World's first open-source enterprise world model.
Rust agent-memorybitemporalgraphragknowledge-baseknowledge-graphllmontologypgvectorpostgresqlragrustself-hostedsemantic-searchtemporal-knowledge-graphworld-model
A date the text spells out is read as a date (#689)
WaylandYang0306d0cf70b5
diff --git a/crates/utopia-extract/src/lib.rs b/crates/utopia-extract/src/lib.rsindex 6e1ca4c8..713bd86c 100644--- a/crates/utopia-extract/src/lib.rs+++ b/crates/utopia-extract/src/lib.rs@@ -295,7 +295,7 @@ pub fn build_messages_with_opening(let attr_rules = if attributes.is_empty() {String::new()} else {- "\n10. Attribute facts carry \"value\" (no \"object\"): number = the figure **as the text writes it, magnitude and currency included** \n (\"86亿元\", \"$5 billion\", \"4,300 人\") — never reduce it to a bare number, the server converts; date = \"YYYY[-MM[-DD]]\" (a zoned clock time only when the text gives one) — a date the text gives only relative to an event \+ "\n11. Attribute facts carry \"value\" (no \"object\"): number = the figure **as the text writes it, magnitude and currency included** \n (\"86亿元\", \"$5 billion\", \"4,300 人\") — never reduce it to a bare number, the server converts; date = \"YYYY[-MM[-DD]]\" (a zoned clock time only when the text gives one) — a date the text gives only relative to an event \(\"45 days after the Trigger Date\", \"within 30 days of closing\") has no calendar date to convert: write it as the text writes it and add \"relative\": true; bool = true/false; \text = a short string. Only attach an attribute to a subject of its listed class. \valid_from = when this value took effect, if the text or the opening of the document says so. \@@ -380,15 +380,14 @@ pub fn build_messages_with_opening(fits — \"purchase_price\", \"job_title\", \"generation_capacity\", \"record_date\". \Attach it to the entity the text attaches it to, and keep the literal as written, \units and all — except a date, which is always written in the format of rule 3 \- (\"June 23, 2020\" is \"2020-06-23\"): the server keeps a date only in that \- format, and a date written any other way is lost. A deadline or a period stated \+ (\"June 23, 2020\" is \"2020-06-23\"). A deadline or a period stated \relative to an event, with no calendar date, is not a date: keep it as written \- and mark it \"relative\" as rule 10 says. \+ and mark it \"relative\" as rule 11 says. \**A stated figure left out is the loss that costs most**: the reader \came for those numbers, and no later step can recover one that was never written \down.\n\- 8c. A listed relation followed by {{…}} can carry those **qualifiers on the edge**: when the same sentence gives both the other entity and a figure for it — an amount, a stake, a price, a share count — write the relation with its \"object\" and put the figure in \"qualifiers\" keyed exactly as listed, **as written in the text, currency and all** (\"€30 million\", \"15亿元人民币\", never a bare number) — except a date, which takes the format of rule 3: {{\"subject\":\"Vega Capital\",\"predicate\":\"invested_in\",\"object\":\"Northwind\", \"qualifiers\":{{\"amount\":\"$5 billion\"}},…}}. Never invent a key that is not listed for that relation, and never drop the figure to keep the edge — a relation without its amount is half the sentence. A relation you name after the text (rule 8) carries its figure the same way — keyed by the listed attribute that fits it, or by the plainest word for it (\"amount\", \"stake\", \"price\") when none does.- 8b. A **listed** relation also takes \"value\" when what the text gives is a \+ 8b. A listed relation followed by {{…}} can carry those **qualifiers on the edge**: when the same sentence gives both the other entity and a figure for it — an amount, a stake, a price, a share count — write the relation with its \"object\" and put the figure in \"qualifiers\" keyed exactly as listed, **as written in the text, currency and all** (\"€30 million\", \"15亿元人民币\", never a bare number) — except a date, which takes the format of rule 3: {{\"subject\":\"Vega Capital\",\"predicate\":\"invested_in\",\"object\":\"Northwind\", \"qualifiers\":{{\"amount\":\"$5 billion\"}},…}}. Never invent a key that is not listed for that relation, and never drop the figure to keep the edge — a relation without its amount is half the sentence. A relation you name after the text (rule 8) carries its figure the same way — keyed by the listed attribute that fits it, or by the plainest word for it (\"amount\", \"stake\", \"price\") when none does.+ 8c. A **listed** relation also takes \"value\" when what the text gives is a \string rather than another entity — a job title, a designation, a ticker, a \model number. Never invent an entity for a string. And when the text introduces \someone by their role — \"X, founder and CEO of Y\", \"Z, co-CEO of W\", \@@ -402,15 +401,15 @@ pub fn build_messages_with_opening((\"appointed … as OpenAI's CTO of applications\") all carry the same \shape — the role is the value, the organization is the other \entity. Past tense and \"former\" give the tie valid_to: \"unknown\".\n\- 8c. A list of named parties is a list of facts — one per name. \"partners \+ 8d. A list of named parties is a list of facts — one per name. \"partners \including A, B, C and D\" is four facts, not one; \"advisors A and B\" is two. \Do not collapse an enumeration into a summary or into its first member. \The same applies to the entities: each named party is its own entity.\n\- 8d. subject_span and object_span are the exact words in quote that name each side. \+ 8e. subject_span and object_span are the exact words in quote that name each side. \Copy them; never paraphrase. When the words that do the thing are a description \rather than a name — \"former X employees\", \"companies using X\" — the span \is that description, whatever you wrote in subject.\n\- 8e. An obligation, a deadline or a right belongs to the agreement, law or decision \+ 8f. An obligation, a deadline or a right belongs to the agreement, law or decision \that imposes it, even when it concerns another agreement or thing. A lease that \sets the last day to sign a second lease gives that deadline to the first lease; \the second lease is only what the deadline is about.\n\@@ -1071,7 +1070,8 @@ pub fn attr_object_value(}/// 属性值按 datatype 归一。失败返回 None——宁缺勿脏,调用方跳过并记日志。-/// number 容忍千分位/空格;date 要求 YYYY[-MM[-DD]] 且保留原精度;bool 宽容 yes/no。+/// number 容忍千分位/空格;date 收规则 3 的格式(YYYY[-MM[-DD]]、带时区的时刻,原样保留),+/// 也收写法说得清是哪天的日期([`written_date`]),换成规则 3 的样子;bool 宽容 yes/no。pub fn normalize_attr_value(datatype: &str, raw: &serde_json::Value) -> Option<serde_json::Value> {match datatype {"number" => match raw {@@ -1104,9 +1104,17 @@ pub fn normalize_attr_value(datatype: &str, raw: &serde_json::Value) -> Option<s}_ => None,},+ // 按规则 3 写的原样留着(精度随写了几位,带时区的时刻也在内);写成别的样子、又读得+ // 出来的日期(#688)换成规则 3 的样子——同一天只该有一种写法,比较和去重才对得上"date" => {let s = raw.as_str()?.trim();- parse_time(s).map(|_| serde_json::Value::String(s.to_string()))+ match written_date(s) {+ Some((date, precision)) => Some(serde_json::Value::String(match precision {+ "month" => date.format("%Y-%m").to_string(),+ _ => date.format("%Y-%m-%d").to_string(),+ })),+ None => parse_time(s).map(|_| serde_json::Value::String(s.to_string())),+ }}"bool" => match raw {serde_json::Value::Bool(b) => Some(serde_json::Value::Bool(*b)),@@ -1134,6 +1142,8 @@ pub fn normalize_attr_value(datatype: &str, raw: &serde_json::Value) -> Option<s/// `YYYY-MM-DDTHH[:MM[:SS]]` 后跟 `Z` 或 `±HH:MM`,精度到 hour / minute / second,/// 值截到那一位。**没有时区的钟点不是时刻**——「14:32」是哪里的 14:32 没人知道——/// 所以只取日期那一半,按天;钟点留在引文里。亚秒一律丢:账本到秒为止。+///+/// 这是规则 3 的契约格式,工具参数、界面上的时刻都只认它。读模型回复用 [`read_time`]pub fn parse_time(s: &str) -> Option<(DateTime<Utc>, &'static str)> {let s = s.trim();if s.is_empty() || s.eq_ignore_ascii_case("null") {@@ -1171,6 +1181,94 @@ pub fn parse_time(s: &str) -> Option<(DateTime<Utc>, &'static str)> {None}+/// 读模型回复里的时间:先按规则 3([`parse_time`]);没照它写、但写法说得清是哪天的日期+/// 也收([`written_date`],#688)。区间端点、日期属性、边上的日期属性都从这里读+pub fn read_time(s: &str) -> Option<(DateTime<Utc>, &'static str)> {+ parse_time(s).or_else(|| {+ let (date, precision) = written_date(s)?;+ Some((+ Utc.from_utc_datetime(&date.and_hms_opt(0, 0, 0)?),+ precision,+ ))+ })+}++/// 没照规则 3 写、但说得清是哪一天(或哪个月)的日期(#688)。+///+/// 合同、公告里的日期多半这么写,模型常常照抄;从前这些值全被当成「不是日期」丢掉。+/// 收两类,精度随写了几位——只写到月的就是月,不替它补一个日:+/// - 月份写成名字的:`March 17, 2020`、`17 March 2020`、`Mar. 17 2020`、`March 2020`。+/// 名字由 chrono 的 `%B` / `%b` 认(整名或三个字母的缩写,不分大小写)+/// - 年在前的数字:`2020/03/17`、`2020.3.17`、`2020年3月17日`、`2020年3月`+///+/// **日、月都是数字而年不在前的不收**:`03/04/2020` 是三月四日还是四月三日,写法本身说+/// 不清,猜错一次就是一个错的截止日。+pub fn written_date(s: &str) -> Option<(NaiveDate, &'static str)> {+ let s = s.trim();+ if let Some(found) = year_first_date(s) {+ return found;+ }+ // 月份写成名字的:句点(缩写后面那个)与逗号只是标点+ let words: Vec<&str> = s+ .split(|c: char| c.is_whitespace() || c == ',' || c == '.')+ .filter(|w| !w.is_empty())+ .collect();+ let is_number = |w: &str| w.chars().all(|c| c.is_ascii_digit());+ let (day, month, year) = match words.as_slice() {+ [m, d, y] if !is_number(m) && is_number(d) && is_number(y) => (Some(*d), *m, *y),+ [d, m, y] if is_number(d) && !is_number(m) && is_number(y) => (Some(*d), *m, *y),+ [m, y] if !is_number(m) && is_number(y) => (None, *m, *y),+ _ => return None,+ };+ if year.len() != 4 || day.is_some_and(|d| d.len() > 2) {+ return None;+ }+ let month = ["%B", "%b"].iter().find_map(|f| {+ NaiveDate::parse_from_str(&format!("1 {month} 2000"), &format!("%d {f} %Y"))+ .ok()+ .map(|d| chrono::Datelike::month(&d))+ })?;+ let year = year.parse().ok()?;+ match day {+ Some(d) => NaiveDate::from_ymd_opt(year, month, d.parse().ok()?).map(|date| (date, "day")),+ None => NaiveDate::from_ymd_opt(year, month, 1).map(|date| (date, "month")),+ }+}++/// 年在前的数字日期。外层 `None` = 不是这种写法,交给下一种;`Some(None)` = 是这种写法+/// 但不是真实的日子+fn year_first_date(s: &str) -> Option<Option<(NaiveDate, &'static str)>> {+ let marked = s.contains('年');+ let separated = s.contains(['/', '.']) && !s.contains(char::is_whitespace);+ if !marked && !separated {+ return None;+ }+ let parts: Vec<&str> = s+ .trim_end_matches('日')+ .split(['/', '.', '年', '月'])+ .filter(|p| !p.is_empty())+ .collect();+ if !parts.iter().all(|p| p.chars().all(|c| c.is_ascii_digit())) {+ return None;+ }+ if parts.first().is_none_or(|y| y.len() != 4) {+ // 年不在前:日月顺序说不清,不交给别的写法去猜+ return Some(None);+ }+ let number = |p: &str| p.parse::<u32>().ok();+ let year = parts[0].parse::<i32>().ok()?;+ Some(match parts[1..] {+ [m, d] if m.len() <= 2 && d.len() <= 2 => {+ NaiveDate::from_ymd_opt(year, number(m)?, number(d)?).map(|date| (date, "day"))+ }+ // 只到月:写了「年」「月」才算(`2020/03` 太像别的东西)+ [m] if marked && s.ends_with('月') && m.len() <= 2 => {+ NaiveDate::from_ymd_opt(year, number(m)?, 1).map(|date| (date, "month"))+ }+ _ => None,+ })+}+/// 钟点后面的时区:`Z` 或 `±HH[:]MM` / `±HH`。返回 (钟点, 相对 UTC 的偏移);/// 没有时区返回 None——调用方据此只记那一天fn split_zone(clock: &str) -> Option<(&str, chrono::Duration)> {@@ -1377,6 +1475,48 @@ mod prompt_shape_tests {assert!(system.contains("except a date, which is always written in the format of rule 3"));}+ /// 规则编号各不相同,「按规则 N」指得到唯一的一条。从前有两条 8c、两条 10,+ /// 「as rule 10 says」说的是哪条要靠猜(#689 评审)+ #[test]+ fn every_rule_has_its_own_number_and_every_reference_lands() {+ let rels = vec![PromptRelation {+ key: "acquired".into(),+ label: "acquired".into(),+ description: String::new(),+ signature: String::new(),+ temporal: "event".into(),+ qualifiers: vec![],+ }];+ let attrs = vec!["lease.option_deadline (date)".to_string()];+ let msgs = build_messages(&[], &rels, &attrs, None, "a.txt", &[], "text");+ let system = &msgs[0].content;+ let mut labels = Vec::new();+ for line in system.lines() {+ let Some((label, _)) = line.trim_start().split_once(". ") else {+ continue;+ };+ let digits = label.trim_end_matches(|c: char| c.is_ascii_lowercase());+ if !digits.is_empty()+ && digits.chars().all(|c| c.is_ascii_digit())+ && label.len() - digits.len() <= 1+ {+ labels.push(label.to_string());+ }+ }+ let unique: std::collections::BTreeSet<_> = labels.iter().collect();+ assert_eq!(unique.len(), labels.len(), "规则编号重复:{labels:?}");+ for (i, _) in system.match_indices("rule ") {+ let n: String = system[i + 5..]+ .chars()+ .take_while(|c| c.is_ascii_alphanumeric())+ .collect();+ assert!(+ labels.contains(&n),+ "「rule {n}」指不到任何一条:{labels:?}"+ );+ }+ }+#[test]fn a_later_chunk_reads_the_opening_of_its_document() {let opening = "FIFTH AMENDMENT TO LEASE AGREEMENT entered into as of February 18, 2020";@@ -1694,6 +1834,71 @@ mod tests {assert!(parse_time("null").is_none());assert!(parse_time("").is_none());assert!(parse_time("下个月").is_none());+ // 契约格式之外的写法不归它:工具参数里的「August 2024」不猜+ assert!(parse_time("June 23, 2020").is_none());+ // 读模型回复的那一个收写出来的日期(#688),精度随写了几位+ let (t, p) = read_time("June 23, 2020").unwrap();+ assert_eq!(+ (t.to_rfc3339(), p),+ ("2020-06-23T00:00:00+00:00".to_string(), "day")+ );+ assert_eq!(read_time("March 2020").unwrap().1, "month");+ assert_eq!(read_time("2024-07").unwrap().1, "month");+ assert!(read_time("03/04/2020").is_none());+ }++ /// 合同与公告里的日期写法(#688):说得清是哪天的都收成规则 3 的样子,说不清的不猜+ #[test]+ fn a_written_date_is_read_only_when_its_form_says_which_day() {+ use serde_json::json;+ let day = |s: &str| normalize_attr_value("date", &json!(s));+ for written in [+ "March 17, 2020",+ "March 17 2020",+ "march 17, 2020",+ "MARCH 17, 2020",+ "Mar 17, 2020",+ "Mar. 17, 2020",+ "17 March 2020",+ "17 Mar. 2020",+ "17 March, 2020",+ " March 17, 2020 ",+ "2020/03/17",+ "2020.3.17",+ "2020年3月17日",+ ] {+ assert_eq!(day(written), Some(json!("2020-03-17")), "{written}");+ }+ // 只写到月的是月,不补日+ for written in ["March 2020", "Mar. 2020", "2020年3月"] {+ assert_eq!(day(written), Some(json!("2020-03")), "{written}");+ }+ // 已经照规则 3 写的原样留着+ assert_eq!(day("2020-03-17"), Some(json!("2020-03-17")));+ assert_eq!(day("2020"), Some(json!("2020")));+ // 日月都是数字、年不在前:说不清是几月几号+ for ambiguous in ["03/04/2020", "3.4.2020", "04-03-2020", "3/4/20"] {+ assert_eq!(day(ambiguous), None, "{ambiguous}");+ }+ // 不是日期,或不是一个真实的日子+ for not_a_date in [+ "45 days after the Trigger Date",+ "Q3 2020",+ "Sometime 2020",+ "February 30, 2020",+ "March 17, 20",+ "March 123, 2020",+ "2020/13/01",+ "2020/03",+ "next March",+ ] {+ assert_eq!(day(not_a_date), None, "{not_a_date}");+ }+ // 区间端点读的是同一个解析+ assert_eq!(+ read_time("17 Mar 2020").map(|(t, p)| (t.date_naive().to_string(), p)),+ Some(("2020-03-17".to_string(), "day"))+ );}#[test]diff --git a/crates/utopia-extract/src/normalize.rs b/crates/utopia-extract/src/normalize.rsindex e26f9df7..e2abbae4 100644--- a/crates/utopia-extract/src/normalize.rs+++ b/crates/utopia-extract/src/normalize.rs@@ -9,7 +9,7 @@//! 每条规则做了什么都返回给服务端记进丢弃表:违约多常见、出在哪个模型,量得出来,//! 契约该怎么改看数说话。-use crate::{parse_time, ExtractedEntity, ExtractedFact, Extraction};+use crate::{read_time, ExtractedEntity, ExtractedFact, Extraction};use std::collections::HashSet;/// 形状检查做了什么;服务端按条记信号@@ -229,9 +229,10 @@ fn value_fact(}}-/// 一侧写的是契约格式的时间(`YYYY` / `YYYY-MM` / `YYYY-MM-DD`,带时区的时刻)-fn is_contract_time(s: &str) -> bool {- parse_time(s.trim()).is_some()+/// 一侧写的是一个时间:规则 3 的格式(`YYYY` / `YYYY-MM` / `YYYY-MM-DD`,带时区的时刻),+/// 或写法说得清是哪天的日期(`written_date`,#688)+fn names_a_time(s: &str) -> bool {+ read_time(s.trim()).is_some()}/// 引文不在这一块、却在附上的文件开头里的事实:丢掉,连同因此没人引用的声明。@@ -355,7 +356,7 @@ pub fn normalize_facts(x: &mut Extraction) -> Vec<Normalization> {// ---- 主语是时间 ----let subject =handle_name(f.subject_ref.as_ref()).unwrap_or_else(|| f.subject.trim().to_string());- if is_contract_time(&subject) {+ if names_a_time(&subject) {out.push(Normalization::TimeAsSubject {predicate: f.predicate.clone(),written: subject,@@ -383,7 +384,7 @@ pub fn normalize_facts(x: &mut Extraction) -> Vec<Normalization> {}// ---- 宾语是时间 ----- if let Some(o) = object.as_deref().filter(|o| is_contract_time(o)) {+ if let Some(o) = object.as_deref().filter(|o| names_a_time(o)) {if values.is_empty() {// 没带数:写出来的那段就是值。`2028`(「2028 年起上线」)、`4000`(人数)都解析// 得成年份,从前整条丢掉,一条信息就没了。宾语那个声明没人引用,下面按孤点去掉diff --git a/crates/utopia-server/src/extraction.rs b/crates/utopia-server/src/extraction.rsindex 3089bf35..e54a12f7 100644--- a/crates/utopia-server/src/extraction.rs+++ b/crates/utopia-server/src/extraction.rs@@ -1583,30 +1583,8 @@ async fn run(state: &AppState, document_id: Uuid, proposer: Proposer) -> anyhow:.await;continue;}- let from = f.valid_from.as_deref().and_then(utopia_extract::parse_time);- let to = f.valid_to.as_deref().and_then(utopia_extract::parse_time);- // **两端各记各的粒度**(见 `facts.valid_to_precision`)。从前一个精度列描述两个端点,- // 于是「2020 年开始、2023-05-06 结束」这种只能共用一个值。- //- // 模型给的 valid_to = "unknown" 表示**原文说它结束了、但没说哪天**。- // parse_time 解不出它(本来就不是日期),落在这里显式认掉——- // 不认的话它退化成 None,那条事实就又变回"仍在持续"了- let ended_unknown = f- .valid_to- .as_deref()- .map(str::trim)- .is_some_and(|v| v.eq_ignore_ascii_case(utopia_store::graph::ENDED_UNKNOWN));- let validity = utopia_store::graph::Validity {- from: from.map(|(t, _)| t),- from_precision: from.map(|(_, p)| p),- to: to.map(|(t, _)| t),- to_precision: to- .map(|(_, p)| p)- .or(ended_unknown.then_some(utopia_store::graph::ENDED_UNKNOWN)),- // 这次观察出自哪一天的文档(0022):没起点的事实从它起成立,结束了- // 不知哪天的到它为止。没有文档日期就是记下的此刻——账本能给的最好的- attested_at: doc.doc_time,- };+ let validity =+ validity_of(f.valid_from.as_deref(), f.valid_to.as_deref(), doc.doc_time);// 属性事实:谓词命中属性 → 字面值通道。datatype 校验失败宁缺勿脏;// domain 校验(含子类上溯)挡住"把 salary 挂到 Organization"这类张冠李戴。@@ -2946,6 +2924,38 @@ fn undeclared_beside_value(////// "2015"、"2023-03"、"6" 认;"杭州"、"首席技术官"、"3M"、"V3" 不认。/// 调用方还额外要求模型**没有**把它声明成实体——两道门一起过才算数。+/// 模型给的区间两端 → 落库的有效区间。+///+/// **两端各记各的粒度**(见 `facts.valid_to_precision`)。从前一个精度列描述两个端点,+/// 于是「2020 年开始、2023-05-06 结束」这种只能共用一个值。两端都用 `read_time` 读:+/// 规则 3 的格式,或写法说得清是哪天的日期(#688),精度随写了几位。+///+/// 模型给的 valid_to = "unknown" 表示**原文说它结束了、但没说哪天**。+/// read_time 解不出它(本来就不是日期),落在这里显式认掉——+/// 不认的话它退化成 None,那条事实就又变回"仍在持续"了+fn validity_of(+ valid_from: Option<&str>,+ valid_to: Option<&str>,+ doc_time: Option<chrono::DateTime<chrono::Utc>>,+) -> utopia_store::graph::Validity<'static> {+ let from = valid_from.and_then(utopia_extract::read_time);+ let to = valid_to.and_then(utopia_extract::read_time);+ let ended_unknown = valid_to+ .map(str::trim)+ .is_some_and(|v| v.eq_ignore_ascii_case(utopia_store::graph::ENDED_UNKNOWN));+ utopia_store::graph::Validity {+ from: from.map(|(t, _)| t),+ from_precision: from.map(|(_, p)| p),+ to: to.map(|(t, _)| t),+ to_precision: to+ .map(|(_, p)| p)+ .or(ended_unknown.then_some(utopia_store::graph::ENDED_UNKNOWN)),+ // 这次观察出自哪一天的文档(0022):没起点的事实从它起成立,结束了+ // 不知哪天的到它为止。没有文档日期就是记下的此刻——账本能给的最好的+ attested_at: doc_time,+ }+}+fn looks_literal(s: &str) -> bool {let s = s.trim();if s.is_empty() {@@ -2966,8 +2976,8 @@ fn looks_literal(s: &str) -> bool {if utopia_extract::parse_quantity(s).is_some() {return true;}- // 日期:复用抽取侧那个解析器,它认 2015 / 2015-03 / 2015-03-01 等- utopia_extract::parse_time(s).is_some()+ // 日期:复用抽取侧那个解析器,它认 2015 / 2015-03 / 2015-03-01,也认写出来的日期(#688)+ utopia_extract::read_time(s).is_some()}/// 提示词里那三段清单:类、关系、属性。@@ -3939,7 +3949,7 @@ mod tests {use super::{incomplete_reason, looks_literal, no_ref_name_binding, referenced_entity, resolve_bare,- resolve_handle, BoundEntity, NoRefNameBinding,+ resolve_handle, validity_of, BoundEntity, NoRefNameBinding,};use std::collections::HashMap;use uuid::Uuid;@@ -3960,6 +3970,10 @@ mod tests {"€1.5 million","52%","35,000",+ // 合同照原文写的日期(#688)+ "June 23, 2020",+ "17 Mar. 2020",+ "2020年3月17日",] {assert!(looks_literal(yes), "{yes} 该认成字面值");}@@ -3983,6 +3997,31 @@ mod tests {}}+ /// 区间两端照合同原文写(#688):读成日期,精度随写了几位;「unknown」仍是结束了不知哪天+ #[test]+ fn a_written_start_keeps_the_precision_it_was_written_with() {+ let day = validity_of(Some("June 8, 2020"), None, None);+ assert_eq!(+ day.from.map(|t| t.date_naive().to_string()).as_deref(),+ Some("2020-06-08")+ );+ assert_eq!(day.from_precision, Some("day"));+ assert!(!day.has_ended());++ let month = validity_of(Some("March 2020"), Some("17 Mar. 2021"), None);+ assert_eq!(month.from_precision, Some("month"));+ assert_eq!(+ month.to.map(|t| t.date_naive().to_string()).as_deref(),+ Some("2021-03-17")+ );+ assert_eq!(month.to_precision, Some("day"));++ // 说不清几月几号的不当起点+ let ambiguous = validity_of(Some("03/04/2020"), Some("unknown"), None);+ assert_eq!((ambiguous.from, ambiguous.from_precision), (None, None));+ assert_eq!(ambiguous.to_precision, Some("unknown"));+ }+#[test]fn no_ref_names_bypass_legacy_map_only_after_two_handle_claims() {let (a, b) = (Uuid::now_v7(), Uuid::now_v7());