[C++][Python] Allow utf8_slice_codeunits to support default start value of None to support strings of different length

[rohanjain101]

Describe the enhancement requested

Related to pandas-dev/pandas#52434. Currently utf8_slice_codeunits doesn't support a default start value. It would be nice to support default of 0 if step > 1 and len -1 if step < 0 for parity with pandas.

Component(s)

Python

[jorisvandenbossche]

@rohanjain101 thanks for the report! I was going to mention that you could use sys.maxsize as start (the largest integer, will always be beyond the end of a single string in the input, and so will always start slicing from the end), but apparently you can easily run into a segfault with that: opened #34928

As a non-ideal workaround, you could check what is the largest string in your input array with pa.compute.max(pa.compute.utf8_length(arr)) (or just take a reasonable large value, but not close to sys.maxsize), and use that as start value:

In [1]: pa.compute.utf8_slice_codeunits("abcdefghijklmnabcdefghijkln", start=1000, stop=8, step=-9)
Out[1]: <pyarrow.StringScalar: 'nd'>