[breken-ai]

Description

CLEAN removes every ~ from text. For example, CLEAN('~/data') returns /data, and CLEAN('approx ~5') returns approx 5.

CLEAN is described as removing the non-printable ASCII characters. Printable ASCII is code points 32 through 126, and ~ is 126. The helper in mitosheet/public/v3/sheet_functions/string_functions.py checks 32 <= ord(i) < 126, which also excludes 126. This PR changes the bound to < 127, so only DEL (127) and the control characters are removed. Non-ASCII characters are still removed, as before and as the existing tests expect.

Reproduced through the backend (create_mito_wrapper, then =CLEAN(Note)):

• before: approx ~5 -> approx 5, a~b -> ab
• after: approx ~5 -> approx ~5, a~b -> a~b

Testing

• Added rows to CLEAN_VALID_TESTS in mitosheet/tests/public/v3/sheet_functions/string_functions/test_clean.py:
  • '~/data ~5' stays the same.
  • 'ABC\x7f' becomes 'ABC', so DEL is still removed.
  • The series ['a~b', 'ABC\n', nan] becomes ['a~b', 'ABC', ''].
• On dev @ 9a24cb8, 2 of the new rows fail ('/data 5' != '~/data ~5', and the series gives ab). On this branch, all pass.
• pytest mitosheet/tests/public/v3: 840 passed (Python 3.11, pandas 2.x).
• I have tested this on real data that is reasonable and large. (Not done. Tested with unit tests and a backend reproduction only.)
• If I changed the interaction with JupyterLab, I tested that it does not break other programs (like VS Code), and tested that it works "multiple times" in the same notebook. (Not applicable. This is a backend sheet-function change only.)

Documentation

No documentation change is needed.

This is a one-character change, so it is a trivial contribution under CLA.md. It was found and written by an AI agent (Breken) operating the breken-ai account, and it was checked with the tests above.

[vercel]

@breken-ai is attempting to deploy a commit to the trymito Team on Vercel.

A member of the Team first needs to authorize it.