perf(state): skip base64 forks for empty subshell-context fields by Chemaclass · Pull Request #768 · TypedDevs/bashunit

Chemaclass · 2026-07-11T12:52:01Z

🤔 Background

Related #762

Every test base64-encoded its title, hook message and output when exporting subshell context, forking base64 three times even though a passing test leaves all three empty.

💡 Changes

Extract encode_field, which returns an empty field for empty input with no fork; base64 of "" is "", so the wire format is byte-identical. Also short-circuit decode_base64 on empty input.
Fork census on a 100-test file: base64 101 -> 1 (three encodes + the output decode removed per test). Platform-independent win.
Behaviour preserved: existing export_subshell_context characterization tests (empty and non-empty fields) plus new round-trip tests (empty, plain, multiline, ANSI) pass; full suite green sequential and --parallel.

Extract encode_field, which returns an empty field for empty input without forking base64 (base64 of "" is "" anyway, so the wire format is unchanged). A passing test's title, hook message and output are all empty, so this removes three forks per test. Also short-circuit decode_base64 on empty input. No behaviour change. Closes#762.

Chemaclass · 2026-07-11T12:52:01Z

🤔 Background

Related #762

Every test base64-encoded its title, hook message and output when exporting subshell context, forking base64 three times even though a passing test leaves all three empty.

💡 Changes

Extract encode_field, which returns an empty field for empty input with no fork; base64 of "" is "", so the wire format is byte-identical. Also short-circuit decode_base64 on empty input.
Fork census on a 100-test file: base64 101 -> 1 (three encodes + the output decode removed per test). Platform-independent win.
Behaviour preserved: existing export_subshell_context characterization tests (empty and non-empty fields) plus new round-trip tests (empty, plain, multiline, ANSI) pass; full suite green sequential and --parallel.

Extract encode_field, which returns an empty field for empty input without forking base64 (base64 of "" is "" anyway, so the wire format is unchanged). A passing test's title, hook message and output are all empty, so this removes three forks per test. Also short-circuit decode_base64 on empty input. No behaviour change. Closes#762.

Chemaclass · 2026-07-11T12:52:01Z

🤔 Background

Related #762

Every test base64-encoded its title, hook message and output when exporting subshell context, forking base64 three times even though a passing test leaves all three empty.

💡 Changes

Extract encode_field, which returns an empty field for empty input with no fork; base64 of "" is "", so the wire format is byte-identical. Also short-circuit decode_base64 on empty input.
Fork census on a 100-test file: base64 101 -> 1 (three encodes + the output decode removed per test). Platform-independent win.
Behaviour preserved: existing export_subshell_context characterization tests (empty and non-empty fields) plus new round-trip tests (empty, plain, multiline, ANSI) pass; full suite green sequential and --parallel.

Extract encode_field, which returns an empty field for empty input without forking base64 (base64 of "" is "" anyway, so the wire format is unchanged). A passing test's title, hook message and output are all empty, so this removes three forks per test. Also short-circuit decode_base64 on empty input. No behaviour change. Closes#762.

Chemaclass · 2026-07-11T12:52:01Z

🤔 Background

Related #762

Every test base64-encoded its title, hook message and output when exporting subshell context, forking base64 three times even though a passing test leaves all three empty.

💡 Changes

Extract encode_field, which returns an empty field for empty input with no fork; base64 of "" is "", so the wire format is byte-identical. Also short-circuit decode_base64 on empty input.
Fork census on a 100-test file: base64 101 -> 1 (three encodes + the output decode removed per test). Platform-independent win.
Behaviour preserved: existing export_subshell_context characterization tests (empty and non-empty fields) plus new round-trip tests (empty, plain, multiline, ANSI) pass; full suite green sequential and --parallel.

Extract encode_field, which returns an empty field for empty input without forking base64 (base64 of "" is "" anyway, so the wire format is unchanged). A passing test's title, hook message and output are all empty, so this removes three forks per test. Also short-circuit decode_base64 on empty input. No behaviour change. Closes#762.

Chemaclass · 2026-07-11T12:52:01Z

🤔 Background

Related #762

Every test base64-encoded its title, hook message and output when exporting subshell context, forking base64 three times even though a passing test leaves all three empty.

💡 Changes

Extract encode_field, which returns an empty field for empty input with no fork; base64 of "" is "", so the wire format is byte-identical. Also short-circuit decode_base64 on empty input.
Fork census on a 100-test file: base64 101 -> 1 (three encodes + the output decode removed per test). Platform-independent win.
Behaviour preserved: existing export_subshell_context characterization tests (empty and non-empty fields) plus new round-trip tests (empty, plain, multiline, ANSI) pass; full suite green sequential and --parallel.

Extract encode_field, which returns an empty field for empty input without forking base64 (base64 of "" is "" anyway, so the wire format is unchanged). A passing test's title, hook message and output are all empty, so this removes three forks per test. Also short-circuit decode_base64 on empty input. No behaviour change. Closes#762.

Chemaclass · 2026-07-11T12:52:01Z

🤔 Background

Related #762

Every test base64-encoded its title, hook message and output when exporting subshell context, forking base64 three times even though a passing test leaves all three empty.

💡 Changes

Extract encode_field, which returns an empty field for empty input with no fork; base64 of "" is "", so the wire format is byte-identical. Also short-circuit decode_base64 on empty input.
Fork census on a 100-test file: base64 101 -> 1 (three encodes + the output decode removed per test). Platform-independent win.
Behaviour preserved: existing export_subshell_context characterization tests (empty and non-empty fields) plus new round-trip tests (empty, plain, multiline, ANSI) pass; full suite green sequential and --parallel.

Extract encode_field, which returns an empty field for empty input without forking base64 (base64 of "" is "" anyway, so the wire format is unchanged). A passing test's title, hook message and output are all empty, so this removes three forks per test. Also short-circuit decode_base64 on empty input. No behaviour change. Closes#762.

Chemaclass · 2026-07-11T12:52:01Z

🤔 Background

Related #762

Every test base64-encoded its title, hook message and output when exporting subshell context, forking base64 three times even though a passing test leaves all three empty.

💡 Changes

Extract encode_field, which returns an empty field for empty input with no fork; base64 of "" is "", so the wire format is byte-identical. Also short-circuit decode_base64 on empty input.
Fork census on a 100-test file: base64 101 -> 1 (three encodes + the output decode removed per test). Platform-independent win.
Behaviour preserved: existing export_subshell_context characterization tests (empty and non-empty fields) plus new round-trip tests (empty, plain, multiline, ANSI) pass; full suite green sequential and --parallel.

Extract encode_field, which returns an empty field for empty input without forking base64 (base64 of "" is "" anyway, so the wire format is unchanged). A passing test's title, hook message and output are all empty, so this removes three forks per test. Also short-circuit decode_base64 on empty input. No behaviour change. Closes#762.

Chemaclass · 2026-07-11T12:52:01Z

🤔 Background

Related #762

Every test base64-encoded its title, hook message and output when exporting subshell context, forking base64 three times even though a passing test leaves all three empty.

💡 Changes

Extract encode_field, which returns an empty field for empty input with no fork; base64 of "" is "", so the wire format is byte-identical. Also short-circuit decode_base64 on empty input.
Fork census on a 100-test file: base64 101 -> 1 (three encodes + the output decode removed per test). Platform-independent win.
Behaviour preserved: existing export_subshell_context characterization tests (empty and non-empty fields) plus new round-trip tests (empty, plain, multiline, ANSI) pass; full suite green sequential and --parallel.

Extract encode_field, which returns an empty field for empty input without forking base64 (base64 of "" is "" anyway, so the wire format is unchanged). A passing test's title, hook message and output are all empty, so this removes three forks per test. Also short-circuit decode_base64 on empty input. No behaviour change. Closes#762.