postgres statement cache behavior with traceparent comments?

[tmm1]

the postgres driver uses prepared statements to prevent sql injection. a new statement is created per query, and then cached for subsequent queries. this works as expected.

however, when you enable tracing it adds sql comments to your statements:

prisma-engines/query-engine/connectors/sql-query-connector/src/sql_trace.rs

Lines 34 to 37 in edd552c

	fn add_trace_id(self, trace_id: Option<&str>) -> Self {
	if let Some(traceparent) = trace_id {
	if should_sample(&traceparent) {
	self.comment(format!("traceparent='{}'", traceparent))

wouldn't this mean the statement cache, which uses the raw sql text as the cache key, would no longer be effective?

prisma-engines/quaint/src/connector/postgres/native/mod.rs

Lines 322 to 323 in edd552c

	async fn fetch_cached(&self, sql: &str, params: &[Value<'_>]) -> crate::Result<Statement> {
	let statement_key = StatementKey::new(sql, params);

prisma-engines/quaint/src/connector/postgres/native/mod.rs

Lines 85 to 86 in edd552c

	let mut hasher = DefaultHasher::new();
	sql.hash(&mut hasher);

cc @aqrln

[tmm1]

maybe cache misses are inevitable here because these are meant to be distinct queries sent the to the server. but i wonder if there's a way to put the comment on just the EXECUTE part and not the PREPARE

[aqrln]

Good question. I'm gonna pull this issue into the ongoing Tracing GA project and think about it.

[aqrln]

By the way let me know if you run into any other issues related to tracing, this is super relevant right now, and we will tackle them asap. A good chunk of important bugfixes might hopefully make it into the next release, and more are coming soon.

[jacek-prisma]

Fixed in #5082