Found multiple performance optimizations for Rewrite

[misrasaurabh1]

What problem are you trying to solve?

Improve speed for Rewrite by merging in optimizations we created with Codeflash

Describe the solution you'd like

Here are the Pull Requests we created and reviewed manually with the optimizations. I have ordered them roughly in the order of ease of merge and impact:

• ⚡️ Speed up method StringUtils.formatUriForPropertiesFile by 265% HeshamHM28/rewrite#54 - Simple speed up for string operations
• ⚡️ Speed up method StringUtils.indexOfNextNonWhitespace by 57% HeshamHM28/rewrite#53 - Speeds up string operations
• ⚡️ Speed up method LocalRemoteArtifactCache.hashUri by 697% HeshamHM28/rewrite#27 - HashURI speedup
• ⚡️ Speed up method StringUtils.indent by 161% HeshamHM28/rewrite#49 - String indent speedup
• ⚡️ Speed up method Cursor.toString by 75% HeshamHM28/rewrite#16 - Cursor.toString speedup
• ⚡️ Speed up method Cursor.firstEnclosing by 94% HeshamHM28/rewrite#15 - Cursor firstEnclosing speedup
• ⚡️ Speed up method ExceptionUtils.containsCircularReferences by 127% HeshamHM28/rewrite#33 - Exception circularReferences
• ⚡️ Speed up method StringUtils.isBlank by 139% HeshamHM28/rewrite#43 - String isBlank whitespace
• ⚡️ Speed up method PathUtils.getEitherOrPatterns by 209% HeshamHM28/rewrite#38 - PathUtils speedups
• ⚡️ Speed up method StringUtils.capitalize by 15% HeshamHM28/rewrite#45 - String capitalize
• ⚡️ Speed up method StringUtils.uncapitalize by 7% HeshamHM28/rewrite#46 - String uncapitalize
• ⚡️ Speed up method RewriteRpcProcess.getLivenessCheck by 38% HeshamHM28/rewrite#19 - RewriteRpcProcess.getLivenessCheck speedup
• ⚡️ Speed up method StringUtils.aspectjNameToPattern by 15% HeshamHM28/rewrite#51 - Speed up string operations
• ⚡️ Speed up method GitRemote.hashCode by 114% HeshamHM28/rewrite#10 - speedup hashCode with a more optimized implementation
• ⚡️ Speed up method MetricsHelper.successTags by 116% HeshamHM28/rewrite#28 - simple speedup
• ⚡️ Speed up method StringUtils.commonMargin by 12% HeshamHM28/rewrite#50 - speeds up string operations by reducing redundant computations

Can you let me know which ones can I create here.

Additional context

We have improved codeflash for java significantly since the runs above last month, and now we can optimize end-to-end performance benchmarks as well. Let me know if you have any performance benchmarks that you would want optimizations for?

[timtebeek]

Analysis: Prioritizing optimizations by real-world value

I went through each of the 16 PRs, checked the current implementations, and cross-referenced with how frequently each code path is executed during LST building and recipe execution. Here's a prioritized breakdown.

Tier 1: High Value (hot paths, meaningful speedup)

PR	Method	Claimed Speedup	Why it matters
#15	Cursor.firstEnclosing()	94%	174 call sites, called thousands of times per recipe run in formatting visitors and trait matchers. Direct parent-chain walk + isInstance() vs allocating a CursorIterator + isAssignableFrom(getClass()). Recipe execution critical path.
#43	StringUtils.isBlank()	139%	178 occurrences across 93 files — parsers, validators, formatters. Fast-path for ASCII chars avoids Character.isWhitespace() overhead. Both LST building and recipe execution. Merged in #7366
#54	StringUtils.formatUriForPropertiesFile()	265%	Pre-compiling a regex pattern — well-known, reliable optimization. Currently calls String.replaceAll() which recompiles the regex on every invocation. Recipe execution (properties manipulation). Merged in #7366
#53	StringUtils.indexOfNextNonWhitespace()	57%	Used in whitespace/indentation processing during parsing and formatting. toCharArray() avoids bounds-checking on each charAt() call. Both LST building and recipe execution. Merged in #7366
#49	StringUtils.indent()	161%	Used in formatting visitors for indentation extraction. Avoids unnecessary StringBuilder allocation when a substring() suffices. Both LST building and recipe execution. Merged in #7366

Tier 2: Moderate Value (less frequent paths or smaller gains)

PR	Method	Claimed Speedup	Why it matters
#27	LocalRemoteArtifactCache.hashUri()	697%	Huge per-call speedup (ThreadLocal MessageDigest reuse + hex lookup table), but only called during remote artifact caching — not on the hot visitor/parser path. Worth it if downloading many dependencies.
#16	Cursor.toString()	75%	Replaces stream-based path collection with manual StringBuilder. However, toString() is primarily used in debugging/logging, not production hot paths.
#10	GitRemote.hashCode()	114%	Avoids Objects.hash() varargs array allocation. Limited call sites; GitRemote objects aren't created/hashed in tight loops.
#38	PathUtils.getEitherOrPatterns()	209%	Note: the regex pattern is already pre-compiled in the current code. The remaining optimization (StringBuilder pre-sizing) is marginal.

Tier 3: Low Value (marginal gains, already efficient, or questionable)

PR	Method	Claimed Speedup	Notes
#33	ExceptionUtils.containsCircularReferences()	127%	Already uses IdentityHashMap efficiently; only called during exception handling.
#45	StringUtils.capitalize()	15%	toCharArray() for a single-char mutation on short strings — tiny improvement.
#46	StringUtils.uncapitalize()	7%	Same pattern, even smaller gain.
#51	StringUtils.aspectjNameToPattern()	15%	Small gain, and the PR changes switch fall-through logic — needs careful review for correctness.
#50	StringUtils.commonMargin()	12%	Micro-optimization with marginal benefit.
#28	MetricsHelper.successTags()	116%	Extracts a string literal to a constant — the JVM already interns string literals, so the claimed speedup seems implausible.
#19	RewriteRpcProcess.getLivenessCheck()	38%	0 direct call sites in the main codebase; only relevant for RPC-based language parsers.

Recommendation

PRs #15, #43, #54, #53, #49 target the hottest code paths (visitor traversal, string utilities used everywhere in parsing and recipe execution) with meaningful speedups and low risk. Together they cover both phases (LST building + recipe execution) and methods that every OpenRewrite run exercises regardless of what recipes are being applied. Merged in #7366 Merged in #7366 Merged in #7366 Merged in #7366

PR #27 (hashUri) is a big win per-call but only matters for remote artifact downloads, so it's a nice-to-have rather than a priority.

[timtebeek]

Thanks again @misrasaurabh1 ! I've posted above a quick analysis of estimated impact; would you want to open one or more PRs for the Tier 1 cases, and perhaps the first Tier 2 case as well?

• ⚡️ Speed up method LocalRemoteArtifactCache.hashUri by 697% HeshamHM28/rewrite#27

Figured you'd prefer opening those PRs yourself for the credit as to the code change; if that's not a concern I can push up the changes as well. Just let me know!

[misrasaurabh1]

Thanks for looking into this! I can help here, but I would prefer if you push the changes yourself. Unfortunately I have too many balls in the air at the moment, so don't want to cause any delays here. If you have any questions, @HeshamHM28 can help - he worked on this originally.

[timtebeek]

Sure thing! Thanks a lot for the proposed changes both! We'll pick and choose from the above then. Keep up the great work!

[timtebeek]

Thanks a lot for the help here all. With the PRs merged and analysis above we've covered what we wanted for now. All the best!