in read_csv, warn_bad_lines and error_bad_lines have misleading names

[rcompton]

read_csv will still fail if *_bad_lines=False. *_bad_lines only applies when the number of delimiters is wrong. This is misleading, a line with a poorly formatted date is what I would think of as a "bad_line" that should be ignored (cf https://stackoverflow.com/questions/22026181 )

[jreback]

this is not the purpose of error_bad_lines

if you are providing a date parser then you need to handle errors
(in this case you you should either put a try except in your parsing function,parse the dates after with to_datetime, or use infer_datetime_format (in 0.13.1))

infering dtypes is inherently an ambiguous and not foolproof process
the user needs to be in charge of this and be responsible

I'll mark this an enhancement for someday

[jorisvandenbossche]

Not sure there is an enhancement here (but feel free to formulate a more concrete feature request). Renaming the *_bad_lines is best done in a reworked API to better deal with such "bad lines", see eg #9549 (comment). Therefore closing this issue.