Make sure column names are valid identifiers

[johnmyleswhite]

We should make sure that readtable guarantees that all column names are valid identifiers. If we're not doing this now (I'm not sure), we should.

[garborg]

Warn when coercing invalid names pulled from file and when invalid names are passed to colnames?

[johnmyleswhite]

I'm cool with warning, but we should be willing to budge on that if enough people complain.

I'd say we should raise errors when setting column names using names!.

[vchuravy]

Currently(0.5.4) the following problem arises:

Given a csv file with the following header:
Generation, Group, BlueTest

The column names are read as
:Generation
: Group
: BlueTest

readTable should either strip white space or allow for multiple-char separator..
I use almost always ", " as a separator for better readability. The data is handled correctly but the the column names result in non-valid identifieres

[garborg]

Thanks for reporting, I'll clean this up.

@johnmyleswhite I'll adapt your makeidentifiers code unless you've done anything else along those lines.

[johnmyleswhite]

Please go ahead. I also noticed a bug with column names the other day, but didn't track it down.

At some point, we need to sync with Base about makeidentifiers to deal with the Unicode normalization work they've been doing.

[garborg]

Good call.