fix: preserve fractional seconds in non-ISO date parsing

[Koan-Bot]

What

Fractional seconds are now preserved when parsing non-ISO date formats.

Why

The time-of-day regex in strptime consumed the fractional seconds text (.5, .123, etc.) but used a non-capturing group, so the value was silently discarded. Only ISO 8601 formats preserved sub-second precision.

This meant str2time("15 Jan 2024 10:30:45.5 UTC") returned 1705314645.000 instead of 1705314645.500 — a quiet data loss that affects any caller relying on sub-second accuracy with non-ISO dates.

How

Changed (?:\.\d+)? to (?:\.(\d+))? in the time parsing regex (one character: \d+ → (\d+)), and updated the three subsequent capture group references ($5→$6, $6→$7) that shifted as a result.

Testing

8 new tests covering ISO 8601, day-month-year, RFC 2822-style, timezone abbreviations, AM/PM suffixes, and integer-only seconds. Full suite passes (1228 tests).

🤖 Generated with Claude Code

---

Quality Report

Changes: 2 files changed, 57 insertions(+), 4 deletions(-)

Code scan: clean

Tests: skipped

Branch hygiene: clean

Generated by Kōan post-mission quality pipeline

[greptile-apps]

Greptile Summary

This PR fixes a silent data-loss bug where the time-of-day regex in strptime consumed fractional seconds (.5, .123, etc.) via a non-capturing group, discarding them for all non-ISO date formats. The one-character change — adding a capture group to produce (?:\.(\d+))? — correctly captures the digits, and the three subsequent $N back-references ($5→$6, $6→$7) are shifted consistently. The $frac variable was already declared in scope and the $ss += "0.$frac" accumulation at line 217 already existed, so the fix slots in cleanly. Eight new tests verify the fix across common format variants.

Confidence Score: 5/5

Safe to merge — minimal, targeted one-character regex change with correct group-reference updates and comprehensive tests.

The change is surgical: one capture group added, three $N references incremented by 1, no new variables introduced, and the pre-existing $frac accumulation path at line 217 handles the value correctly. Regex trace confirms group numbering is exact. Eight new tests cover all impacted format paths. No regressions to pre-existing ISO behavior.

No files require special attention.

Important Files Changed

Filename	Overview
lib/Date/Parse.pm	Adds one capture group to the time-of-day regex ((?:\.(\d+))?) and increments the three downstream $N references by 1 to correctly populate $frac for non-ISO date strings.
t/frac-seconds.t	New test file with 8 cases covering ISO 8601 (already worked), day-month-year (the bug), RFC 2822-style, timezone abbreviation, AM/PM, and integer-only seconds regression check.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["strptime(dtstr)"] --> B{ISO compact YYYYMMDDTHHMMSS?}
    B -- yes --> C["Set hh,mm,ss,frac from capture groups (already captured .frac)"]
    B -- no --> D{ISO with delimiter YYYY-MM-DD?}
    D -- yes --> C
    D -- no --> E{Boost-style YYYY-Mon-DD?}
    E -- yes --> C
    C --> F{hh defined?}
    F -- yes --> G
    E -- no --> F
    F -- no --> H["Time-of-day regex [:\\s](hh):(mm)(:ss(?:\\.frac)?)?"]
    H -- "matched (FIXED) captures frac into $5; $6=z; $7=ap" --> I["Set hh,mm,ss,frac; $zone if $6; $merid if $7"]
    I --> G
    H -- no match --> G["Parse date components year/month/day/zone"]
    G --> J["ss += '0.' + frac if frac defined"]
    J --> K["Return ss,mm,hh,day,month,year,zone,century"]

Loading

_{Reviews (1): Last reviewed commit: "fix: preserve fractional seconds in non-..." | Re-trigger Greptile}