:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
use std::ops::Range;
use chrono::{DateTime, Datelike, SecondsFormat, Utc};
use serde::Serialize;
use thiserror::Error;
use usage_ingestion_proto::usage_ingestion::v1::BillingUsageRecord;
use uuid::Uuid;
const CLICKHOUSE_DATETIME64_YEARS: Range<i32> = 1900..2300;
/// A record_id may have to mirror a whole dedup identity: the analytics producers compose
/// theirs from the events table's sorting key, whose event name and distinct_id are each
/// capped at 200. Rejecting one would drop the record and under-bill.
const MAX_IDENTIFIER_LENGTH: usize = 512;
#[derive(Debug, Error, PartialEq, Eq)]
pub enum ValidationError {
#[error("{0} must not be empty")]
Empty(&'static str),
#[error("{0} exceeds the maximum length")]
TooLong(&'static str),
#[error("team_id must be positive")]
InvalidTeamId,
#[error("quantity must be positive")]
InvalidQuantity,
#[error("timestamp_ms must be milliseconds since the epoch, between years 1900 and 2300")]
InvalidTimestamp,
}
#[derive(Debug, Serialize)]
pub struct KafkaBillingUsageRecord {
pub schema_version: u8,
pub record_id: String,
pub producer_id: String,
pub team_id: i64,
pub organization_id: Uuid,
pub usage_key: String,
pub unit: String,
pub quantity: i64,
pub timestamp: String,
pub inserted_at: String,
#[serde(skip)]
pub usage_timestamp: DateTime<Utc>,
}
impl KafkaBillingUsageRecord {
pub fn from_proto(
record: BillingUsageRecord,
organization_id: Uuid,
inserted_at: DateTime<Utc>,
) -> Result<Self, ValidationError> {
validate_identifier("record_id", &record.record_id)?;
validate_identifier("producer_id", &record.producer_id)?;
validate_identifier("usage_key", &record.usage_key)?;
validate_identifier("unit", &record.unit)?;
if record.team_id <= 0 {
return Err(ValidationError::InvalidTeamId);
}
// Every record is a delta, so nothing meaningful is zero. A snapshot producer would
// need this exemption back alongside the mode column.
if record.quantity <= 0 {
return Err(ValidationError::InvalidQuantity);
}
// This protects the Kafka engine rather than defining a billing-time policy. Tighten the
// accepted range when producers have an explicit backfill and future-skew contract.
let timestamp = DateTime::from_timestamp_millis(record.timestamp_ms)
.filter(|value| CLICKHOUSE_DATETIME64_YEARS.contains(&value.year()))
.ok_or(ValidationError::InvalidTimestamp)?;
Ok(Self {
schema_version: 1,
record_id: record.record_id,
producer_id: record.producer_id,
team_id: record.team_id,
organization_id,
usage_key: record.usage_key,
unit: record.unit,
quantity: record.quantity,
timestamp: timestamp.to_rfc3339_opts(SecondsFormat::Millis, true),
inserted_at: inserted_at.to_rfc3339_opts(SecondsFormat::Millis, true),
usage_timestamp: timestamp,
})
}
}
fn validate_identifier(field: &'static str, value: &str) -> Result<(), ValidationError> {
if value.is_empty() {
return Err(ValidationError::Empty(field));
}
if value.len() > MAX_IDENTIFIER_LENGTH {
return Err(ValidationError::TooLong(field));
}
Ok(())
}