ERR: csv parsing error consistency of exceptions

[jreback]

xref #12551

a) change CParserError to an actual ValueError. this is used in a very small number of cases in the engine='c' parser. (in #12551 it will inherit from ValueError)

b) So we either change ALL ValueError to ParserError(ValueError) (e.g. a custom exception) and make the changes in both the engine='c' and engine='python'.

[jreback]

cc @gfyoung @jorisvandenbossche

thoughts?

[gfyoung]

xref #12506 as well

Hmmm...from a semantic point of view, I would prefer ParserError over the more generic ValueError. However, it does seem like converting everything to ParserError is going to break a lot more things compared to ValueError I'd imagine.

[gfyoung]

@jreback : Does the merged #12506 handle this issue satisfactorily?

[jreback]

yes, ty.

closed by #12506

[jreback]

actually. What I eventually wanted was to make ALL parser exceptions the same, which is an API change.

[gfyoung]

By which you mean across both the CParser and the PythonParser or that they just always raise one exception whenever there is an error in the reading?

[jreback]

yes I think we need to have a single ParserException (which is just a ValueError). (and remove the CParserException moniker, that's why its an API change.

[gfyoung]

Okay, that makes sense. Just wanted to clarify!