[quinnj]

Delimiter detection could elect a candidate that does not appear in the header row, so a one-column file of times or timestamps was split on : and kept only the hour (with long-row problems):

Input	CSV 0.10.16	CSV 1.1.0	This PR
ts\n2026-09-28T01:02:03.123456789\n	2026-09-28T01	2026-09-28T01	the timestamp
event time\n10:30:00\n11:45:00\n	Time values	[10, 11]	Time values
Created Date\n2026-09-28T01:02:03\n2026-09-28T01:02:04\n	DateTime values	2026-09-28T01 twice	timestamps

The first row passed the byte-count "count divisible by the line count" tier: two colons over two lines. The other two passed the data-only fallback, which 1.0 added, so they are 1.0 regressions.

Change: a delimiter that does not split the first surviving row would leave the header one column wide, so detection keeps , in that case. The divisibility tier could only elect such candidates, so it is removed. An explicit delim is unaffected. The existing test that pinned : for "header text\n1:2\n3:4\n" now expects ,, which is what CSV 0.10 returns.

Validation:

• Regression tests: one, two, and ten rows of times and timestamps under a plain (ts) and a spaced (event time) header, both scanners, other punctuation, every reader, and colon-delimited controls with and without a header.
• Full suite, Julia 1.12.6, bounds checks, four threads: 331,778 passed.

Co-authored by Codex

AI disclosure: This work was prepared with assistance from OpenAI Codex and Claude Code.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.79%. Comparing base (a0335ff) to head (f5abfe9).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1211      +/-   ##
==========================================
+ Coverage   95.75%   95.79%   +0.04%     
==========================================
  Files           9        9              
  Lines        5489     5566      +77     
==========================================
+ Hits         5256     5332      +76     
- Misses        233      234       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[quinnj]

I refreshed the source review against base a0335ff5bcbd3abe41a2f6ee9cb1b00187503937 and ran the complete package suite on Julia 1.12.6 with bounds checks and four threads: all 331,670 checks pass. The shared detector handles the single-column timestamp reproducer and the existing reader-mode/colon-delimiter regressions without an additional source finding.

All 16 executed Actions jobs pass on this exact head; performance was intentionally skipped. The current codecov/project check reports a small coverage drop (95.73% versus 95.75%), while patch coverage passes. That check and the required independent approval remain outstanding, so this is not yet ready to merge.

AI disclosure: This work was prepared with assistance from OpenAI Codex.