perf: optimize metadata generator

[DjDeveloperr]

Summary

• speed up the Objective-C metadata generator while preserving generated output
• add an opt-in Rust/Oxc analyzer that scans emitted JavaScript or TypeScript bundles and generates a symbol-based NativeScript metadata whitelist
• retain the existing whitelist.mdg and blacklist.mdg filtering behavior
• harden build integration so the analyzer is packaged for the host architecture without causing redundant metadata-generator rebuilds

How automatic filtering works

The new ns-metadata-symbols tool parses bundle files with Oxc, resolves unbound native symbols, and analyzes multiple files/chunks in parallel with Rayon. It emits deterministic NativeScript .mdg rules that the existing metadata-generator build step consumes.

This feature is opt-in:

NS_METADATA_AUTO_FILTER=1
NS_METADATA_BUNDLE_PATHS="$CONFIGURATION_BUILD_DIR/app/bundle.js"

The packaged analyzer is used automatically. NS_METADATA_SYMBOL_ANALYZER can override its path for development.

Safety and compatibility behavior:

• existing manual whitelist entries are merged into the generated whitelist
• blacklist.mdg is still applied after the whitelist
• Foundation and Runtime metadata are retained by default
• parse/semantic failures, dynamic global access, and dynamic code execution fail open by retaining all metadata
• symbol matching intentionally prefers false positives over removing required metadata

Performance

On the full macOS SDK workload, median warm-cache results improved while all 207 generated artifacts remained byte-for-byte identical:

Measurement	Baseline	Optimized	Change
Wall time	6.98 s	4.94 s	-29.2%
User CPU	5.62 s	4.35 s	-22.6%
System CPU	0.80 s	0.55 s	-31.2%
Maximum RSS	1.21 GB	1.18 GB	-2.6%

For a representative AppKit bundle using NSView, automatic filtering reduced binary metadata by 92.4%, signature bindings by 84.3%, and TypeScript output by 93.1%. These figures are workload-specific rather than size guarantees.

The analyzer averaged 6.5 ms across the 184-file TestRunner corpus with 18 workers, versus 9.9 ms with one worker, while producing identical output.

Validation

• Rust analyzer unit and CLI tests: 8 passed
• C++ metadata-filter regression test: passed
• source, minified, and code-split bundle analysis produced byte-identical whitelists
• malformed/dynamic TestRunner inputs verified conservative fail-open behavior
• macOS CI: 714 tests, 0 failures
• iOS CI: 714 specs, 0 failures
• full CI: passed

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 240b3475-a81f-4cd0-832e-487e42a814c5

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}