Bug: implicit conversion of strings to exponent. (tabulate converting types)

This is a bug is converting strings like this 20200407_2 to exponent values.

>>> print(tabulate.tabulate([ [93, '20200310.2'], [65, '20200407_2'] ]))
--  -----------
93  2.02003e+07
65  2.02004e+08
--  -----------

surprising that float will convert this one at all as it has an underscore in it.

>>> float('20200407_2')
202004072.0

Anyway tabulate should not be converting types at all.

[MikeBusuttil]

@pete312 I was able to fix the issue by invoking the disable_numparse argument: print(tabulate.tabulate([ [93, '20200310.2'], [65, '20200407_2'] ], disable_numparse=True))

[pete312]

@astanin The end point of tabulate is text align and is display only. Auto convert int to str is fine but going back is not consistent in regression. Meaning tabulate will output different results python 3.6 than prior versions of python 3. because python 3.6 allows an int to be defined as 20200407_2 and not before 3.6. However “20200407_2” and 20200407_2 are not equal. Also the output of print(20200407_2) is not the same as string “20200407_2”.

>>> 20200407_2 == "20200407_2"
False
>>> print(20200407_2), print("20200407_2")
202004072
20200407_2
(None, None)

To me this means tabulate is incorrectly converting a previously string type to an int type. Output will differ after 3.6 with this default set to True. It will not break if set to False. Usually regression test failure are considered as bugs. “Undefined behavior” or “Inconsistant Feature”