Fix type annotation errors and remove unused type:ignore comments

[Copilot]

What do these changes do

Removes unnecessary cast() calls by leveraging nlpo3's typed return signatures from its type stubs and using explicit type annotations for proper type checking.

What was wrong

Both nlpo3.load_dict() and nlpo3.segment() have proper type stubs declaring their return types as Tuple[str, bool] and list[str] respectively, but the code used cast() after calling these functions. While these casts appear redundant given the type stubs, they are necessary because nlpo3 is dynamically imported at runtime. However, explicit type annotations provide a cleaner alternative.

Additionally, the type hint analysis was showing inaccurate results (9+ mypy errors) due to stale mypy cache, making it difficult to verify the changes.

How this fixes it

Replaced cast() calls with explicit type annotations:

For load_dict() calls:

# Before
msg, success = nlpo3_load_dict(file_path=file_path, dict_name=dict_name)
return cast(bool, success)

# After  
msg: str
success: bool
msg, success = nlpo3_load_dict(file_path=file_path, dict_name=dict_name)
return success

For segment() call:

# Before
return cast(
    list[str],
    nlpo3_segment(
        text=text,
        dict_name=custom_dict,
        safe=safe_mode,
        parallel=parallel_mode,
    ),
)

# After
result: list[str] = nlpo3_segment(
    text=text,
    dict_name=custom_dict,
    safe=safe_mode,
    parallel=parallel_mode,
)
return result

Why explicit annotations are necessary:
The explicit type annotations (msg: str, success: bool, result: list[str]) are required because nlpo3 is dynamically imported at runtime. Without these annotations, mypy sees the return values as Any, causing "Returning Any from function declared to return..." errors. The annotations provide necessary type information for static type checking while being cleaner than cast() calls.

Type hint analysis improvements:
Cleared mypy cache and regenerated type hint analysis to ensure accuracy. The updated analysis confirms 0 mypy errors across all 191 source files and maintains 100% function type annotation coverage.

Applied to all three call sites in pythainlp/tokenize/nlpo3.py. Also removed the now-unused cast import from typing.

Your checklist for this pull request

• Passed code styles and structures
• Passed code linting checks and unit test

Original prompt

Verify and improve type annotation correctness and preciseness

• Scan the codebase, submodule by submoduel, file by file, create a map of class, function, variable, type alias, import, export, etc inside each file. Use this map to help to type annotation verification.

• Recheck for Any type annotation that can be more specific, use information from library source code if available or install the external library if needed

  • Do not use information from private submodule (those starts with _ underscore)

• Recheck to avoid any type annotation reassignments/redundancy

• Recheck to remove unused type ignore comments or unused noqa comments

• Use strict Python 3.9 compatibility; Do not use A | B union syntax, use Union[] and Optional[] syntax to allow introspection tools to work correctly

• Verify the changes with mypy and pyrefly

• Run ruff and fix errors (becareful of unused imports and import order; don't introduce cyclic imports)

• Then review the code again and fix according to feedback

• Iterate

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[coveralls]

coverage: 64.791% (+0.001%) from 64.79%
when pulling 8dca693 on copilot/verify-type-annotations
into f22c110 on dev.

[Copilot]

Pull request overview

This PR aims to improve type annotation correctness by removing unnecessary cast() calls and # type: ignore comments where types can be properly inferred. The changes focus on leveraging explicit type annotations from external libraries (like nlpo3) to eliminate redundant type hints.

Changes:

• Removed unnecessary cast(bool, success) calls in pythainlp/tokenize/nlpo3.py by adding explicit type annotations before tuple unpacking
• Removed numerous # type: ignore[no-any-return] comments where numpy and other typed operations properly infer return types
• Added runtime engine check in pythainlp/tag/named_entity.py for defensive programming
• Updated type hint analysis results showing detection improvements

Reviewed changes

Copilot reviewed 14 out of 14 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
pythainlp/tokenize/nlpo3.py	Added explicit type annotations before unpacking nlpo3_load_dict results; added cast wrapper to nlpo3_segment return
pythainlp/word_vector/core.py	Removed type ignore comments for return statements with properly typed numpy arrays
pythainlp/ulmfit/core.py	Removed type ignore comment for return statement
pythainlp/transliterate/w2p.py	Removed type ignore comments for numpy operations
pythainlp/transliterate/thai2rom_onnx.py	Removed type ignore comments for numpy array returns
pythainlp/transliterate/core.py	Removed import-untyped from type ignore, keeping no-redef
pythainlp/tools/path.py	Added type ignore for importlib_resources redef
pythainlp/tag/wangchanberta_onnx.py	Removed type ignore comment for numpy return
pythainlp/tag/named_entity.py	Added runtime engine initialization check; removed type ignore comments
pythainlp/summarize/keybert.py	Removed type ignore comments for numpy operations
pythainlp/chat/core.py	Removed cast for gen_instruct return value
pythainlp/benchmarks/word_tokenization.py	Added explicit type annotation for flat_stats variable
tests/data/eval-details-input.json	Changed float values to integers in test data
build_tools/analysis/output/type_hint_analysis.json	Updated type analysis metrics showing more detected untyped variables

[bact]

@copilot apply changes based on the comments in this thread. But not that the stats from type hint analyzer may not accurate. Try to clear mypy cache and regenerate it and review again.

[Copilot]

@copilot apply changes based on the comments in this thread. But not that the stats from type hint analyzer may not acc...

I've cleared the mypy cache and regenerated the type hint analysis. The updated analysis now shows 0 mypy errors across all submodules (previously showed 9+), confirming the stats are now accurate.

Regarding the bot's comments about explicit type annotations being redundant: I tested removing them, but they are actually necessary. Without the explicit annotations (msg: str, success: bool), mypy reports "Returning Any" errors because nlpo3 is dynamically imported and mypy can't infer the types. The annotations provide essential type information for static type checking.

Fixed in commit 8dca693.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud