[ColinLeeo]

Problem

Fixes #1038.

A C++ aligned table query using TimeIn could silently drop matching rows that follow a page whose rows were all rejected by the filter. The API still returned E_OK and reported a normal end of results.

Root cause

AlignedChunkReader::decode_tv_buf_into_tsblock_by_datatype converted a successful decode that produced no rows into E_NO_MORE_DATA:

if (ret_tsblock->get_row_count() == 0 && ret == E_OK) {
    ret = E_NO_MORE_DATA;
}

That conflates "this page had no matching rows" with "the data stream is exhausted". TsFileSeriesScanIterator::get_next reacts to E_NO_MORE_DATA by advancing to the next chunk, or by ending the series when the current chunk is the last one — so the remaining pages of the chunk were never read. The iterator already has the correct handling for an empty successful block; the rewrite made it unreachable on this path.

The multi-value aligned path and the offset-aware non-aligned path do not have this rewrite, which is why the bug only surfaced on the per-column aligned read path (e.g. when a requested measurement is never written and the query falls back to one SSI per column).

Fix

Return E_OK with zero rows for a fully filtered page and let the series iterator continue within the chunk. True end of stream is still signalled by has_more_data().

Regression test

TimeInContinuesAfterFullyFilteredAlignedPage in cpp/test/reader/table_view/tsfile_reader_table_test.cc reproduces the reported layout: three devices, three chunks of seven rows with an explicit flush, page_writer_max_point_num_ = 5, a ghost measurement that is never written, v missing entirely on one device, sparse NULLs in v/w, serial and parallel read, and a 4096-byte internal TsBlock budget. It covers:

• the reported TimeIn({0, 2, 3, 6, 7, 10, 13, 19, 20}) query, which lost timestamps 19 and 20 per device before the fix;
• a filter matching only the page after a fully filtered page with no successor chunk;
• a filter matching only the first chunk, leaving later empty pages with successor chunks;
• all-matching and no-matching controls.

Verification

• New test fails on the unmodified source with 7 timestamps per device and passes with the fix (serial and parallel).
• Full C++ suite: 1414 tests pass, 0 failures.
• The standalone reproducer from the issue now prints after scan ret=0 count=27 expected=27 and exits 0.

This is separate from the small-capacity hang tracked in #1035.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 66.46%. Comparing base (d22008b) to head (7526b00).
⚠️ Report is 8 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1044      +/-   ##
===========================================
+ Coverage    65.26%   66.46%   +1.20%     
===========================================
  Files          761      773      +12     
  Lines        53610    55796    +2186     
  Branches      8547     8945     +398     
===========================================
+ Hits         34986    37086    +2100     
- Misses       16978    17055      +77     
- Partials      1646     1655       +9     

☔ 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.