Source map consumer: Search only the target line in originalPositionFor

[robhogan]

MappingsConsumer#originalPositionFor binary-searched every mapping in the map:

metro/packages/metro-source-map/src/Consumer/MappingsConsumer.js

Lines 54 to 94 in b7c2055

	originalPositionFor(
	generatedPosition: GeneratedPositionLookup,
	): SourcePosition {
	const {line, column} = generatedPosition;
	if (line == null || column == null) {
	return {...EMPTY_POSITION};
	}
	if (generatedPosition.bias != null) {
	invariant(
	generatedPosition.bias === GREATEST_LOWER_BOUND,
	`Unimplemented lookup bias: ${lookupBiasToString(
	// $FlowFixMe[incompatible-type]
	generatedPosition.bias,
	)}`,
	);
	}
	const mappings = this._decodeAndCacheMappings();
	const index = greatestLowerBound(
	mappings,
	{line, column},
	(position, mapping) => {
	if (position.line === mapping.generatedLine) {
	return get0(sub(position.column, mapping.generatedColumn));
	}
	return get0(sub(position.line, mapping.generatedLine));
	},
	);
	if (
	index != null &&
	mappings[index].generatedLine === generatedPosition.line
	) {
	const mapping = mappings[index];
	return {
	source: mapping.source,
	name: mapping.name,
	line: mapping.originalLine,
	column: mapping.originalColumn,
	};
	}
	return {...EMPTY_POSITION};
	}

This records the index of the first mapping on each generated line while decoding, and searches only that line. Results are unchanged, including a column before a line's first mapping (unmapped) and repeated generated columns (the last wins).

composeSourceMaps is 20.5% faster than #1979 (95% CI 20.1-21.1%) and 2.38x faster than main (1,107ms -> 878ms on our benchmark app), with identical output. Memory is up 1.2% (+11MB), most of it the per-line index over 1.18M generated lines.

Benchmark

AI-driven.

	Time (median)	vs previous	vs main	Composition memory	vs previous	vs main	Peak RSS
main	2,088ms (2,082 to 2,094)			1,070MB (1,070 to 1,070)			1,664MB
#1979	1,107ms (1,099 to 1,111)		-47.1% (-47.4 to -46.7)	898MB (897 to 899)		-15.9% (-16.1 to -15.6)	1,492MB
This PR	878ms (872 to 883)	-20.5% (-21.1 to -20.1)	-58.0% (-58.1 to -57.8)	908MB (908 to 909)	+1.2% (+1.0 to +1.3)	-15.1% (-15.2 to -15.0)	1,502MB

Changelog

 - **[Performance]**: Faster position lookups in `metro-source-map`'s `Consumer`

Test plan

New tests for a column before a line's first mapping and for repeated generated columns. They pass on main too.

Benchmark (AI-driven): Mattermost Mobile 2.45.0 (React Native 0.83.9, Metro 0.83.7), iOS release bundle, unminified: 7,086 sources, 52.6MB bundle, 82.4MB flat source map. Composed with its hermesc -O -output-source-map map (hermes-compiler 0.14.1, 10.7MB, 1.84M segments). Timings are composeSourceMaps alone, excluding JSON parsing, over 105 rounds on Node 22.13.1 on an M5 Pro. Each round runs main, every PR in this stack and a memory baseline, in a shuffled order. Figures are medians with 95% bootstrap confidence intervals, for changes of the per-round paired ratio. Composition memory is peak RSS less that of the same process parsing both maps without composing them (594MB).

[Copilot]

Copilot review overview

🟢 Approval recommended

No unresolved review issues were identified.

Review effort: Lite
Findings: None

What changed in this PR

Improves metro-source-map lookup performance by restricting searches to the requested generated line while preserving behavior.

Changes:

• Adds per-line mapping indexes during decoding.
• Uses line-scoped binary searches.
• Adds regression tests for lookup edge cases.

File	Description
packages/​metro-source-map/​src/​Consumer/​MappingsConsumer.js	Implements line-indexed position lookups.
packages/​metro-source-map/​src/​__tests__/​Consumer-test.js	Adds coverage for unmapped leading columns and duplicate generated columns.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.