l10n: replace the hash table by a long match per file to faciliate th…

[sylvestre]

…e lookup

[sylvestre]

should improve the perf?!

[codspeed-hq]

CodSpeed Performance Report

Merging #8734 will improve performances by 24.17%

_{Comparing sylvestre:l10n-perf (3b9d0a8) with main (12e875c)}

Summary

⚡ 17 improvements
✅ 28 untouched
⏩ 1 skipped¹

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	tsort_complex_dag[10000]	155.6 µs	125.3 µs	+24.17%
⚡	tsort_complex_dag[1000]	155.1 µs	126.1 µs	+23%
⚡	tsort_complex_dag[50000]	155.2 µs	126.7 µs	+22.54%
⚡	tsort_complex_dag[5000]	155.6 µs	126.5 µs	+23.02%
⚡	tsort_input_parsing_heavy[10000]	148.8 µs	120 µs	+24.03%
⚡	tsort_input_parsing_heavy[50000]	154.5 µs	125 µs	+23.64%
⚡	tsort_linear_chain[1000000]	155.6 µs	126.2 µs	+23.31%
⚡	tsort_linear_chain[100000]	155.3 µs	126.3 µs	+22.97%
⚡	tsort_linear_chain[10000]	156.1 µs	126.1 µs	+23.78%
⚡	tsort_linear_chain[1000]	155 µs	126.4 µs	+22.61%
⚡	tsort_tree_dag[(4, 3)]	155.6 µs	126 µs	+23.44%
⚡	tsort_tree_dag[(5, 3)]	155.3 µs	125.6 µs	+23.58%
⚡	tsort_tree_dag[(6, 2)]	155.2 µs	126 µs	+23.16%
⚡	tsort_tree_dag[(7, 2)]	155.4 µs	126.2 µs	+23.17%
⚡	tsort_wide_dag[100000]	154.9 µs	125.7 µs	+23.16%
⚡	tsort_wide_dag[10000]	154.9 µs	125.1 µs	+23.87%
⚡	tsort_wide_dag[50000]	154.8 µs	125.5 µs	+23.36%

Footnotes

1. 1 benchmark was skipped, so the baseline result was used instead. If it was deleted from the codebase, click here and archive it to remove it from the performance reports. ↩

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/tail/overlay-headers. tests/tail/overlay-headers is passing on 'main'. Maybe you have to rebase?

[sylvestre]

sweet, it works!
tsort is so fast that most of the time is spent on l10n work!

[cakebaker]

Nice performance improvement :)