[ariostas]

🤖 AI text below 🤖

Part of the split of #1814 (the list is in #1834). These bugs were found while writing the coverage tests of #1814, and tests/test_1827_fix_writing.py holds their regression tests: every test in it fails on main (for parametrized tests, at least one of their cases).

• Directories: without cycle numbers, a name stands for its highest cycle that passes the filters, in both ReadOnlyDirectory and WritableDirectory (keys, items, classnames), and WritableDirectory.values() lists each name once, as its highest cycle, like ReadOnlyDirectory.values(). The writable directory also stops listing a name twice and filters subdirectories on their full paths.
• Keys and streamers
  • A key that moves can become a big key, so keys that move past 2 GB are written correctly.
  • Growing the list of streamers allocates the size it uses and sets the file length.
  • Streamers that aren't TStreamerInfo raise a TypeError.
  • FileSink.tell doesn't reopen a file that is open.
• TTrees
  • A branch name that is defined twice raises a ValueError. Before, the second one was silently left out, and when it was a field of a record whose flattened name collided with another branch ({"r_x": ..., "r": {"x": ..., "y": ...}}), the TTree could never be extended. A scalar integer branch can still name the counter of a jagged branch before it.
  • Branch types can be any Mapping, and type strings that make NumPy raise a SyntaxError are understood.
  • Categorical strings (an Awkward IndexedArray) can be written.
  • Jagged arrays that can't be written raise a TypeError that names the branch.
  • extend and add_branches refuse trees with branches they can't interpret, instead of leaving those branches behind. These include custom interpretations without a dtype, so Interpretation's base methods now raise NotImplementedError naming the subclass, instead of a bare AssertionError.
• Histograms and other objects
  • to_TArray picks the class from the dtype's kind and size, so long/long long/intc arrays work on every platform. On Windows with NumPy 1, int32 data was rejected.
  • Histogram-like objects whose values() doesn't take flow can be written: their fallback still called values(flow=False).
  • Tuples of arrays with 0 or more than 3 dimensions raise a TypeError.
  • The as_TGraph error about maxY names maxY.
• Copying: copy_from doesn't try to make a directory named "" for the root of the copy.

[codecov]

Codecov Report

❌ Patch coverage is 93.07692% with 9 lines in your changes missing coverage. Please review.
✅ Project coverage is 78.76%. Comparing base (b4acc0f) to head (551ded0).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
src/uproot/writing/_cascadetree.py	89.28%	4 Missing and 2 partials ⚠️
src/uproot/writing/identify.py	50.00%	1 Missing and 1 partial ⚠️
src/uproot/writing/_cascade.py	94.73%	0 Missing and 1 partial ⚠️

❌ Your patch check has failed because the patch coverage (93.07%) 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/__init__.py	96.96% <100.00%> (+18.18%)	⬆️
src/uproot/reading.py	83.18% <100.00%> (+0.12%)	⬆️
src/uproot/sink/file.py	91.66% <100.00%> (+0.82%)	⬆️
src/uproot/writing/interpret.py	72.81% <ø> (+19.41%)	⬆️
src/uproot/writing/writable.py	89.46% <100.00%> (+1.80%)	⬆️
src/uproot/writing/_cascade.py	91.02% <94.73%> (+2.62%)	⬆️
src/uproot/writing/identify.py	82.75% <50.00%> (+8.89%)	⬆️
src/uproot/writing/_cascadetree.py	90.22% <89.28%> (+2.74%)	⬆️

... and 1 file with indirect coverage changes

[ianna]

@ariostas -- Thanks! Looks good. The coverage is low but the subsequent PRs should fix it I think. Let's merge the tests first and update this one to check it. Thanks!