fix: keep TTree branch lookup indices valid when a generated counter collides

[ariostas]

🤖 AI text below 🤖

Addresses finding 1 of "PR 5" in #1688.

Tree._branch_lookup maps a branch name to its index in Tree._branch_data. When a jagged branch's generated counter name collides with a branch that was already declared, the colliding datum was removed with del:

https://github.com/scikit-hep/uproot5/blob/main/src/uproot/writing/_cascadetree.py#L203-L207

which shifts every later datum down by one, while _branch_lookup keeps pointing at the old indices.

f.mktree("t", {"nx": np.dtype("int32"), "y": np.dtype("float64"), "x": <jagged>})
# _branch_lookup == {'nx': 1, 'y': 1, 'x': 2}

y aliases the generated counter, and extending the tree fails.

Changes, depending on what the counter collides with:

• A scalar integer branch (the documented case of providing a counter explicitly): the datum is replaced in place instead of deleted. That also keeps the counter ahead of the jagged branch it counts, since a colliding name must have been declared before the branch that generates it.
• Another generated counter (jagged branches sharing a counter, e.g. counter_name=lambda n: "nJet" for Jet_pt and Jet_eta): the existing counter datum is reused. Previously the earlier branch kept a reference to a counter that was never written, and writing the tree failed with a struct.error. extend also failed when comparing the shared counter, because awkward.to_numpy rejects its big-endian dtype; it now uses numpy.asarray.
• Anything else (a float, a fixed-size array, a jagged branch, a record, or a record field): mktree now raises a ValueError. Previously the declared branch was silently replaced, leaving orphaned branches (the jagged branch's own counter, or the record's fields) that made extend fail with unrelated errors.
• A record or record field declared after the counter: mktree also raises a ValueError. Previously the record or field was silently dropped, or, for a dict record, appended as a second branch with the counter's name.

Note this needs mktree to reproduce: since v5.7.0 dict assignment produces an RNTuple, which does not go through this code.

[codecov]

Codecov Report

❌ Patch coverage is 73.91304% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 76.16%. Comparing base (e598b61) to head (d75479e).
⚠️ Report is 22 commits behind head on main.

Files with missing lines	Patch %	Lines
src/uproot/writing/_cascadetree.py	73.91%	3 Missing and 3 partials ⚠️

❌ Your patch check has failed because the patch coverage (73.91%) 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/writing/_cascadetree.py	85.10% <73.91%> (+0.99%)	⬆️

... and 20 files with indirect coverage changes

[Copilot]

Copilot review overview

🟡 Changes recommended

Record/counter collision cases can still be silently ignored or create duplicate/orphaned branches.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1

Open (1)

• Reject counter collisions with previously registered record fields · New

What changed in this PR

Fixes TTree branch lookup corruption caused by generated counter-name collisions during mktree and extend.

Changes:

• Replaces scalar counter collisions in place.
• Reuses compatible generated counters.
• Rejects incompatible collisions and adds regression tests.

File	Summary
src/​uproot/​writing/​_cascadetree.py	Updates counter collision and comparison handling.
tests/​test_1698_cascadetree_counter_collisions.py	Adds collision regression tests.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.