[Bug-datamash] New feature: extended identifiers

[Stefan Vargyas]

Dear Assaf,

First of all, let me thank you for this useful tool!

This email is about one very useful feature of 'datamash': that of fields
specified by name.

Using 'datamash' programmatically in bash scripts, it is very handy that
I need not compute column indexes based on the column names that I know.
The scripts are generating tables of which set of columns is a function
of the input parameters. For example, the column 'total-node-mem' will
be the 5th one for one running instance of the script, while for other
-- the 6th.

Now comes in my issue with 'datamash': my dynamically generated tables
specify column names by the pattern '[a-z0-9-]+', as already suggested
above. Currently, passing in 'total-node-mem' as a field to 'datamash',
makes the program to exit with error (saying that field ranges must be
numeric).

Reading your code, I learned that extending the program's notion of
identifiers to a more liberal variety would not impede neither its
internal nor its external logic. For a simple proof of concept, see
the patch attached.

If you consider this notion of extended identifiers to be beneficial
for a wider range of users, I will gladly work with you on a complete
implementation.

Sincerely,

Stefan Vargyas.

extended-identifiers.patch
Description: Binary data