Re Development Guide

Re is linear-time regular expressions in ECMA-262's syntax. WEP: Re holds the design and why.

Sources

The rules come from the specification, vendor/ecma262/spec.html: the chapter RegExp (Regular Expression) Objects, read as the u flag reads it. Annex B does not apply. A rule in the code cites the production or the section it follows.

vendor/test262 is the conformance corpus, read as data. No engine's source is read, whatever its license. Node's RegExp may be run as a black-box oracle.

\p{…} takes what core:icu knows, at its Unicode version. Where that differs from ECMA-262's tables or from an engine, core:icu is right.

Layout

• src/ast.wado — the tree a pattern parses to.
• src/parse.wado — the parser, rex's syntax included, PatternError, and validate_ecma262_syntax.
• src/charset.wado — code point sets, and what each named set holds.
• src/utf8.wado — a code point range as the UTF-8 byte sequences encoding it.
• src/program.wado — the tree compiled to Pike VM instructions over bytes.
• src/pike.wado — the Pike VM.
• src/regex.wado — Regex, its searches, replace and split.
• src/template.wado — the tags re and rex, and PatternPart.

Tests

A test writes a pattern as a raw string, #"\d+"#, so that it reads as the pattern it is. A cooked one needs a reason, such as a newline ending a rex comment. A program uses re and rex, whose templates are raw already.

Running tests

wado test package-re

tests/generated/ holds what the oracles answered, committed so that CI needs neither:

• mise run oracle-re asks Node's RegExp about each case in tests/oracle/cases.json and tests/generated/test262.json.
• mise run test262-re reads test262's parse-phase negative tests, and runs test262's RegExp tests in Node to collect the patterns and inputs they use into tests/generated/test262.json. It then runs oracle-re.