fix: format_o (%o) now uses language-specific ordinal suffixes

[Koan-Bot]

What

The %o format specifier now correctly uses each language module's @Dsuf array for ordinal suffixes, instead of always falling back to English (st/nd/rd/th).

Why

format_o in Generic.pm closed over a lexical my @Dsuf, making it impossible for language modules to provide their own ordinal suffixes through inheritance. Languages like Spanish, Brazilian, Romanian, Arabic, and Occitan declared custom @Dsuf arrays but these were silently ignored — users always got English suffixes.

How

Changed format_o to dynamically resolve @Dsuf from the object's package via symbolic reference, falling back to the English default when no package-level @Dsuf exists. This follows the same pattern already used by Date::Language::format_Z.

Testing

• Extended t/time2str-lang.t from 8 to 20 tests covering %o across English, French, German, Spanish, Romanian, and Occitan
• Full test suite passes (28 test files)

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

This PR fixes format_o (%o) in Generic.pm to dynamically look up @Dsuf from the calling object's package via a symbolic reference (with a fallback to the English @Dsuf lexical in Generic.pm), mirroring the pattern already used by format_Z. The implementation is correct for all language modules that declare our @Dsuf, and the fix genuinely resolves the long-standing bug where Spanish, Romanian, Occitan, and similar languages always received English ordinal suffixes.

Confidence Score: 5/5

Safe to merge — the fix is correct and all remaining findings are minor style/test-quality suggestions with no runtime impact.

No P0 or P1 issues found. The symbolic-reference lookup is correct for all our @Dsuf language modules, the English fallback works properly, and the edge cases (empty $pkg, unknown package) are handled. All three comments are P2: test scope clarity, assertion strength, and a pre-existing uninitialized-value risk for day 31.

t/time2str-lang.t — the Spanish and Occitan assertions could be tightened to exact equality checks for better regression protection.

Important Files Changed

Filename	Overview
lib/Date/Format/Generic.pm	format_o updated to dynamically resolve @Dsuf from the object's package via symbolic reference, with correct English fallback; logic is sound for all language modules using our @Dsuf
t/time2str-lang.t	12 new %o tests added; French and German tests exercise pre-existing per-module format_o overrides rather than the new Generic.pm code path; Spanish and Occitan assertions use unlike instead of exact value checks

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["format_o called\n(Generic.pm)"] --> B["pkg = ref(obj)"]
    B --> C{pkg truthy?}
    C -- No --> F["dsuf = undef"]
    C -- Yes --> D["no strict 'refs'\ndsuf = \\@{pkg::Dsuf}"]
    D --> E{"dsuf && @$dsuf non-empty?"}
    F --> G["dsuf = \\@Dsuf\n(English fallback)"]
    E -- No --> G
    E -- Yes --> H["Use language-specific @Dsuf\nfrom pkg symbol table"]
    G --> I["sprintf('%2d%s', day, dsuf->[day])"]
    H --> I
    I --> J["Return ordinal string\ne.g. ' 1st', ' 1do', ' 1er'"]
    style G fill:#f9f,stroke:#333
    style H fill:#9f9,stroke:#333

Loading

_{Reviews (1): Last reviewed commit: "fix: format_o (%o) now uses language-spe..." | Re-trigger Greptile}

[greptile-apps]

French and German tests exercise pre-existing overrides, not the new fix

Both French.pm and German.pm define their own format_o method that overrides Generic.pm::format_o entirely — so these tests prove the pre-existing overrides still work, not the new symbolic-reference lookup. The new code path in Generic.pm is only exercised by the Spanish, Romanian, and Occitan tests below. This doesn't cause test failures, but the coverage statement in the PR description is slightly misleading: adding an assertion on a language whose format_o relies solely on Generic.pm (e.g., Dutch or Finnish) would provide more direct validation.

[greptile-apps]

Weak assertion — consider testing the exact expected value

The unlike assertion only verifies that Spanish day 1 doesn't end in the English "st" suffix; it passes even if the result is garbled or empty. Since the comment on the preceding line already states the exact expected suffix ('do'), a precise equality check would make the test self-documenting and catch regressions:

Suggested change

	{
	my $result = time2str('%o', $jan1, 'GMT', 'Spanish');
	unlike( $result, qr/st$/, 'Spanish %o day 1 is not English "st"' );
	is( $result, ' 1do', 'Spanish %o day 1 uses Spanish suffix "do"' );

The same applies to the Occitan block below (expected value ' 1er').

[greptile-apps]

$dsuf && guard is redundant, and day-31 is unguarded

After the ternary, $dsuf is either undef (when ref($_[0]) is falsy) or a reference — never a non-reference falsy value — so $dsuf && adds nothing beyond @$dsuf. More practically, several @Dsuf arrays (Romanian, Occitan, Spanish, Swedish, Finnish, Portuguese, etc.) have exactly 31 elements (indices 0–30). Day 31 uses index 31, which is out of range, so $dsuf->[31] returns undef and sprintf emits an "uninitialized value" warning under use warnings. This is pre-existing but the fix now propagates it to more locales. Adding // '' when indexing would silence the warning safely.