This is perhaps conceptually best as two filters; -group and -sort. OTOH there may be performance reasons to have -sort --group=RE instead.
I think the ideal output would still subunit, rather than (say) just the ids, so that one can immediately look at the traces, or pipe it through -ls to get the ids. OTOH this suggests potentially large buffering.
Some ways to address that:
- spool to N temp files (one per group)
- output a single case of a group and just enumerate the others
- don't worry about it for now
This is perhaps conceptually best as two filters; -group and -sort. OTOH there may be performance reasons to have -sort --group=RE instead.
I think the ideal output would still subunit, rather than (say) just the ids, so that one can immediately look at the traces, or pipe it through -ls to get the ids. OTOH this suggests potentially large buffering.
Some ways to address that:
- spool to N temp files (one per group)
- output a single case of a group and just enumerate the others
- don't worry about it for now