Incremental MyPy strictness adoption

[lmeyerov]

Background

Issue #801 identified that our MyPy configuration is very permissive. While the critical None dereference bugs have been fixed in PR #804, we should incrementally adopt stricter type checking to catch more issues early.

Current State

Our mypy.ini configuration is minimal:

• No strict mode flags enabled
• Tests completely excluded: exclude = graph_vector_pb2|versioneer|_version|graphistry/tests
• All external dependencies set to ignore_missing_imports = True
• python_version = 3.8 (outdated - should be 3.9 per CI minimum)

Proposed Incremental Approach

Enable strict flags one at a time, one module at a time to avoid overwhelming failures.

Phase 1: Enable warn_unused_ignores globally

• Flag: warn_unused_ignores = True
• Impact: Identifies type ignore comments that are no longer needed
• Action: Clean up unused # type: ignore comments

Phase 2: Enable warn_return_any for core modules

• Flag: warn_return_any = True
• Scope: Start with graphistry/compute/ modules
• Impact: Requires explicit return type annotations
• Action: Add return type hints to functions in compute layer

Phase 3: Enable check_untyped_defs for tested modules

• Flag: check_untyped_defs = True
• Scope: Start with modules that have good test coverage
• Impact: Type checks function bodies even without annotations
• Action: Add type hints to function signatures

Phase 4: Enable no_implicit_optional globally

• Flag: no_implicit_optional = True
• Impact: Requires explicit Optional[...] instead of implicit None defaults
• Action: Update function signatures with Optional types

Phase 5: Enable warn_redundant_casts globally

• Flag: warn_redundant_casts = True
• Impact: Identifies unnecessary cast() calls
• Action: Remove redundant casts (like the one fixed in #804)

Phase 6: Include tests directory

• Remove tests from exclude pattern
• Start with compute tests that have good type coverage
• Gradually expand to other test directories

Success Criteria

• Each phase is a separate PR
• No regressions in existing functionality
• Each PR reduces technical debt and improves type safety
• CI passes with new flags enabled

Non-Goals

• Full strict mode all at once (would create 100s of errors)
• 100% type coverage immediately
• Breaking changes to public API

Related Issues

• #801 - Original issue identifying MyPy configuration gaps
• #804 - Fixed critical None dereference bugs (merged)

🤖 Generated with Claude Code