[Aias00]

Closed: #1149

Summary

• make numRows(0) follow the documented unlimited behavior
• cover workbook-level and sheet-level numRows(0) reads across formats

Verification

• ./mvnw -pl fesod-sheet -Dmaven.test.skip=false -DskipTests=false -Dtest=ParameterDataTest#numRowsZeroReadsAllRows test
• ./mvnw -pl fesod-sheet spotless:check

[BigDataDZ]

Small suggestion from a fellow traveler on #1149 (my overlapping #1189 — closing it in favor of this one, the core guard is identical). While testing I also covered two neighboring cases that might be worth guarding here: numRows omitted entirely (must keep reading all rows) and a positive limit (must still stop early), plus a headless-sheet setup so the limit maps 1:1 onto data rows. Happy to send these as a follow-up PR if useful.

[BigDataDZ]

The core guard is exactly right - limit > 0 matches the documented sentinel, and it is the same shape I had shipped in the closed #1189 (verified there against 9 format/level combinations).

Two of the neighboring cases from the Oct-9 comment are still uncovered, and they protect the other half of the condition:

1. A positive limit must still stop the read - limit > 0 exists to disable the zero sentinel, but nothing pins that a legitimate numRows(1) still truncates. Without that case, a future "simplification" to limit == null || limit <= 0 || index < limit-style refactors could silently read everything.
2. Omitted numRows stays unlimited - the limit == null fall-through from sheet holder to workbook holder.

One testing detail from #1189 that will save you an iteration on case 1: lastRowIndex counts the head row (a headed file's first data row sits at index 1), so numRows(1) on a headed default read delivers zero data rows. Writing a header-less file (or headRowNumber(0)) makes the limit map one-to-one onto data rows - my closed branch has ready-made parameterized tests for all three cases (zero / omitted / positive across XLSX, XLS, CSV) if you want to lift them: BigDataDZ/fesod@f4e0ec7 (NumRowsReadLimitTest).

Happy to review the follow-up commit. Otherwise LGTM on the fix itself.

[bengbengbalabalabeng]

LGTM.