Skip to content

Reduce redundant compiler work and release 0.12.2 - #63

Merged
gregjotau merged 2 commits into
mainfrom
perf/compiler-small-wins
Sep 19, 2026
Merged

gregjotau merged 2 commits into
mainfrom
perf/compiler-small-wins

Conversation

@gregjotau

@gregjotau gregjotau commented Sep 19, 2026 •

Copy link
Copy Markdown
Contributor

Repeated template compilation rescans the same controller routes and copies every completed renderer source through indentation and regex passes. Cache path matches for one compilation and generate source at its final indentation. HTTP methods, enum coverage and source diagnostics are still checked per use; generated files and runtime behavior stay unchanged. Release as 0.12.2.

On ReAI/M5 Max, eight warmed runs per version give median complete builds of 6.684 → 6.151 s for static edits that change Java offsets, and 5.597 → 5.533 s for same-length resource-only edits. Timings overlap and vary; this is a modest reduction of redundant work, not a promised percentage. The audit includes raw results, warmups, and rejected experiments.

Validation:

  • Full clean build --no-build-cache and the CI isolated-project checks pass.
  • New regressions cover repeated URLs with differing HTTP methods, source locations and enum bindings; existing golden renders pass.
  • All 67 generated ReAI web-app files match byte for byte across both edit sequences, with the build cache disabled.
  • Packaged consumer builds pass with the composite candidate (application tests excluded). All 304 generated files match 0.12.1 exactly: ReAI 133, Utin 64, Eteo 46, Ecomtools 61.

@gregjotau
gregjotau merged commit 45c6e0d into main Sep 19, 2026
1 check passed
@gregjotau
gregjotau deleted the perf/compiler-small-wins branch September 19, 2026 10:44
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant