567890,
"ratio": 0.123456789012345678901234567890,
"estimate": 1e400
},
"properties": { "name": { "type": "keyword" } }
}
}
}"#;
let result = super::parse_elasticsearch_rest_response(200, body, std::time::Instant::now()).unwrap();
assert_eq!(result.columns, vec!["status", "response"]);
assert_eq!(result.rows[0][0], json!(200));
assert_eq!(result.rows[0][1].as_str(), Some(body));
}
#[tokio::test]
async fn execute_rest_query_preserves_index_specific_cat_json_response() {
use tokio::io::{AsyncReadExt, AsyncWriteExt};
let body = r#"[{"health":"green","index":"app-log-2026-07","docs.count":"42"}]"#;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let mut request = [0_u8; 1024];
let read = socket.read(&mut request).await.unwrap();
let request = String::from_utf8_lossy(&request[..read]);
assert!(request.starts_with("GET /_cat/indices/data_pack_and_box_index_v1?format=json "));
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "GET /_cat/indices/data_pack_and_box_index_v1").await.unwrap();
server.await.unwrap();
assert_eq!(result.columns, vec!["status", "response"]);
assert_eq!(result.rows, vec![vec![json!(200), json!(body)]]);
}
#[tokio::test]
async fn execute_rest_query_preserves_explicit_cat_json_response() {
use tokio::io::{AsyncReadExt, AsyncWriteExt};
let body = r#"[{"index":"data_pack_and_box_index_v1","docs.count":"42"}]"#;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let mut request = [0_u8; 1024];
let read = socket.read(&mut request).await.unwrap();
let request = String::from_utf8_lossy(&request[..read]);
assert!(request.starts_with("GET /_cat/indices/data_pack_and_box_index_v1?format=json "));
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "GET /_cat/indices/data_pack_and_box_index_v1?format=json")
.await
.unwrap();
server.await.unwrap();
assert_eq!(result.columns, vec!["status", "response"]);
assert_eq!(result.rows, vec![vec![json!(200), json!(body)]])
}
#[tokio::test]
async fn execute_rest_query_keeps_default_cat_text_when_server_does_not_return_json() {
use tokio::io::{AsyncReadExt, AsyncWriteExt};
let body = "health status\ngreen open\n";
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let mut request = [0_u8; 1024];
let read = socket.read(&mut request).await.unwrap();
let request = String::from_utf8_lossy(&request[..read]);
assert!(request.starts_with("GET /_cat/indices?format=json "));
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: text/plain\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "GET /_cat/indices").await.unwrap();
server.await.unwrap();
assert_eq!(result.columns, vec!["response"]);
assert_eq!(result.rows, vec![vec![json!("health status")], vec![json!("green open")]]);
}
#[tokio::test]
async fn execute_rest_query_keeps_explicit_text_cat_response() {
use tokio::io::{AsyncReadExt, AsyncWriteExt};
let body = "health status\ngreen open\n";
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let mut request = [0_u8; 1024];
let read = socket.read(&mut request).await.unwrap();
let request = String::from_utf8_lossy(&request[..read]);
assert!(request.starts_with("GET /_cat/indices?format=txt "));
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: text/plain\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "GET /_cat/indices?format=txt").await.unwrap();
server.await.unwrap();
assert_eq!(result.columns, vec!["response"]);
assert_eq!(result.rows, vec![vec![json!("health status")], vec![json!("green open")]]);
}
#[tokio::test]
async fn execute_rest_query_preserves_numeric_literals_from_http_body() {
use tokio::io::AsyncWriteExt;
let response_body = r#"{"largest_id":123456789012345678901234567890,"ratio":0.123456789012345678901234567890,"estimate":1e400}"#;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("GET /products/_mapping "));
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "GET /products/_mapping").await.unwrap();
server.await.unwrap();
assert_eq!(result.columns, vec!["status", "response"]);
assert_eq!(result.rows[0][0], json!(200));
assert_eq!(result.rows[0][1].as_str(), Some(response_body));
}
#[tokio::test]
async fn index_metadata_endpoints_read_their_index_scoped_paths() {
use tokio::io::AsyncWriteExt;
for (kind, endpoint) in [("mapping", "_mapping"), ("settings", "_settings"), ("stats", "_stats")] {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let expected = format!("GET /my-index/{endpoint} ");
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with(&expected), "unexpected request: {request}");
let body = r#"{"my-index":{}}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let value = match kind {
"mapping" => super::get_index_mapping(&client, "my-index").await.unwrap(),
"settings" => super::get_index_settings(&client, "my-index").await.unwrap(),
_ => super::get_index_stats(&client, "my-index").await.unwrap(),
};
server.await.unwrap();
assert_eq!(value, json!({ "my-index": {} }));
}
}
#[tokio::test]
async fn delete_all_documents_posts_match_all_and_reports_skipped_documents() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
// The index itself must survive: this is a delete-by-query, not a DELETE /index.
assert!(
request.starts_with("POST /logs/_delete_by_query?conflicts=proceed&refresh=true "),
"unexpected request: {request}"
);
assert!(request.contains(r#""match_all""#), "unexpected body: {request}");
let body =
r#"{"total":10,"deleted":8,"version_conflicts":2,"timed_out":false,"failures":[{"index":"logs"}]}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::delete_all_documents(&client, "logs").await.unwrap();
server.await.unwrap();
assert_eq!(result.total, 10);
assert_eq!(result.deleted, 8);
assert_eq!(result.version_conflicts, 2);
assert!(!result.timed_out);
assert_eq!(result.failures, vec![r#"{"index":"logs"}"#.to_string()]);
}
/// ES 6.x through 9.x all answer `_delete_by_query` with the same counter
/// names wrapped in a larger, version-dependent envelope (`retries`,
/// `throttled_millis`, `slices`, …). Parsing must ignore the envelope
/// instead of failing on fields a given server happens to add.
#[tokio::test]
async fn delete_all_documents_ignores_version_specific_response_fields() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
read_http_request(&mut socket).await;
let body = r#"{"took":42,"timed_out":false,"total":5,"deleted":5,"batches":1,"version_conflicts":0,"noops":0,"retries":{"bulk":0,"search":0},"throttled_millis":0,"requests_per_second":-1.0,"throttled_until_millis":0,"slices":[],"failures":[]}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
body.len(),
body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::delete_all_documents(&client, "logs").await.unwrap();
server.await.unwrap();
assert_eq!(result.total, 5);
assert_eq!(result.deleted, 5);
assert_eq!(result.version_conflicts, 0);
assert!(!result.timed_out);
assert!(result.failures.is_empty());
}
#[tokio::test]
async fn execute_rest_head_request_returns_http_status() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("HEAD /orders "));
socket.write_all(b"HTTP/1.1 200 OK\r\nContent-Length: 0\r\nConnection: close\r\n\r\n").await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "HEAD /orders").await.unwrap();
server.await.unwrap();
assert_eq!(result.columns, vec!["status", "response"]);
assert_eq!(result.rows[0][0], json!(200));
assert_eq!(result.rows[0][1], json!("null"));
}
#[tokio::test]
async fn kibana_proxy_rewrites_method_path_and_preserves_elasticsearch_status() {
use std::collections::HashMap;
use tokio::io::AsyncWriteExt;
let response_body = r#"{"error":{"type":"index_not_found_exception"},"status":404}"#;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
let (headers, body) = request.split_once("\r\n\r\n").unwrap();
let request_target = headers.lines().next().unwrap().split_whitespace().nth(1).unwrap();
let url = reqwest::Url::parse(&format!("http://localhost{request_target}")).unwrap();
let query = url.query_pairs().into_owned().collect::<HashMap<_, _>>();
assert_eq!(url.path(), "/kibana/s/analytics/api/console/proxy");
assert_eq!(query.get("path").map(String::as_str), Some("/missing/_doc/1?refresh=true"));
assert_eq!(query.get("method").map(String::as_str), Some("DELETE"));
assert!(headers.lines().any(|line| line.eq_ignore_ascii_case("kbn-xsrf: true")), "{headers}");
assert!(headers.lines().any(|line| line.eq_ignore_ascii_case("osd-xsrf: true")), "{headers}");
assert!(headers.lines().any(|line| line.eq_ignore_ascii_case("authorization: Basic ZWxhc3RpYzpzZWNyZXQ=")));
assert_eq!(serde_json::from_str::<serde_json::Value>(body).unwrap(), json!({ "reason": "cleanup" }));
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nx-console-proxy-status-code: 404\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let external_config = json!({ "mode": "kibana", "kibanaBasePath": "/kibana/s/analytics" });
let client = EsClient::from_config(
&format!("http://{addr}"),
Some("elastic"),
Some("secret"),
false,
None,
Some(&external_config),
Duration::from_secs(1),
None,
None,
None,
)
.expect("failed to build Elasticsearch test client");
let result =
super::execute_rest_query(&client, "DELETE /missing/_doc/1?refresh=true\n{\"reason\":\"cleanup\"}")
.await
.unwrap();
server.await.unwrap();
assert_eq!(result.rows[0][0], json!(404));
assert_eq!(result.rows[0][1], json!(response_body));
}
#[test]
fn direct_transport_omits_proxy_xsrf_headers() {
let client = EsClient::new("http://localhost:9200", None, None, false, Duration::from_secs(1));
let request = client.get("/").build().unwrap();
assert!(!request.headers().contains_key("kbn-xsrf"));
assert!(!request.headers().contains_key("osd-xsrf"));
}
#[tokio::test]
async fn execute_rest_query_sends_encoded_date_math_path() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("GET /%3Clogs-%7Bnow%2Fd%7D%3E/_search?pretty "), "{request}");
socket.write_all(b"HTTP/1.1 200 OK\r\nContent-Length: 2\r\nConnection: close\r\n\r\n{}").await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
super::execute_rest_query(&client, "GET /<logs-{now/d}>/_search?pretty").await.unwrap();
server.await.unwrap();
}
#[tokio::test]
async fn execute_rest_msearch_sends_ndjson_content_type_and_trailing_newline() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
let (headers, body) = request.split_once("\r\n\r\n").unwrap();
assert!(headers.starts_with("POST /_msearch "));
assert!(headers.lines().any(|line| line.eq_ignore_ascii_case("content-type: application/x-ndjson")));
assert_eq!(body, "{\"index\":\"orders\"}\n{\"size\":0}\n");
let response_body = r#"{"responses":[]}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(
&client,
"// run two searches\nPOST /_msearch\n{\"index\":\"orders\"}\n{\"size\":0}",
)
.await
.unwrap();
server.await.unwrap();
assert_eq!(result.rows[0][0], json!(200));
}
#[tokio::test]
async fn execute_rest_delete_sends_json_body() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
let (headers, body) = request.split_once("\r\n\r\n").unwrap();
assert!(headers.starts_with("DELETE /_search/scroll "));
assert_eq!(serde_json::from_str::<serde_json::Value>(body).unwrap(), json!({ "scroll_id": ["scroll-1"] }));
let response_body = r#"{"succeeded":true,"num_freed":1}"#;
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result =
super::execute_rest_query(&client, "DELETE /_search/scroll\n{\"scroll_id\":[\"scroll-1\"]}").await.unwrap();
server.await.unwrap();
assert_eq!(result.columns, vec!["status", "response"]);
assert_eq!(result.rows[0][0], json!(200));
assert_eq!(
serde_json::from_str::<serde_json::Value>(result.rows[0][1].as_str().unwrap()).unwrap(),
json!({ "succeeded": true, "num_freed": 1 })
);
}
#[test]
fn parses_search_rest_response_as_source_table() {
let body = r#"{"took":1,"hits":{"total":{"value":2,"relation":"eq"},"hits":[{"_id":"1","_source":{"name":"Alice","age":30,"active":true,"deleted_at":null,"profile":{"team":"core"},"tags":["admin","reader"]}},{"_id":"2","_routing":"shard-a","_source":{"name":"Bob","city":"NYC"}}]}}"#;
let result = super::parse_elasticsearch_rest_response(200, body, std::time::Instant::now()).unwrap();
assert_ne!(result.columns, vec!["status", "response"]);
assert!(result.columns.contains(&"name".to_string()));
assert!(result.columns.contains(&"_id".to_string()));
assert_eq!(result.rows.len(), 2);
let name_idx = result.columns.iter().position(|c| c == "name").unwrap();
let id_idx = result.columns.iter().position(|c| c == "_id").unwrap();
assert_eq!(result.rows[0][name_idx], json!("Alice"));
assert_eq!(result.rows[0][id_idx], json!("1"));
let age_idx = result.columns.iter().position(|c| c == "age").unwrap();
let active_idx = result.columns.iter().position(|c| c == "active").unwrap();
let deleted_at_idx = result.columns.iter().position(|c| c == "deleted_at").unwrap();
let profile_idx = result.columns.iter().position(|c| c == "profile").unwrap();
let tags_idx = result.columns.iter().position(|c| c == "tags").unwrap();
assert_eq!(result.rows[0][age_idx], json!(30));
assert_eq!(result.rows[0][active_idx], json!(true));
assert_eq!(result.rows[0][deleted_at_idx], serde_json::Value::Null);
assert_eq!(result.rows[0][profile_idx], json!(r#"{"team":"core"}"#));
assert_eq!(result.rows[0][tags_idx], json!(r#"["admin","reader"]"#));
assert_eq!(result.column_types[age_idx], "number");
assert_eq!(result.column_types[active_idx], "boolean");
assert_eq!(result.column_types[profile_idx], "json");
assert_eq!(result.column_types[tags_idx], "json");
let city_idx = result.columns.iter().position(|c| c == "city").unwrap();
assert_eq!(result.rows[1][city_idx], json!("NYC"));
let routing_idx = result.columns.iter().position(|c| c == "_routing").unwrap();
assert_eq!(result.rows[1][routing_idx], json!("shard-a"));
assert_eq!(result.affected_rows, 2);
assert_eq!(result.elasticsearch_raw_body.as_deref(), Some(body));
}
#[test]
fn parses_empty_search_rest_response_as_empty_table() {
let body = r#"{"took":1,"hits":{"total":{"value":0,"relation":"eq"},"hits":[]}}"#;
let result = super::parse_elasticsearch_rest_response(200, body, std::time::Instant::now()).unwrap();
assert_eq!(result.columns, vec!["_id"]);
assert!(result.rows.is_empty());
assert_eq!(result.affected_rows, 0);
}
#[test]
fn sparse_search_response_over_table_cell_limit_falls_back_to_raw_json() {
let hits = (0..450)
.map(|index| {
let mut source = serde_json::Map::new();
source.insert(format!("field_{index}"), json!(index));
json!({ "_id": index.to_string(), "_source": source })
})
.collect::<Vec<_>>();
let body = json!({ "hits": { "hits": hits } }).to_string();
let result = super::parse_elasticsearch_rest_response(200, &body, std::time::Instant::now()).unwrap();
assert_eq!(result.columns, vec!["status", "response"]);
assert_eq!(result.rows[0][0], json!(200));
assert_eq!(result.rows[0][1], json!(body));
assert_eq!(result.elasticsearch_raw_body, None);
}
#[tokio::test]
async fn execute_rest_query_keeps_json_error_response() {
use tokio::io::AsyncWriteExt;
let response_body = r#"{"error":{"type":"index_not_found_exception","reason":"no such index"},"status":404}"#;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("GET /missing/_mapping "));
let response = format!(
"HTTP/1.1 404 Not Found\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "GET /missing/_mapping").await.unwrap();
server.await.unwrap();
assert_eq!(result.columns, vec!["status", "response"]);
assert_eq!(result.rows[0][0], json!(404));
assert_eq!(
serde_json::from_str::<serde_json::Value>(result.rows[0][1].as_str().unwrap()).unwrap(),
json!({ "error": { "type": "index_not_found_exception", "reason": "no such index" }, "status": 404 })
);
}
#[tokio::test]
async fn execute_select_query_keeps_search_response_tabular() {
use tokio::io::AsyncWriteExt;
let response_body = r#"{"hits":{"hits":[{"_id":"product-1","_source":{"name":"Notebook"}}]}}"#;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("POST /products/_search "));
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
response_body.len(),
response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result = super::execute_rest_query(&client, "SELECT * FROM products LIMIT 1").await.unwrap();
server.await.unwrap();
assert_ne!(result.columns, vec!["status", "response"]);
let name_idx = result.columns.iter().position(|column| column == "name").unwrap();
assert_eq!(result.rows[0][name_idx], json!("Notebook"));
}
#[tokio::test]
async fn update_document_uses_legacy_type_path_without_storing_metadata() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("PUT /orders/order/abc?routing=tenant-1&refresh=true "));
assert!(request.ends_with(r#"{"name":"Alice"}"#));
assert!(!request.contains(r#""_type""#));
assert!(!request.contains(r#""_routing""#));
let response =
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: 2\r\nConnection: close\r\n\r\n{}";
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
super::update_document(
&client,
"orders",
"abc",
r#"{"_id":"abc","_type":"order","_routing":"tenant-1","name":"Alice"}"#,
None,
)
.await
.unwrap();
server.await.unwrap();
}
#[tokio::test]
async fn delete_document_uses_legacy_type_path() {
use tokio::io::AsyncWriteExt;
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let request = read_http_request(&mut socket).await;
assert!(request.starts_with("DELETE /orders/order/abc?routing=tenant-1&refresh=true "));
let response =
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: 2\r\nConnection: close\r\n\r\n{}";
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
super::delete_document(&client, "orders", "abc", Some("order"), Some("tenant-1")).await.unwrap();
server.await.unwrap();
}
#[tokio::test]
async fn execute_rest_search_preserves_full_json_response() {
use tokio::io::{AsyncReadExt, AsyncWriteExt};
let body = json!({
"took": 3,
"hits": {
"total": { "value": 1, "relation": "eq" },
"max_score": 1.0,
"hits": [{
"_index": "products",
"_id": "product-1",
"_score": 1.0,
"_source": { "name": "Notebook", "price": 1299 },
"highlight": { "name": ["<em>Note</em>book"] }
}]
},
"aggregations": {
"by_category": {
"buckets": [{ "key": "electronics", "doc_count": 1 }]
}
}
});
let response_body = body.to_string();
let server_response_body = response_body.clone();
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let server = tokio::spawn(async move {
let (mut socket, _) = listener.accept().await.unwrap();
let mut request = [0_u8; 4096];
let read = socket.read(&mut request).await.unwrap();
let request = String::from_utf8_lossy(&request[..read]);
assert!(request.starts_with("POST /products/_search "));
let response = format!(
"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{}",
server_response_body.len(),
server_response_body
);
socket.write_all(response.as_bytes()).await.unwrap();
});
let client = EsClient::new(&format!("http://{addr}"), None, None, false, Duration::from_secs(1));
let result =
super::execute_rest_query(&client, "POST /products/_search\n{\"query\":{\"match_all\":{}}}").await.unwrap();
server.await.unwrap();
// The response now parses hits+aggs into a tabular result (aggregation
// columns) instead of the raw status/response JSON panel.
assert!(result.columns.contains(&"key".to_string()));
assert!(result.columns.contains(&"doc_count".to_string()));
assert_ne!(result.columns, vec!["status", "response"]);
// Raw body is attached for JSON toggle in the UI.
assert!(result.elasticsearch_raw_body.is_some());
}
#[tokio::test]
#[ignore = "requires DBX_TEST_KIBANA_URL pointing to a reachable Kibana instance"]
async fn live_kibana_proxy_supports_metadata_queries_and_document_writes() {
let url = std::env::var("DBX_TEST_KIBANA_URL").expect("DBX_TEST_KIBANA_URL is required");
let username = std::env::var("DBX_TEST_KIBANA_USERNAME").ok();
let password = std::env::var("DBX_TEST_KIBANA_PASSWORD").ok();
let base_path = std::env::var("DBX_TEST_KIBANA_BASE_PATH").unwrap_or_default();
let external_config = json!({ "mode": "kibana", "kibanaBasePath": base_path });
let mut client = EsClient::from_config(
&url,
username.as_deref(),
password.as_deref(),
false,
None,
Some(&external_config),
Duration::from_secs(20),
None,
None,
None,
)
.expect("failed to build Elasticsearch test client");
super::test_connection(&mut client, Duration::from_secs(20)).await.unwrap();
let index = format!("dbx-kibana-proxy-{}", uuid::Uuid::new_v4().simple());
let create = super::execute_rest_query(
&client,
&format!(
"PUT /{index}\n{{\"mappings\":{{\"properties\":{{\"name\":{{\"type\":\"keyword\"}},\"price\":{{\"type\":\"double\"}}}}}}}}"
),
)
.await
.unwrap();
assert_eq!(create.rows[0][0], json!(200));
let id = super::insert_document(&client, &index, r#"{"name":"Notebook","price":12.5}"#, None).await.unwrap();
assert!(!id.is_empty());
assert!(super::list_indices(&client).await.unwrap().contains(&index));
let columns = super::get_columns(&client, &index).await.unwrap();
assert!(columns.iter().any(|column| column.name == "name" && column.data_type == "keyword"));
assert!(columns.iter().any(|column| column.name == "price" && column.data_type == "double"));
let documents = super::find_documents(&client, &index, 0, 10, None, None).await.unwrap();
assert_eq!(documents.total, 1);
assert_eq!(documents.documents[0]["name"], json!("Notebook"));
super::update_document(&client, &index, &id, r#"{"name":"Notebook Pro","price":15.0}"#, None).await.unwrap();
let sql_result = super::execute_rest_query(&client, &format!("SELECT * FROM {index} LIMIT 10")).await.unwrap();
let name_index = sql_result.columns.iter().position(|column| column == "name").unwrap();
assert_eq!(sql_result.rows[0][name_index], json!("Notebook Pro"));
super::delete_document(&client, &index, &id, None, None).await.unwrap();
let delete = super::execute_rest_query(&client, &format!("DELETE /{index}")).await.unwrap();
assert_eq!(delete.rows[0][0], json!(200));
}
#[test]
fn document_body_removes_elasticsearch_id_metadata() {
let (doc, _) = super::elasticsearch_document_body_and_routing_from_json(
r#"{"_id":"abc","_routing":"tenant-1","name":"Alice"}"#,
None,
)
.unwrap();
assert_eq!(doc, json!({ "name": "Alice" }));
}
#[test]
fn document_body_extracts_elasticsearch_routing_metadata() {
let (doc, routing) = super::elasticsearch_document_body_and_routing_from_json(
r#"{"_id":"abc","_routing":"tenant-1","name":"Alice"}"#,
None,
)
.unwrap();
assert_eq!(doc, json!({ "name": "Alice" }));
assert_eq!(routing.as_deref(), Some("tenant-1"));
}
#[test]
fn update_document_body_extracts_legacy_type_metadata() {
let (doc, routing, document_type) = super::elasticsearch_update_document_body_and_metadata(
r#"{"_id":"abc","_type":"order","_routing":"tenant-1","name":"Alice"}"#,
None,
)
.unwrap();
assert_eq!(doc, json!({ "name": "Alice" }));
assert_eq!(routing.as_deref(), Some("tenant-1"));
assert_eq!(document_type.as_deref(), Some("order"));
}
#[test]
fn explicit_elasticsearch_routing_overrides_document_metadata() {
let (doc, routing) = super::elasticsearch_document_body_and_routing_from_json(
r#"{"_id":"abc","_routing":"tenant-1","name":"Alice"}"#,
Some("tenant-2"),
)
.unwrap();
assert_eq!(doc, json!({ "name": "Alice" }));
assert_eq!(routing.as_deref(), Some("tenant-2"));
}
#[test]
fn document_body_preserves_user_field_order() {
let (doc, _) =
super::elasticsearch_document_body_and_routing_from_json(r#"{"z":1,"_id":"abc","a":2}"#, None).unwrap();
assert_eq!(serde_json::to_string(&doc).unwrap(), r#"{"z":1,"a":2}"#);
}
// Integration test against a real TLS-enabled Elasticsearch. Gated by env vars
// so it stays a no-op in CI:
// DBX_ELASTICSEARCH_TLS_TEST_URL e.g. https://localhost:9200
// DBX_ELASTICSEARCH_TLS_TEST_CA_CERT_PATH path to the server's CA cert (PEM)
// Mirrors the ZooKeeper TLS integration test: tls_enabled=true loads the custom
// CA (and any client cert) through apply_tls_certificates, then connects over HTTPS.
#[tokio::test]
async fn tls_integration_connects_with_custom_ca() {
let url = match std::env::var("DBX_ELASTICSEARCH_TLS_TEST_URL") {
Ok(u) => u,
Err(_) => {
eprintln!("skipping ES TLS integration test: DBX_ELASTICSEARCH_TLS_TEST_URL not set");
return;
}
};
let ca_cert_path = match std::env::var("DBX_ELASTICSEARCH_TLS_TEST_CA_CERT_PATH") {
Ok(c) => c,
Err(_) => {
eprintln!("skipping ES TLS integration test: DBX_ELASTICSEARCH_TLS_TEST_CA_CERT_PATH not set");
return;
}
};
// Optional basic-auth credentials (required when the server has security enabled).
let username = std::env::var("DBX_ELASTICSEARCH_TLS_TEST_USERNAME").ok();
let password = std::env::var("DBX_ELASTICSEARCH_TLS_TEST_PASSWORD").ok();
let timeout = std::time::Duration::from_secs(15);
let mut client = EsClient::from_config(
&url,
username.as_deref(),
password.as_deref(),
true,
None,
None,
timeout,
Some(&ca_cert_path),
None,
None,
)
.expect("failed to build Elasticsearch TLS client");
super::test_connection(&mut client, timeout)
.await
.expect("Elasticsearch TLS connection with custom CA should succeed");
}
}