perf(coverage): scan function declarations in one awk pass by Chemaclass · Pull Request #1085 · TypedDevs/bashunit

Chemaclass · 2026-08-13T04:56:50Z

🤔 Background

Related #1084

extract_functions walked each source file in Bash, counting braces with two
pattern substitutions per line — 17.5 ms per file against a 3.5 ms fork floor,
and the report phase pays it once per file per renderer.

💡 Changes

The same state machine runs as one awk pass; rules unchanged, quirks included
extract_functions 2238 ms → 399 ms over 128 files, report_lcov 5052 ms → 2638 ms, and a --coverage run over src 9.23 s → 6.81 s
Output diffed against the Bash implementation over all 463 shell files in the repo, and the LCOV report is byte-identical on a fixed corpus
Characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF, and shapes that are not declarations

extract_functions walked each file in Bash and counted braces with two pattern substitutions per line: 17.5ms per file against a 3.5ms fork floor, paid once per file per renderer. The same state machine in awk is 2238ms to 399ms over 128 files, and report_lcov 5052ms to 2638ms. The rules are unchanged, quirks included, and the output was diffed against the Bash implementation over all 463 shell files in the repo. The characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF and the shapes that are not declarations. Closes#1084

Chemaclass · 2026-08-13T04:56:50Z

🤔 Background

Related #1084

extract_functions walked each source file in Bash, counting braces with two
pattern substitutions per line — 17.5 ms per file against a 3.5 ms fork floor,
and the report phase pays it once per file per renderer.

💡 Changes

The same state machine runs as one awk pass; rules unchanged, quirks included
extract_functions 2238 ms → 399 ms over 128 files, report_lcov 5052 ms → 2638 ms, and a --coverage run over src 9.23 s → 6.81 s
Output diffed against the Bash implementation over all 463 shell files in the repo, and the LCOV report is byte-identical on a fixed corpus
Characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF, and shapes that are not declarations

extract_functions walked each file in Bash and counted braces with two pattern substitutions per line: 17.5ms per file against a 3.5ms fork floor, paid once per file per renderer. The same state machine in awk is 2238ms to 399ms over 128 files, and report_lcov 5052ms to 2638ms. The rules are unchanged, quirks included, and the output was diffed against the Bash implementation over all 463 shell files in the repo. The characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF and the shapes that are not declarations. Closes#1084

Chemaclass · 2026-08-13T04:56:50Z

🤔 Background

Related #1084

extract_functions walked each source file in Bash, counting braces with two
pattern substitutions per line — 17.5 ms per file against a 3.5 ms fork floor,
and the report phase pays it once per file per renderer.

💡 Changes

The same state machine runs as one awk pass; rules unchanged, quirks included
extract_functions 2238 ms → 399 ms over 128 files, report_lcov 5052 ms → 2638 ms, and a --coverage run over src 9.23 s → 6.81 s
Output diffed against the Bash implementation over all 463 shell files in the repo, and the LCOV report is byte-identical on a fixed corpus
Characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF, and shapes that are not declarations

extract_functions walked each file in Bash and counted braces with two pattern substitutions per line: 17.5ms per file against a 3.5ms fork floor, paid once per file per renderer. The same state machine in awk is 2238ms to 399ms over 128 files, and report_lcov 5052ms to 2638ms. The rules are unchanged, quirks included, and the output was diffed against the Bash implementation over all 463 shell files in the repo. The characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF and the shapes that are not declarations. Closes#1084

Chemaclass · 2026-08-13T04:56:50Z

🤔 Background

Related #1084

extract_functions walked each source file in Bash, counting braces with two
pattern substitutions per line — 17.5 ms per file against a 3.5 ms fork floor,
and the report phase pays it once per file per renderer.

💡 Changes

The same state machine runs as one awk pass; rules unchanged, quirks included
extract_functions 2238 ms → 399 ms over 128 files, report_lcov 5052 ms → 2638 ms, and a --coverage run over src 9.23 s → 6.81 s
Output diffed against the Bash implementation over all 463 shell files in the repo, and the LCOV report is byte-identical on a fixed corpus
Characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF, and shapes that are not declarations

extract_functions walked each file in Bash and counted braces with two pattern substitutions per line: 17.5ms per file against a 3.5ms fork floor, paid once per file per renderer. The same state machine in awk is 2238ms to 399ms over 128 files, and report_lcov 5052ms to 2638ms. The rules are unchanged, quirks included, and the output was diffed against the Bash implementation over all 463 shell files in the repo. The characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF and the shapes that are not declarations. Closes#1084

Chemaclass · 2026-08-13T04:56:50Z

🤔 Background

Related #1084

extract_functions walked each source file in Bash, counting braces with two
pattern substitutions per line — 17.5 ms per file against a 3.5 ms fork floor,
and the report phase pays it once per file per renderer.

💡 Changes

The same state machine runs as one awk pass; rules unchanged, quirks included
extract_functions 2238 ms → 399 ms over 128 files, report_lcov 5052 ms → 2638 ms, and a --coverage run over src 9.23 s → 6.81 s
Output diffed against the Bash implementation over all 463 shell files in the repo, and the LCOV report is byte-identical on a fixed corpus
Characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF, and shapes that are not declarations

extract_functions walked each file in Bash and counted braces with two pattern substitutions per line: 17.5ms per file against a 3.5ms fork floor, paid once per file per renderer. The same state machine in awk is 2238ms to 399ms over 128 files, and report_lcov 5052ms to 2638ms. The rules are unchanged, quirks included, and the output was diffed against the Bash implementation over all 463 shell files in the repo. The characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF and the shapes that are not declarations. Closes#1084

Chemaclass · 2026-08-13T04:56:50Z

🤔 Background

Related #1084

extract_functions walked each source file in Bash, counting braces with two
pattern substitutions per line — 17.5 ms per file against a 3.5 ms fork floor,
and the report phase pays it once per file per renderer.

💡 Changes

The same state machine runs as one awk pass; rules unchanged, quirks included
extract_functions 2238 ms → 399 ms over 128 files, report_lcov 5052 ms → 2638 ms, and a --coverage run over src 9.23 s → 6.81 s
Output diffed against the Bash implementation over all 463 shell files in the repo, and the LCOV report is byte-identical on a fixed corpus
Characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF, and shapes that are not declarations

extract_functions walked each file in Bash and counted braces with two pattern substitutions per line: 17.5ms per file against a 3.5ms fork floor, paid once per file per renderer. The same state machine in awk is 2238ms to 399ms over 128 files, and report_lcov 5052ms to 2638ms. The rules are unchanged, quirks included, and the output was diffed against the Bash implementation over all 463 shell files in the repo. The characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF and the shapes that are not declarations. Closes#1084

Chemaclass · 2026-08-13T04:56:50Z

🤔 Background

Related #1084

extract_functions walked each source file in Bash, counting braces with two
pattern substitutions per line — 17.5 ms per file against a 3.5 ms fork floor,
and the report phase pays it once per file per renderer.

💡 Changes

The same state machine runs as one awk pass; rules unchanged, quirks included
extract_functions 2238 ms → 399 ms over 128 files, report_lcov 5052 ms → 2638 ms, and a --coverage run over src 9.23 s → 6.81 s
Output diffed against the Bash implementation over all 463 shell files in the repo, and the LCOV report is byte-identical on a fixed corpus
Characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF, and shapes that are not declarations

extract_functions walked each file in Bash and counted braces with two pattern substitutions per line: 17.5ms per file against a 3.5ms fork floor, paid once per file per renderer. The same state machine in awk is 2238ms to 399ms over 128 files, and report_lcov 5052ms to 2638ms. The rules are unchanged, quirks included, and the output was diffed against the Bash implementation over all 463 shell files in the repo. The characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF and the shapes that are not declarations. Closes#1084

Chemaclass · 2026-08-13T04:56:50Z

🤔 Background

Related #1084

extract_functions walked each source file in Bash, counting braces with two
pattern substitutions per line — 17.5 ms per file against a 3.5 ms fork floor,
and the report phase pays it once per file per renderer.

💡 Changes

The same state machine runs as one awk pass; rules unchanged, quirks included
extract_functions 2238 ms → 399 ms over 128 files, report_lcov 5052 ms → 2638 ms, and a --coverage run over src 9.23 s → 6.81 s
Output diffed against the Bash implementation over all 463 shell files in the repo, and the LCOV report is byte-identical on a fixed corpus
Characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF, and shapes that are not declarations

extract_functions walked each file in Bash and counted braces with two pattern substitutions per line: 17.5ms per file against a 3.5ms fork floor, paid once per file per renderer. The same state machine in awk is 2238ms to 399ms over 128 files, and report_lcov 5052ms to 2638ms. The rules are unchanged, quirks included, and the output was diffed against the Bash implementation over all 463 shell files in the repo. The characterization tests pin the spans that carry the risk: nested braces, single-line bodies, every declaration spelling, an unclosed body at EOF and the shapes that are not declarations. Closes#1084

Chemaclass added the enhancement New feature or request label Aug 13, 2026

Chemaclass self-assigned this Aug 13, 2026

Chemaclass merged commit 3e30c70 into mainAug 13, 2026
37 checks passed