You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Optimize tryToDoStringReplacementWithExtraIndentation with first-line fast bailout
Summary
• In packages/agent-runtime/src/generate-diffs-prompt.ts, optimized tryToDoStringReplacementWithExtraIndentation to eliminate redundant string splits and array joins during indentation mismatch recovery.
• Previously, the recovery loop evaluated 1..12 space indents followed by 1..6 tab indents (up to 18 iterations). On every iteration, it called searchContent.split('\n'), mapped over all lines, and joined them back together. For a 50-line block on non-matching text, this produced 18 full string splits and 900 line mappings.
• Pre-split searchLines once outside the loops.
• Added a fast-bailout check: if firstNonEmptyLine is present, it tests whether prefix + firstNonEmptyLine exists in oldFileContent. If the first indented line is not present, the full block cannot possibly match, bypassing string formatting and joining for that indentation level.
• Evaluated replaceContent.split('\n') lazily only when a match is found.
• Verified 100% exact parity across spaces, tabs, leading/trailing empty lines, and partial matches.
• Added unit test coverage for first-line matching with subsequent line mismatch, and leading empty lines in packages/agent-runtime/src/__tests__/generate-diffs-prompt.test.ts.
Good find. The fast bailout in generate-diffs-prompt.ts (checking prefix + firstNonEmptyLine before doing the full split/map/join) is sound: since the full indented block must contain that substring if it matches, skipping the expensive indentLines/includes work when the substring is absent can't produce a false negative — only a possible false positive that falls through to the real check, which is harmless.
The refactor into indentLines and hoisting searchLines out of the loop is a clean, low-risk change, and the two new tests (subsequent-line mismatch, leading empty lines) cover the interesting edge cases around firstNonEmptyLine detection. I traced through empty search content and all-empty-lines cases and didn't find a correctness gap.
Minor nit: the PR description claims replaceContent.split('\n') is now 'evaluated lazily only when a match is found,' but that was already true in the original code (it was inside the if block) — not a bug, just an inaccurate description of what changed. Worth fixing in the write-up but not the code.
This is small, in-scope (agent-runtime, not touching forbidden paths), has a plausible benchmark, and includes tests. Worth porting.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
codebuff-team
added
bot:triaged
Optimize tryToDoStringReplacementWithExtraIndentation with first-line fast bailout
Summary
• In
packages/agent-runtime/src/generate-diffs-prompt.ts, optimizedtryToDoStringReplacementWithExtraIndentationto eliminate redundant string splits and array joins during indentation mismatch recovery.• Previously, the recovery loop evaluated 1..12 space indents followed by 1..6 tab indents (up to 18 iterations). On every iteration, it called
searchContent.split('\n'), mapped over all lines, and joined them back together. For a 50-line block on non-matching text, this produced 18 full string splits and 900 line mappings.• Pre-split
searchLinesonce outside the loops.• Added a fast-bailout check: if
firstNonEmptyLineis present, it tests whetherprefix + firstNonEmptyLineexists inoldFileContent. If the first indented line is not present, the full block cannot possibly match, bypassing string formatting and joining for that indentation level.• Evaluated
replaceContent.split('\n')lazily only when a match is found.• Verified 100% exact parity across spaces, tabs, leading/trailing empty lines, and partial matches.
• Added unit test coverage for first-line matching with subsequent line mismatch, and leading empty lines in
packages/agent-runtime/src/__tests__/generate-diffs-prompt.test.ts.Verification & Benchmark Results
1. Benchmark
2. Test Suite & Hygiene
bun test packages/agent-runtime/src/__tests__/generate-diffs-prompt.test.tspassed 8/8 tests (0 fail).