@williamjallen @JManion32
This PR fixes #10702 — tolerance grading was incorrectly accepting positive numbers when negative values were expected, due to the minus sign being stripped in [isolateAlphanumAndNumberPunctuation]. Would appreciate a review when you get a chance.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 21.64%. Comparing base (7f44c21) to head (c088c06).
⚠️ Report is 1 commits behind head on main.

@@            Coverage Diff            @@
##               main   #12569   +/-   ##
=========================================
  Coverage     21.64%   21.64%           
  Complexity     9654     9654           
=========================================
  Files           268      268           
  Lines         36244    36244           
  Branches        487      487           
=========================================
  Hits           7845     7845           
  Misses        27916    27916           
  Partials        483      483           

Hi @Akarshkushwaha
I tried to test this PR locally, but the problem doesn't seem to be fixed, as shown in the screenshot below:

I'm not sure if it's an issue in the code or with the way I tested the problem.
Here is my testing config.json:

{
    "assignment_message": "Calculate average and standard deviation given a list of number by filtering only positive numbers. Result expected with a precision of three decimal places.",
    "required_capabilities": "python",
    "autograding_method": "docker",
    "container_options": {
        "container_image": "submitty/python:latest"
    },
    "testcases": [
    {
      "title" : "Allow Minor Numeric Noise (+/-0.05 Tolerance)",
      "command" : [ "python3 *.py" ],
      "points" : 10,
      "validation": [
        {
          "description": "diff w/ tolerance=0.05",
          "method": "diff",
          "tolerance": 0.05,
          "actual_file": "STDOUT.txt",
          "expected_file": "test_output.txt",
          "deduction": 1,
          "failure_message":"These files have numeric errors > 0.05 or whitespace or text differences."
        }
      ]
    }
  ]
}

I rebuilt the autograder using submitty_install, but it's possible this didn't recompile the autograder correctly. Can you provide more specific details on how you compiled you change?

Hi @Eli-J-Schwartz Thankyou for testing this out locally.

Your config.json is perfectly correct! I believe the issue is simply with the way the C++ autograder builds local executables.

When you ran sudo submitty_install, the build system successfully ran CMake to compile my updated C++ changes into the core libsubmitty_grading.a library.

However, because Submitty's autograder relies on static linking, existing gradeables do not automatically inherit these changes. Their validate.out executables are still statically linked to the old, unpatched version of the library. Running submitty_install updates the core files, but it doesn't automatically recompile existing gradeables.

How to test the fix:

1. Keep your exact config.json—it is correct!
2. Go to your neg_tol gradeable in the web UI -> Edit Gradeable -> Autograding.
3. Click the "Build Gradeable" button at the bottom.

Clicking "Build Gradeable" forces Submitty to re-run CMake specifically for that assignment, linking the newly patched libsubmitty_grading.a into a fresh validate.out executable.

Once I clicked "Build Gradeable" on my end using a config identical to yours, I submitted 5.5 against an expected -5.5. The autograder successfully preserved the minus sign, detected the error, and correctly gave me 0 points

Let me know if rebuilding the gradeable from the UI works for you