There are a few issues with the implementation proposed by this branch:

• We seek a list of file paths relative to the current directory. However, this branch only adds file names.

• This branch updates filenames using the assignment operator = instead of .append(), meaning that the list of filenames is dropped with each iteration of the for loop.

• filenames is also being used by the for block itself. This means that even if the previous issue is fixed, every iteration of this for loop will still delete the value of filenames set by the previous iteration. Take this as a simplified example:

>>> for i in range(5):
...   print(i)
...   i = 1
...
0
1
2
3
4

This implementation can be corrected and simplified greatly. Here are my suggestions.

1. The logic within the for filename in filenames: ... block on line 69 should be extracted to a separate split_file(path, splitter) function.

2. Revert the other changes, and simply add this block at the very top of this split() function definition:

if os.path.isfile(path):
    return split_file(path, splitter)

I checked all this info by adding print() statements in the definition of split() to verify the value of filenames. To test, I ran jupyter lab from the root of this Git repo and called /learn docs to learn all of the Jupyter AI documentation.

Can you do the same before I review this again? Thanks in advance!

@dlqqq The code for the function split() can be simplified to the following form of the original function:

I have tested this separately and it works for a single file or a directory.