fix: entry-range arithmetic in iterate/concatenate and basket selection

[henryiii]

🤖 AI text below 🤖

This PR fixes three coupled entry-range arithmetic bugs in iterate, concatenate, and basket selection.

1. Off-by-one extra basket on exact entry_stop boundaries

When entry_stop landed exactly on a basket boundary, the basket collector read, decompressed, and interpreted one extra basket contributing zero entries.

Reproducer (3 baskets of 10 entries, offsets [0,10,20,30]):

b.entries_to_ranges_or_baskets(0, 10)  # selected baskets [0, 1], should be [0]

The condition start <= entry_stop in entries_to_ranges_or_baskets and _hasbranches_num_entries_for (src/uproot/behaviors/TBranch.py) selected a basket whose start == entry_stop. The matching per-basket trim loops in the interpretations (numerical.py, jagged.py, strings.py, objects.py) had the same start <= entry_stop <= stop / start <= entry_stop boundary conditions.

Because the interpretation expects exactly the baskets the collector supplied, both sides are fixed together (start < entry_stop), and the degenerate empty-range case (entry_start == entry_stop) is handled explicitly so empty reads still work. custom.py uses a numpy.where-based selection that does not have this off-by-one, so it was left unchanged.

After the fix, ranges are minimal and results are byte-identical for non-boundary ranges.

2. concatenate(..., allow_missing=True) shifted offsets after a skipped file

When hasbranches.arrays raised KeyInFileError (a requested branch missing from one file) and the loop continued, it skipped advancing global_start = global_stop, so every subsequent file was read at offsets shifted by the skipped file's entry count.

Reproducer (3 files x 10 entries, middle file missing branch x, entry_start=5, entry_stop=25):

• Before: 15 entries returned (values 5-9, 20-29, including out-of-range 25-29)
• After: 10 entries (values 5-9, 20-24), correct

Fixed by advancing global_start = global_stop in the allow_missing continue path, in both TBranch.concatenate and RNTuple.concatenate.

3. RNTuple.HasFields.iterate did not clamp per-step entry_stop

iterate passed entry_stop=start + step_size per step without clamping to the user's entry_stop (only clamped to num_entries inside arrays()).

Reproducer (50000-entry RNTuple):

nt.iterate(step_size=30000, entry_stop=40000)  # yielded 50000 entries, should be 40000

Fixed with entry_stop=min(start + step_size, entry_stop). With report=True the ranges now come out as [(0, 30000), (30000, 40000)].

Testing

• prek -a --quiet: clean
• Existing suites pass: tests/test_0034*, tests/test_0017*, test_0016_interpretations, test_0043_iterate_function, test_0044_concatenate_function, test_0067_common_entry_offsets, test_0194_fix_lost_cuts_in_iterate, test_0335_empty_ttree_division_by_zero, test_1573_out_of_bounds_slicing, test_1630_rntuple_iterate_report, test_1406_improved_rntuple_methods, test_1250_rntuple_improvements, plus a wide keyword sweep (iterate/concatenate/basket/interpret/string/jagged/entry): all pass.

Regression tests will be added in a follow-up commit on this branch.

🤖 Generated with Claude Code

Part of #1646.

[henryiii]

🤖 AI text below 🤖

Regression tests added in tests/test_1656_entry_range_arithmetic.py (commit 142b01f). They cover: no extra basket on exact entry_stop boundaries (numerical/jagged/strings interpretations), correct boundary and empty-range reads, concatenate(allow_missing=True) keeping global offsets aligned when a file is skipped, and RNTuple.iterate clamping the per-step entry_stop. Three of the five tests fail on the pre-fix code and pass after.

Note on the RNTuple concatenate change: it is applied for symmetry with the TTree fix and matches the reported issue, but in current code an RNTuple arrays() call with a missing field returns empty records rather than raising KeyInFileError, so the allow_missing continue path there is not reachable via a black-box reproducer and is not covered by a dedicated test.

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 73.80%. Comparing base (1c06db0) to head (2af1b83).
⚠️ Report is 15 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/uproot/behaviors/RNTuple.py	0.00%	1 Missing ⚠️
src/uproot/behaviors/TBranch.py	66.66%	0 Missing and 1 partial ⚠️

❌ Your patch check has failed because the patch coverage (83.33%) is below the target coverage (98.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

Files with missing lines	Coverage Δ
src/uproot/interpretation/jagged.py	90.27% <100.00%> (ø)
src/uproot/interpretation/numerical.py	82.21% <100.00%> (ø)
src/uproot/interpretation/objects.py	83.96% <100.00%> (ø)
src/uproot/interpretation/strings.py	71.36% <100.00%> (+0.93%)	⬆️
src/uproot/behaviors/RNTuple.py	70.84% <0.00%> (-0.14%)	⬇️
src/uproot/behaviors/TBranch.py	83.73% <66.66%> (+0.42%)	⬆️

... and 1 file with indirect coverage changes