Alt cells double counting

[erdogmush]

@erdogmush commented on Mon Sep 12 2022

When an alt-cell results in a correct answer, the score appears to be multiplied. If more than one alt-cell is good, the score should still be the score specified in the rubric.

An example is for the formula rubric, when the same formula is copied to the alt-cell and both original cell and alt-cell are correct.

Test-mode should have test cases for this.

[erdogmush]

@colelloa I pulled the changes and ran the test. This is working for the formula rubric type, but not for others (because possibly the mistake is replicated in the same copy-paste code for other rubric types -- which makes refactoring the high-level rubric processing important.) To see it run this from the excel subfolder:

$> pysheetgrader TestModeKey.xlsx TestModeSubmission.xlsx --html-report-output Report.html

Check the html report by dragging it to your browser, and see that in RConstant sheet, B17 has a score of 3, while the max score is 1.5. This is because the an alt-cell is also correct so it's double-scoring.

Now run in --test-mode

$> pysheetgrader TestModeKey.xlsx TestModeSubmission.xlsx --test-mode --html-report-output Report.html

Check the report again: I made B17 to fail by setting the Expected Result column to incorrect deliberately so we won't forget this bug. When it's fixed, we can set it to correct so it doesn't fail any more.

Add similar tests to RCheck and RRelative to see whether this happens with those rubric types. The score should not double (you'll see the score if you run it in normal mode, without the --test-mode flag). If doubling make the tests fail so we will know by setting expected result to incorrect.

[erdogmush]

Let's improve the test-mode first so we can easily easily detect regressions before fixing this bug and refactoring.

[colelloa]

issue fixed in refactor, was due to cells not being marked correct properly