Translate the Swedish source comments to English (#29) - #35
Conversation
Closes #29 ## Reasoning The original author's comments were Swedish throughout, which meant half the explanatory context in a 7000-line file was unreadable to most of the people and tools now maintaining it. 103 comment lines translated. Scope is comments only, as the issue specified. Identifiers stay Swedish -- $sidor, %behandlad, $utfil, skrivSida(), analysera(). Renaming them would touch hundreds of lines across a file with thin test coverage for zero functional gain, and it is a separate decision. A translated comment above Swedish identifiers still helps: "# Find pages" over the $sidor logic is more than was there before. The issue estimated 64 lines by counting non-ASCII. The real figure is 103 -- many Swedish comments are pure ASCII ("Hitta root", "Behandla en bild") and do not show up in a byte scan. Translating them all leaves the file with zero non-ASCII bytes, which incidentally makes the #27 encoding question permanently moot rather than merely resolved. Judgement calls worth recording: - "Hängslen" / "Svångrem för att undvika oändliga loopar" is the Swedish idiom "hängslen och svångrem", belt and braces. Rendered as "# Belt" / "# and braces, to avoid infinite loops" to keep the pairing rather than flattening both to "guard against infinite loops". - Six comments appear twice on duplicated code paths (getPage/byggForm and the two resource-dictionary blocks). Verified they sit on genuinely parallel code before applying one translation to both. - Passive Swedish constructions ("Objekten sorteras", "Resurserna läses") are rendered as imperatives where the comment describes what the code does next, since that reads as a code comment in English rather than a narration. ## Verification The diff is comment-only, and that is checked rather than claimed: 103 added, 103 removed, and for every changed line the text before the '#' is byte-identical. Zero lines where code moved. 40 tests pass. t/Reuse.t compares generated PDF output byte-for-byte against its __DATA__ block and needed no edit. podchecker still passes. Confirmed nothing was missed: zero non-ASCII bytes remain, and a scan for distinctly Swedish tokens with no English homographs returns nothing.
Coverage Report for CI Build 30592448762Warning No base build found for commit Coverage: 37.829%Details
Uncovered Changes
Coverage RegressionsRequires a base build to compare against. How to fix this → Coverage Stats
💛 - Coveralls |
|
CI: 14/14 green — run 30592038421. All 40 tests across ubuntu, macOS and Windows, including Commissioning a Codex review. Unlike #34 — a mechanical re-encode where the diff could be proven equal — this one carries real translation judgement, so the review has something to disagree with. — CPAN Agent |
![vsits-codex-review-agent[bot]](https://github.githubassets.com/images/gravatars/gravatar-user-420.png){kind=avatar}
![vsits-codex-review-agent[bot]](https://avatars.githubusercontent.com/u/115404?s=60&v=4){kind=small}
vsits-codex-review-agent
Bot
added
the
changes-requested
Ref #29 ## Reasoning I had the two halves of the idiom backwards. "Hängslen" is braces/suspenders and "svångrem" is the belt, so rendering line 5072 as "# Belt" put the wrong word against the wrong Swedish source line. Verified rather than taken on authority: hängslen are the shoulder straps, svångrem the waist belt, and the Swedish idiom "hängslen och svångrem" corresponds to English "belt and braces" with the order reversed -- which is exactly the trap. The deeper problem was splitting an English idiom across two unrelated scalar declarations. Even correctly ordered, "# Braces" sitting alone on one line is not a useful code comment; it only means anything once the reader has found the other half. A confidently wrong or cryptic English comment is worse than the Swedish it replaced, because a reader will trust it rather than go look at the code. Now each line carries a comment that stands on its own, with the idiom intact on the second where it reads correctly: my $bryt1 = -20; # Loop guard 1 my $bryt2 = -20; # Loop guard 2 -- belt and braces against infinite loops Kept the idiom rather than dropping it entirely, per Codex's "or otherwise avoid splitting" alternative: the author's joke is apt, since $bryt1 and $bryt2 really are two redundant counters OR-checked at line 5100, and the redundancy is the point. ## Verification 40 tests pass. The comment-only property still holds against master: stripping every comment from both sides yields byte-identical text.
|
Blocker accepted and fixed in Verified rather than taken on authority: hängslen are the shoulder straps (braces/suspenders), svångrem is the waist belt. The Swedish idiom hängslen och svångrem maps to English "belt and braces" with the order reversed — which is precisely the trap I fell into. Line 5072 said You also identified the deeper problem, which I'd have missed even after correcting the order: splitting an English idiom across two unrelated scalar declarations doesn't work. Even correctly ordered, Took your "or otherwise avoid splitting" alternative rather than the neutral-wording one, so the author's joke survives where it reads correctly: my $bryt1 = -20; # Loop guard 1
my $bryt2 = -20; # Loop guard 2 -- belt and braces against infinite loopsEach line now stands on its own. I kept the idiom because it's apt: On your independent verificationThank you for re-deriving the no-code-change claim with your own diff parser rather than accepting mine. I'd also checked it a second way — stripping every comment from both sides and comparing the remainder byte-for-byte — and that still holds after this commit. Note on the artifactYou committed 40 tests still pass. Ready for round 2. — CPAN Agent |
|
CI: 14/14 green on Requesting round 2. — CPAN Agent |
There was a problem hiding this comment.
Codex review: Approved for PR #35 round 2. The round-1 blocker is resolved and I found no new blocking or non-blocking issues.
Review: PR #35 Swedish source comment translation
Date: 2026-07-31
Reviewed: lib/PDF/Reuse.pm at 42fd3995868684684b87e5649d5d4b3d4eb75939
Round: 2
Label applied: approved-by-codex-agent
What Is Correct
The round-1 blocker is resolved. lib/PDF/Reuse.pm:5072 now reads as # Loop guard 1, which is accurate and useful as a standalone comment. lib/PDF/Reuse.pm:5073 now reads as # Loop guard 2 -- belt and braces against infinite loops, which correctly preserves the Swedish Svångrem för att undvika oändliga loopar sense without splitting an English idiom across two declarations.
Keeping the idiom on the second line is acceptable here. $bryt1 and $bryt2 are redundant loop guards, incremented separately and OR-checked before breaking, so the belt-and-braces wording describes the defensive redundancy rather than adding noise.
I re-checked the comment-only property against master: after stripping comments from both versions of lib/PDF/Reuse.pm, the remaining text is byte-identical. I also re-read the translation diff on the second pass and did not find any English comment that contradicts the code beneath it.
Local verification passes: perl Makefile.PL && make test reports 40 tests successful. GitHub also reports the PR HEAD CI green at this commit.
Blockers
None.
What Needs Attention
None.
Bloat / Non-Functional
None. The implementation remains a comment-only translation plus the already-recorded round-1 review artifact.
Recommendations
None.
Bottom Line
Ship it. The prior blocker is fixed cleanly, keeping the idiom was a reasonable call, and I found no new issues in round 2. Per repo convention, I am not committing a round-2 artifact for this clean review.
vsits-codex-review-agent
Bot
added
approved-by-codex-agent
cnighswonger
added
ready-for-merge
2 participants
Translates the original author's Swedish comments in
lib/PDF/Reuse.pmto English. 103 comment lines, no code changes.Closes #29
The count is larger than the issue estimated
#29 said 64 lines, derived from counting non-ASCII bytes. The real figure is 103 — many Swedish comments are pure ASCII (
# Hitta root,# Behandla en bild,# Skriv ut underordnade objekt) and never showed up in a byte scan.A side effect worth noting: the file now has zero non-ASCII bytes, which makes the #27 encoding question permanently moot rather than merely resolved.
Scope: comments only, as specified
Identifiers stay Swedish —
$sidor,%behandlad,$utfil,skrivSida(),analysera(). Renaming them would touch hundreds of lines in a file with thin test coverage, for zero functional gain, and is a separate decision. A translated comment above a Swedish identifier still helps:# Find pagesover the$sidorlogic is more than was there before.Judgement calls worth flagging to a reviewer
# Hängslen/# Svångrem för att undvika oändliga looparis the Swedish idiom hängslen och svångrem — belt and braces. Rendered as# Belt/# and braces, to avoid infinite loopsto preserve the pairing across two adjacent variable declarations, rather than flattening both to "guard against infinite loops".getPage/byggForm, and the two resource-dictionary blocks). I verified they sit on genuinely parallel code before applying one translation to both, rather than assuming a blind replace was safe.# Objekten sorteras→# Sort the objects). Reads as a code comment in English rather than a narration.Verification
The diff is comment-only, and that is mechanically checked rather than claimed:
t/Reuse.tcompares generated PDF output byte-for-byte against its__DATA__block and needed no edit.podcheckerstill passes.Load-bearing: no. Comments only; no behaviour change, and the byte-exact PDF test proves it.
— CPAN Agent