(enh) AgentSkills: fix \n handling with file editing (iPython)

[tobitege]

Short description of the problem this fixes or functionality that this introduces. This may be used for the CHANGELOG

Fixed handling of literal "\n" in file editing operations via iPython.

---

Give a summary of what the PR does, explaining any non-trivial design decisions

• in file_ops.py and CodeActAgent.py fixed handling of literal line breaks
• helps fix errors when LLM calls edit_file_by_replace but errors with [No exact match found in xxx]
  this contributed to an excessive amount of steps needed (and often failing the bench test instance)
• with this change the Aider bench instance 57 is finally successful as it doesn't run out of steps
• added several more unit tests

TODO:

• Aider bench (done)
• SWE bench

In below example, the to_replace value with this PR has the \n correctly treated as text, whereas in main it would be shown as an actual line break:

THOUGHT: Thank you for providing the content of the `accumulate.py` file. Now, let's implement the `accumulate` function according to the given instructions. We'll modify the file to include the correct implementation:
CODE:
edit_file_by_replace(
    'accumulate.py',
    to_replace="def accumulate(collection, operation):\n    pass",
    new_content="""def accumulate(collection, operation):
    return [operation(item) for item in collection]"""
)

---

Link of any specific issues this addresses

Fixes #3650

[tobitege]

I'm still looking into formatting issues with linebreaks with this.
Even though the fix is working to have more correct search hits in edit_file_by_replace in e.g. IPython code snippets, it doesn't correctly replace "\n" for display in logs.

[tobitege]

Preliminary Aider bench result (unit tests enabled, 30-steps limit):
Passed 124 tests, failed 9 tests, resolve rate = 93.23%

[xingyaoww]

Thanks for taking a stab at this! This is a very tricky issue that somewhat lead me to believe we need to implement an EditAction (#3777) 😢

[xingyaoww]

Shouldn't this Line 2 get replaced? because to_replace='Line 2\r\nLine 3\\r\\n',

[tobitege]

Actually not in this case as the search term is Line 2\r\nLine 3\\r\\n (note the \r)

[xingyaoww]

Yeah, since "Line 2\r\nLine 3\r\n" is the search term, shouldn't Line 2 completely goes way in the new fine, since it it replaced?

[tobitege]

@xingyaoww - this should be ready for you eval, if you like

[xingyaoww]

Thanks a lot -- this looks much cleaner! Just had a few questions before i can start kick off eval

[xingyaoww]

Yeah, since "Line 2\r\nLine 3\r\n" is the search term, shouldn't Line 2 completely goes way in the new fine, since it it replaced?

[xingyaoww]

Kicked off eval on the latest commit!

[tobitege]

Kicked off eval on the latest commit!

Thank you! 🚀