[logaretm]

sqlite in Node only has the sqlite.db.query "profiling" channel, which can't give APMs real query spans, using a tracing channels allows for:

• Adds clear start event.
• Clearer span timings based on the actual operation runtime.
• Makes use of error channel, which allows APMs to mark failing query spans.
• Queries can be more easily parsed and sanitized.

This PR publishes run(), get(), all(), exec() and the tag store equivalents to a sqlite.query TracingChannel.

The tests check the context for statement, tag store and exec queries, error events for SQL that fails to compile in exec() and for runtime failures, that a store bound to start stays active during the query, that iterate() isn't traced, and that every query method publishes exactly once.

The end result is APM's and users can build spans/metrics/logs telemetry out of these channels easily with a simple subscriber. This is an example of a subscriber that creates spans out of it (using disk driver):

Regarding tag stores

For tracing tag store queries, we needed to cache the SQLTagStore template, since createTagStore() built a new one per store, so every store had its own prototype. Stores now share one prototype, which is visible from JS.

Happy to split this and tag store tracing into a follow-up PR if that's easier to review.

Benchmarks

ns/call, no subscribers	native	this PR
get()	336.7	330.4
run()	627.1	625.5
get(), cold	384.4	397.7
run() that throws	1817.0	1910.6
run() that throws, cold	1690.0	1875.8

Median of 30 interleaved runs on an M3 Pro, in-memory db, 10k calls, with a 300k-call warm-up except for the cold rows. "native" is the binding's native method, the one main puts on the prototype. Like #66575, calls that throw pay for the extra JS frame, about 90 ns once optimized.

'use strict';
// Usage: node --expose-internals --no-warnings sqlite-query-bench.js <native|pr> <get|run|throw> <cold|warm>
// Prints ns per call. "native" calls the binding's native method directly,
// which is what main installs on Statement.prototype.
const { internalBinding } = require('internal/test/binding');
const { Database, Statement } = require('node:sqlite');

const [variant, kind, temp] = process.argv.slice(2);
const method = kind === 'get' ? 'get' : 'run';

const db = new Database(':memory:');
db.exec('CREATE TABLE t (x INTEGER PRIMARY KEY); INSERT INTO t VALUES (1)');
const stmt = db.prepare({
  get: 'SELECT x FROM t WHERE x = ?',
  run: 'INSERT OR REPLACE INTO t VALUES (?)',
  throw: 'INSERT INTO t VALUES (?)', // Always violates the primary key.
}[kind]);

// Both variants are installed as own properties so lookups are identical.
stmt[method] = variant === 'native' ?
  internalBinding('sqlite').statementQueryMethods[method] :
  Statement.prototype[method];

function call() {
  try {
    stmt[method](1);
  } catch {
    // Expected for 'throw'.
  }
}

if (temp === 'warm') for (let i = 0; i < 3e5; i++) call();
const n = 1e4;
const t0 = process.hrtime.bigint();
for (let i = 0; i < n; i++) call();
console.log(Number(process.hrtime.bigint() - t0) / n);

Refs: #62241
Refs: #66575

[nodejs-github-bot]

Review requested:

• @nodejs/performance
• @nodejs/sqlite

[codecov]

Codecov Report

❌ Patch coverage is 93.78531% with 11 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.65%. Comparing base (d630b5d) to head (1d8998c).
⚠️ Report is 12 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_sqlite.cc	78.84%	4 Missing and 7 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #66581      +/-   ##
==========================================
+ Coverage   90.59%   90.65%   +0.06%     
==========================================
  Files         797      797              
  Lines      278230   278439     +209     
  Branches    53482    53549      +67     
==========================================
+ Hits       252066   252424     +358     
+ Misses      16558    16395     -163     
- Partials     9606     9620      +14     

Files with missing lines	Coverage Δ
lib/sqlite.js	100.00% <100.00%> (ø)
src/node_sqlite.h	87.27% <ø> (ø)
src/node_sqlite.cc	82.25% <78.84%> (+0.04%)	⬆️

... and 39 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Qard]

LGTM. The hasSubscribers checks could be simplified, but also fine to leave as-is. Using the shared one would check the other async channels too unnecessarily.