Major Release v3.0.0

What's Changed

New v3 major release. This is a major version with breaking changes covering two significant architectural
improvements.

ANTLR4 Grammar-Based Parsing (PR #149)

Sequence pattern parsing has been migrated from regex to a shared ANTLR4 grammar, aligning the Python
implementation with the Go and C++ implementations for consistency and maintainability.

Removed API:

• FileSequence.SPLIT_RE and FileSequence.DISK_RE class variables
• constants.SPLIT_PATTERN, constants.SPLIT_RE, constants.SPLIT_SUB_PATTERN, constants.SPLIT_SUB_RE

Removed files:

• setup.py → replaced by pyproject.toml
• src/fileseq/__version__.py → version now derived from git tags via setuptools-scm
• requirements.txt → replaced by pyproject.toml optional dependency groups (pip install ".[dev]", pip install ".[docs]")

Behavioral changes:

• Auto-padding now only applies to single-frame files without explicit padding — foo.1@@@@.exr preserves
  @@@@ rather than re-padding

Range-Based FrameSet (PR #150)

FrameSet now stores frame ranges compactly as Range objects rather than expanding all frames upfront,
reducing memory usage by ~99.9%+ for large ranges (e.g. 100k frames drops from ~7.8MB to ~536 bytes).

Deprecated API (raises DeprecationWarning):

• FrameSet.items — use set(fs) instead
• FrameSet.order — use list(fs) or fs[i] instead

Bug fixes:

• isConsecutive() now correctly handles interleaved ranges and empty framesets
• hasSubFrames() correctly detects decimal notation like "1.0-5.0"
• Stagger modifier (:) now properly deduplicates frames

Pull Requests

• Implement ANTLR4 grammar parsing and migration from regex focus (refs #147) by @justinfx in
  #149
• Refactor FrameSet to use range-based storage (fixes #148) by @justinfx in
  #150

Full Changelog: v2.3.1...v3.0.0