Merge an out-of-order fragment that extends an AoT element

[dchaudhari7177]

Fixes #577.

Problem

content = """\
[[y.a]]
[b.d.x]
[[y.a.c]]
"""
doc = parse(content)
assert doc.as_string() == content   # passes
doc.unwrap()                        # KeyAlreadyPresent: Key "a" already exists

tomllib reads this as {"y": {"a": [{"c": [{}]}]}, "b": {"d": {"x": {}}}}. Parsing and rendering are fine; only the read path fails.

Cause

[b.d.x] splits y, so reading it builds an OutOfOrderTableProxy over the two y fragments. The final header extends the last element of the existing y.a array rather than appending a new element, so the second a fragment arrives at the proxy as an implicit super table holding c — not as another AoT.

_merge_aot_fragment only handles AoT + AoT. It returned None for the super-table shape, the caller fell through to _raw_append, and that refused a as a duplicate key.

Fix

Handle the super-table shape in the same place: copy the array's last element, merge the fragment's body into the copy, and hand back a new AoT. Copying is what keeps the fragments the document renders from untouched — the same reason the existing AoT + AoT path builds a new AoT instead of mutating one.

The guard stays narrow: only a super table (never a concrete [y.a] redefinition) and only when the array actually has an element, so genuine duplicates still raise.

Tests

• test_out_of_order_table_merges_aot_element_extension — the reported document, with values on both sides, checking round-trip, unwrap() against the tomllib shape, and indexed access.
• test_out_of_order_table_extends_last_of_several_aot_elements — two elements before the split, confirming the extension lands on the last one only.

I also checked the neighbouring shapes against tomllib (a plain out-of-order AoT split, [y.a.c] as a plain table rather than an AoT, and two [[y.a.c]] headers separated by another table); all agree.

Full suite passes (1053 tests).

🤖 Generated with Claude Code

[dimbleby]

why is this better or worse than the existing pull request #578?

@dchaudhari7177 you should likely have your bot check for already-open pull requests before creating new ones