feat: support insert_embed shared types in Text

[samuelstecher]

Summary

Support embedding shared types (Array, Map, Text) in a Text via insert_embed(), returning a live, mutable handle to the integrated child.

This brings Text to parity with XmlText, whose insert_embed already integrates shared types (added in #390, released in v0.13.1). Text.insert_embed still hardcoded the immutable-JSON path (py_to_any) and discarded the integrated reference returned by yrs.

The underlying yrs library already returns the integrated ArrayRef/MapRef/TextRef from Text::insert_embed for preliminary types, so this simply exposes the existing capability through the pyo3 bindings, reusing the same _do_and_integrate path already used by Map/Array/XmlText.

Motivation

Enables inline, mutable nested structures inside a single collaborative Text avoiding expensive re-inserts.

Changes

• src/text.rs: Add insert_array_prelim, insert_map_prelim, insert_text_prelim to Text (mirror the existing XmlText equivalents).
• python/pycrdt/_text.py: Dispatch BaseType embeds through _do_and_integrate (same pattern as XmlText.insert_embed); primitives keep the existing path.
• tests/test_text.py: Add test_insert_embed_shared_types (integration + mutability + attrs + diff() types).
• CHANGELOG.md: Add entry.

Tests

> python -m pytest tests/test_text.py -v
============================= test session starts ==============================
platform darwin -- Python 3.12.12, pytest-9.1.1, pluggy-1.6.0
collected 14 items

tests/test_text.py::test_iterate PASSED                                  [  7%]
tests/test_text.py::test_str PASSED                                      [ 14%]
tests/test_text.py::test_api PASSED                                      [ 21%]
tests/test_text.py::test_to_py PASSED                                    [ 28%]
tests/test_text.py::test_prelim PASSED                                   [ 35%]
tests/test_text.py::test_slice PASSED                                    [ 42%]
tests/test_text.py::test_formatting PASSED                               [ 50%]
tests/test_text.py::test_insert_embed_shared_types PASSED                [ 57%]
tests/test_text.py::test_observe PASSED                                  [ 64%]
tests/test_text.py::test_iterate_events[asyncio] PASSED                  [ 71%]
tests/test_text.py::test_iterate_events[trio] PASSED                     [ 78%]
tests/test_text.py::test_sticky_index[to_json] PASSED                    [ 85%]
tests/test_text.py::test_sticky_index[encode] PASSED                     [ 92%]
tests/test_text.py::test_sticky_index_transaction PASSED                 [100%]

============================== 14 passed in 0.15s ==============================

The full suite passes (238 passed) with 100% coverage maintained; mypy python and ruff/ruff-format are clean.

Made with Cursor

[davidbrochart]

Sorry @samuelstecher, my suggestion in the test was wrong. The types are not the Python types but the integrated types indeed (e.g. not Text but _pycrdt.Text), which raises the question whether they should be the former instead of the latter?

[samuelstecher]

@davidbrochart its a good point. I originally submitted it like this because these assertions match the test suite of XmlText.insert_embed (see test_xml.py L149-151 in def test_text())

I agree that it would be more convenient if Text.diff() returned the python types instead of the integrated ones. I'll see what that entails for both Text and XmlText.

This would be a breaking change though right?

[davidbrochart]

@davidbrochart its a good point. I originally submitted it like this because these assertions match the test suite of XmlText.insert_embed (see test_xml.py L149-151 in def test_text())

Indeed, I didn't remember I wrote that.

This would be a breaking change though right?

Unfortunately yes, but I think it's better like that since as you said "this matches how Map and
Array already return values from __getitem__".

[samuelstecher]

@davidbrochart I pushed the changes that align both Text and XmlText.
Please review and let me know if there's anything more you need from me to land this PR :)

[davidbrochart]

Thanks @samuelstecher, that looks good to me.
I don't think this can be considered a breaking change, I rather see it as a fix since previously the returned integrated types were pretty useless, and I doubt anyone ever used them.