Type checking for channel_metrics

[purva-thakre]

Description

Related to #299

We use --explicit-package-bases here to disable mypy errors due to 2 files of the same name existing in different modules.

Changes

Right now ~/toqito$ mypy toqito/channel_metrics --explicit-package-bases is failing due to the following. These are mostly due to incompatibilities with type annotations in cvxpy compared to how we use them.

toqito/channel_metrics/diamond_norm.py:102: error: Argument 1 to "Parameter" has incompatible type "list[int]"; expected "int | tuple[int, ...]"  [arg-type]
toqito/channel_metrics/completely_bounded_trace_norm.py:67: error: Argument "dims" to "partial_trace" has incompatible type "tuple[int, int]"; expected "tuple[int]"  [arg-type]
toqito/channel_metrics/completely_bounded_spectral_norm.py:35: error: Module not callable  [operator]

Will need to refactor portions of:

• diamon_norm.py:

toqito/toqito/channel_metrics/diamond_norm.py

Line 102 in 1122a6e

j_var = cvxpy.Parameter([dim_squared, dim_squared], complex=True)

Here, [dim_squared, dim_Squared] is of type List[int] whereas cxpy.Parameter expects shape: int | tuple[int, ...]
https://www.cvxpy.org/api_reference/cvxpy.expressions.html#parameter

• completely_bounded_trace_norm.py

Same as the previous item, we use an incompatible type compared to what's expected by cvpy type annotation. It expects tuple[int] but we use tuple[int. int].

toqito/toqito/channel_metrics/completely_bounded_trace_norm.py

Line 67 in 1122a6e

cp.norm(cp.partial_trace(y0, dims=(dim, dim), axis=1)) + cp.norm(cp.partial_trace(y1, dims=(dim, dim), axis=1))

https://www.cvxpy.org/api_reference/cvxpy.atoms.affine.html#partial-trace

• completely_bounded_spectral_norm.py

For some reason, completely_bounded_trace_norm or dual_channel shows up as a module not callable.

toqito/toqito/channel_metrics/completely_bounded_spectral_norm.py

Line 35 in 1122a6e

return completely_bounded_trace_norm(dual_channel(phi))

If I specify the full path for completely_bounded_trace_norm as completely_bounded_trace_norm.completely_bounded_trace_norm, mypy fails with the following error:

toqito/channel_metrics/completely_bounded_spectral_norm.py:35: error: Argument 1 to "completely_bounded_trace_norm" has incompatible type "ndarray[Any, Any] | list[list[ndarray[Any, Any]]]"; expected "ndarray[Any, Any]"  [arg-type]

Checklist

Before marking your PR ready for review, make sure you checked the following locally. If this is your first PR, you might be notified of some workflow failures after a maintainer has approved the workflow jobs to be run on your PR.

Additional information is available in the documentation.

• Use ruff for errors related to code style and formatting.
• Verify all previous and newly added unit tests pass in pytest.
• Check the documentation build does not lead to any failures. Sphinx build can be checked locally for any failures related to your PR
• Use linkcheck to check for broken links in the documentation
• Use doctest to verify the examples in the function docstrings work as expected.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.1%. Comparing base (1122a6e) to head (a7381f9).

Additional details and impacted files

@@          Coverage Diff           @@
##           master    #623   +/-   ##
======================================
  Coverage    98.1%   98.1%           
======================================
  Files         162     162           
  Lines        3112    3115    +3     
  Branches      759     759           
======================================
+ Hits         3054    3057    +3     
  Misses         37      37           
  Partials       21      21           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[purva-thakre]

added this because not all libraries use type annotations.

Skipping analyzing "picos": module is installed, but missing library stubs or py.typed marker

[purva-thakre]

this is an inline type annotation as requested by mypy

Need type annotation for "sys" (hint: "sys: list[<type>] = ...") [var-annotated]

[purva-thakre]

I think the type annotation for dims argument in partial_trace is wrong.

It is dims: Tuple[int] when it should be dims: Tuple[int, ...]. We see the latter being used by cvxpy.partial_transpose.

https://www.cvxpy.org/api_reference/cvxpy.atoms.affine.html#partial-transpose

To Do: Understand if I need to raise this in the cvxpy repo.

[purva-thakre]

Don't know if this is a best practice hack.

Due to dual_channel, the return type of dual_channel(phi) is np.ndarray | List[List[np.ndarray]] when completely_bounded_trace_norm expects just a np.ndarray. So, added an additional step of converting dual_channel(phi).

If completely_bounded_trace_norm.completely_bounded_trace_norm was not specified, mypy failed with toqito/channel_metrics/completely_bounded_spectral_norm.py:35: error: Module not callable [operator] but python could not import this function to run unit tests.

[purva-thakre]

temporarily ignoring this line due to #623 (comment)