I would do it in a regular expression. Take a look at the "Positional" and "Lookaround" operators:
You should be able to get the first columns and second columns this way without a problem.
You can do this in whatever language you want, as long as it supports regular expressions -- M, Perl, shell (sed/awk), etc. The actual syntax will vary depending on which on you pick.
You could also do this without regular expressions, counting character places, but you might need to do some transformations of the data type.
The "best" way is the one that works with whatever you are doing with the data next. Ideally, the output will be in a format that is readily useful to the next processing step.
