[jayhemnani9910]

A regex on its own, as a pattern or in an expression, never matches. It is lowered as 0 ~ /re/ (the number 0) instead of $0 ~ /re/. gawk 5.3.2 vs main:

$ printf 'a b\nc\nab\n' | awk '/b/'
gawk: a b
      ab
main: (nothing)

$ printf 'a b\nc\nab\n' | awk '!/a/'
gawk: c
main: a b
      c
      ab

awk '/re/' is the most common awk program, and if (/re/), n += /re/ and /a/ || /b/ fail the same way. $0 ~ /re/ already works.

The lowering of Atom::Regex now loads field 0 into the register before the Matches, the way $0 itself is loaded.

New tests bare_regex_pattern_matches_record and bare_regex_in_expression_matches_record fail without the change. cargo test --workspace: 210 passed, clippy and fmt clean. Eleven pattern forms (/x/, !/a/, /a/ || /x/, /^$/, /[[:digit:]]+/, if (/[0-9]/), x = /a/, n += /a/, ...) print the same as gawk.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing. Review the Copilot code review billing settings.

[github-actions]

GNU awk testsuite comparison:

Test results comparison:
  Current:   TOTAL: 666 / PASSED: 100 / FAILED: 468 / SKIPPED: 98
  Reference: TOTAL: 666 / PASSED: 97 / FAILED: 471 / SKIPPED: 98

Changes from main branch:
  TOTAL: +0
  PASSED: +3
  FAILED: -3

New test failures (3):
  - litoct
  - noloop1
  - noloop2

Test improvements (6):
  + datanonl
  + dfacheck1
  + dfacheck2
  + escapebrace
  + gensub3
  + reint2

[jayhemnani9910]

The three new failures come from the bare regex now being matched against $0, not from a regression:

• noloop1 and noloop2 run /''/ { sub(...) }. Before this change the pattern never matched, so sub never ran and the empty expected output passed. Now the pattern matches and hits the todo!() for sub.
• litoct runs /a\52b/ on ab with --traditional. gawk treats octal escapes in a regex constant literally in traditional mode, so it prints "no match"; with --posix or no flag it prints "match". Here --traditional maps to the same mode as --posix, so it now matches. Before, it passed only because the regex was never tested against the record.

I can take either of these in a follow-up if you want.

[sylvestre]

a bunch of jobs are failing

[Alonely0]

The three new failures come from the bare regex now being matched against $0, not from a regression:

* `noloop1` and `noloop2` run `/''/ { sub(...) }`. Before this change the pattern never matched, so `sub` never ran and the empty expected output passed. Now the pattern matches and hits the `todo!()` for `sub`.

* `litoct` runs `/a\52b/` on `ab` with `--traditional`. gawk treats octal escapes in a regex constant literally in traditional mode, so it prints "no match"; with `--posix` or no flag it prints "match". Here `--traditional` maps to the same mode as `--posix`, so it now matches. Before, it passed only because the regex was never tested against the record.

I can take either of these in a follow-up if you want.

That would be nice, thanks! Let me take a look at the two PRs tomorrow, I've had a long day.

[Alonely0]

Nice! This must have slipped through one of the thousand refactorings of the IR emitter. We could do this a single instruction in the near future, given how common of a pattern this is.

[Alonely0]

Merged, thanks for your contribution!