Coverage report from unit tests is inaccurate

Using the matlab.unittest.plugins.CodeCoveragePlugin with class-based unit tests produces coverage reports that are not always accurate. I see them repeatedly (but not always) occur when function handles are split across multiple lines. The following image provides an example (code to reproduce it is attached).

The "missed line" goes away if you replace isequal(x) with x == x in line 4. Also, if you just remove the ellipsis and have the function handle defined entirely on one line, the report does not display a "missed line."

Answer 1

[Neeraj Badamikar]

Hello Micah,

Thank you for reporting this. We've had a few other users reporting about this issue and we are working on addressing it.

One possible solution we're looking at is to report code coverage on entire MATLAB statements as opposed to lines. That would ensure that a statement like this -

function_handle_with_wrapping_line_for_isequal =  @(x) ~isequal(x, x) || ... 
      isequal(x, x);

would either be marked as covered or not covered in entirety.

Is that something that you would find useful?

Regards,

Neeraj

[Micah Fry]

My main concern is seeing a reported "lack of coverage" when there isn't one. That messes with the code coverage metrics as well. Whether the report is by statement or by line, as long as I'm not seeing erroneous red lines (and getting lower coverage percentages), my issue is resolved.

The example was purposefully made to illustrate that the execution MUST be going through the last line (even though it reports that the line wasn't touched). So if the MATLAB's reporting plugin continues to focus on lines rather than statements, a fix would need to show 100% coverage of the example I provided.

[Neeraj Badamikar]

Thank you for your feedback, Micah.

[Micah Fry]

Is there a bug tag I can keep track of for this issue? It looks like this may be related to another question (https://www.mathworks.com/matlabcentral/answers/575389-some-lines-missed-in-code-coverage-report).

[Micah Fry]

I see with R2022a this issue is fixed! Thanks @Neeraj Badamikar!