feat: add multi-engine regex flavour support

This commit is contained in:
2026-07-27 11:43:51 +02:00
parent 7079cde15f
commit 7f3a91ad37
340 changed files with 643286 additions and 483 deletions

View File

@@ -2,6 +2,69 @@
## Unreleased
## 0.4.0 — 2026-07-27
- Expand the workbench from four to twelve selectable execution profiles:
JavaScript/ECMAScript, standalone PCRE2, PHP, Perl, Python, Ruby, Java, C++,
Go, Rust, .NET and Scala/JVM compatibility.
- Add actual PHP 8.5.8 `preg_*` execution through the self-hosted
`@php-wasm/web-8-5` 3.1.46 runtime. Preserve native UTF-8 byte ranges and
identify PHP's embedded PCRE2 as 10.44, independently of the standalone
PCRE2 10.47 pack.
- Add actual Perl 5.28.1 execution through unmodified WebPerl 0.09-beta assets
as an explicitly legacy beta profile. Pass user values through a fixed
data-only JSON request, stream pre-tokenized bounded replacement output into
a separately verified MEMFS binary file, disable runtime regex evaluation,
reject code assertions and retain the upstream loader's dormant
interoperability `eval` site as a documented CSP caveat.
- Add actual CRuby 4.0.0 `Regexp` execution through a minimal ruby.wasm
2.9.3-2.9.4 pack. Native `String#gsub`/`String#sub` block iteration drives a
fixed byte-bounded CRuby-compatible replacement expander. Use separate
worker-local WASI request/output files and metadata-only inert JSON instead
of output escaping or ruby.wasm's JavaScript-eval object conversion; reject
lossy lone-surrogate requests and pass the strict-CSP browser gate.
- Add Emscripten 6.0.4 libc++ `std::wregex` execution with the modified
ECMAScript default grammar, alternate standard grammars and a build
configured with dynamic execution disabled. Stream a fixed bounded
replacement formatter through the byte cap and verify its grammar against
libc++ at load.
- Add Go 1.26.5 standard-library `regexp` and Rust `regex` crate 1.13.1
execution, preserving their native UTF-8 byte ranges and non-backtracking
semantics. Use bounded streaming replacement expanders that are
differentially checked against `Regexp.ExpandString` and
`Captures::expand`, with predictable base64 output transport. Measure exact
encoded request bytes and reserve native JSON limits for worst-case escaping
so control-heavy inputs retain the documented decoded limits.
- Add .NET 10.0.10 `System.Text.RegularExpressions` in browser WebAssembly,
with invariant-culture execution, native UTF-16 ranges and a bounded native
regex timeout inside the killable worker. Stream bounded replacement tokens
without materializing `Match.Result`, verify token semantics against native
.NET at load, and transport output as bounded base64-encoded UTF-8.
- Add an explicitly labelled Scala/JVM compatibility profile backed by the
verified TeaVM 0.15.0 `java.util.regex` engine. No Scala runtime, Scala
wrapper API or OpenJDK runtime is claimed.
- Add partial application-owned lexical syntax and replacement providers,
flavour-specific quick-reference content and capability labels for the eight
new profiles while keeping each real runtime authoritative.
- Add deterministic, closed-file-set runtime packs, checksums, licences,
notices, exact toolchain/source identities and release verification for the
new self-hosted engines.
- Close the existing Pyodide pack's full base-runtime provenance: exact
primary archive or Git tree identities, vendored JavaScript-port hashes,
CPython/Emscripten patch and recipe hashes, a preferred-form HACL* route,
independently anchored notices, and executable identity checks for the
linked compression/database/parser components.
- Transport bounded Python replacement output as canonical base64 with strict
declared-length, request-cap and UTF-8 validation instead of
content-dependent JSON escaping.
- Close the PHP pack's complete native/source/legal inventory with exact
package, binary, build-recipe, archive and Git identities; independently
pin 38 legal assets; audit linked component versions; and explicitly exclude
the GMP, libsodium, tidy, ICU/intl, FreeType, ImageMagick and adjacent AOM
false-positive or unshipped detections.
- Bound retained heavyweight workers with least-recently-used eviction so
selecting many profiles does not leave every runtime resident.
## 0.3.0 — 2026-07-27
- Add a self-hosted Pyodide 314.0.3 worker containing CPython 3.14.2 `re`, with