combine columns stringwise

[bjpop]

It can be useful to combine columns as (if they were) strings.

embark_town,class,size
Cherbourg,First,85
Cherbourg,Second,17
Cherbourg,Third,66
Queenstown,First,2
Queenstown,Second,3
Queenstown,Third,72
Southampton,First,127
Southampton,Second,164
Southampton,Third,353

For example it would be useful to be able to combine embark_town and class into a single column by concatenating their values together as strings.

This ought to be possible with eval, but currently eval doesn't seem to work with strings when using +.

There is an open ticket for this: pandas-dev/pandas#47734

We could wait for that bug to be fixed, or we could consider having a combine operation.

Sometimes it would be useful to be able to combine columns and throw away the originals, other times it might be useful to keep the originals as well.