BUG: Read_Table and Read_Csv does not raise when delim_whitespace=True and sep=default is given

[phofl]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

import io
f = io.StringIO("a  b  c\n1 -2 -3\n4  5   6")
df = pd.read_table(f, sep="\t", delim_whitespace=True)
df = pd.read_csv(f, sep=",", delim_whitespace=True)

Problem description

Both statement should raise a ValueError

Expected Output

ValueError

Must also edit docs. See #36560 for discussion.

cc @simonjayhawkins

Output of pd.show_versions()

master

[OlehKSS]

Hi, does anyone work on this issue? I want to contribute

[phofl]

Hi, not as far as I now. Please check the discussion in the PR mentioned above and the related issues before starting

[OlehKSS]

Thanks for the quick reply, I'll take it

[OlehKSS]

As far as I see, this bug is due to the fact that in both cases the explicitly provided values of separators are equal/identical to their default values. When pandas compares whether a separator/delimiter value was given, it checks it against these default values. No ValueError was raised in both statements because no default value was changed as if no separator was provided explicitly.

One possible solution might be setting all default values to None and assigning their proper values later. In my opinion, it is possibly going to break automatical detection of separators, i.e. if sep is None, the Python parsing engine (not C) will be used to automatically detect the separator.

[phofl]

None won‘t work. Please read the discussions in #36381 and #36560. The Solution is already discussed there

[OlehKSS]

Based on those discussions I see that you use lib.no_default as a sentinel object. I'll apply it here