v0.6.1

This release focuses on closing behavioral gaps with the upstream TypeScript ProseMirror modules. Over 25 bugs have been fixed to improve correctness and compatibility.

Upstream Compatibility

UTF-16 Semantics

• Fix UTF-16 slicing in TextNode.text_between and Fragment.text_between — Python string slicing uses codepoints, but ProseMirror positions are in UTF-16 code units
• Fix UTF-16 character counting in diff.py — find_diff_start and find_diff_end produced wrong positions for text with non-BMP characters (e.g. emoji)

DOM Parsing

• Fix parse_slice() missing the lxmltext wrapping needed for lxml to handle text nodes (was only applied in parse())
• Remove incorrect preserve_whitespace=True default in parse_slice() — upstream lets it inherit from context
• Fix ruleFromNode fallback logic — when ruleFromNode returned falsy, the matchTag result was dropped instead of used as fallback

DOM Serialization

• Handle {dom, contentDOM} dict variant in render_spec, matching the TypeScript DOMOutputSpec type
• Coerce attribute values to string in render_spec — TypeScript's setAttribute does this implicitly
• Cache DOMSerializer in from_schema() (upstream caches on schema.cached)

Node & Fragment

• Fix Node.cut() not resolving default to parameter before the early-return check
• Add text attribute to base Node class (returns None), matching Node.prototype.text = undefined in TypeScript
• Reject negative indices in child() and maybe_child() — Python lists silently accept them, TypeScript arrays don't
• Make Fragment.cut_by_index require both parameters to match upstream API surface

Schema & Attributes

• Distinguish explicit None from missing in compute_attrs using a sentinel — TypeScript distinguishes null from undefined
• Map dict and list to "object" in _validate_type to match JavaScript's typeof
• Add else clause for unknown expression types in NFA compile (upstream throws, Python silently returned None)

Resolved Positions

• Make ResolvedPos.index_after depth parameter optional, matching upstream

Content Matching & Replace

• Fix check_content usage in close() to match upstream error semantics — was raising ReplaceError instead of ValueError
• Fix infinite loop in matches_context()
• Fix operator precedence in NodeContext.finish()
• Fix logic inversion in ignore_fallback()

Other API Fixes

• Change textBetween defaults to None to match upstream (was defaulting to empty strings)
• Add diagnostic hint to Fragment.from_() error message about multiple versions
• Add missing exports to prosemirror.model: AttributeSpec, GenericParseRule, HTMLOutputSpec, MarkSpec, NodeSpec, ParseOptions, ParseRule, SchemaSpec, StyleParseRule, TagParseRule

Bug Fixes

• Fix memory leak in resolve cache — added weakref finalizer to clean up entries when documents are garbage collected

CI & Tooling

• Add Python 3.14 to test matrix
• Add pre-commit config for ruff, ty, and pytest