Python File Splitter

Requires ast-grep: npm install -g @ast-grep/cli

Full Workflow

Phase 1: Parse & Plan

0. If you have not yet read the file in question DO NOT READ IT YET. IT MIGHT BE TOO BIG, WAIT AFTER YOU RUN parse.
1. Copy scripts/split_module.py to working directory (using cp)
2. Run parse:
   bashCopy

   python split_module.py parse path/to/module.py
   

3. Analyze output, apply grouping heuristics (see below)
4. Create groupings.json
5. Present proposed structure to user, wait for confirmation
   • (Mode 1 stops here if user just wants to see the plan)

Phase 2: Split & Iterate

6. Run split:
   bashCopy

   python split_module.py split path/to/module.py groupings.json package.module Class1 Class2
   

7. If imports succeed → done!
8. If imports fail → rollback happens automatically, then:
   • Fix groupings.json (add missing items to base, adjust groups)
   • Re-run step 6
   • Repeat until imports pass

The Retry Loop

split → test fails → rollback → fix groupings.json → repeat

NEVER edit generated files. They get deleted on rollback. Only groupings.json survives - that's your source of truth.

CLI Reference

# Parse: print definitions JSON
python split_module.py parse <file.py>

# Split: write files, test imports, rollback on failure
python split_module.py split <file.py> <groupings.json> <module> <Name1> [Name2 ...]

Groupings JSON format

{
  "base": ["BaseClass", "helper_func"],
  "groups": {
    "group_name": ["ClassA", "ClassAVariant"],
    "other": ["ClassB"]
  },
  "init_extras": ["get_instance", "create_thing"]
}

Grouping heuristics

Apply in order:

1. Base classes → base: parent is ABC/Protocol/BaseModel, or name contains Base/Mixin
2. Inheritance chains → same group: class + all subclasses that inherit from it
3. Naming patterns → same group: FooRenderer, FooVLRenderer, FooDisableThinkingRenderer
4. Helper functions → with related class if name matches, else base
5. Factory functions → init_extras: get_*, create_*, make_*, build_*

Output structure

module.py      → re-export stub (backwards compat)
module/
├── __init__.py   # re-exports + factory functions
├── base.py       # imports + base classes + shared utils
└── <group>.py    # from .base import * + group members

FAQ

Q: Imports failed. Can I just edit base.py to fix it? A: No. Rollback deletes generated files. Edit groupings.json instead, then re-run. That's the whole point of the retry loop.

Q: Why does rollback exist? A: It protects you. Failed state = broken code in the repo. Rollback restores clean state so you can iterate on groupings.json safely.

Q: cmd_parse didn't capture something (type aliases, constants, etc). What do I do? A: This will happen. When imports fail, read the error, figure out what's missing, and either add it to base in groupings.json or fix the script's parse_definitions() if it's a pattern you'll hit often.

Q: What if I need to edit split_module.py to capture new patterns? A: Prefer using ast-grep patterns over regex. The script already uses run_ast_grep() for classes and functions. Add new patterns following the existing style (see parse_definitions()).