On Mar 14, 8:15 am, "tbh" <fem...@[EMAIL PROTECTED]
> wrote:
> "Janis" wrote in message
>
> news:18784f90-e0de-450c-8c93-a91d5dee36d4@[EMAIL PROTECTED]
>
> > > i guess it's instructive to me that this appears non-trivial (and
even a
> > > bit
> > > controversial?) in a language as mature as awk.
>
> > No, it is trivial. Any "non-triviality" here stems from shell
> > quoting issues and from mixing shell level and awk level code.
> > I also think it hasn't been controversial (at least comparing
> > suggestions amongst posters of a similar experience).
>
> > (Some non-trivialities will appear if your dynamic regexps
> > contains regexp meta-characters.)
>
> thanks, that was perhaps careless of me. maybe we could agree on
"trivial"
> in theory, sometimes non-trivial (or non-portable? or potentially
> troublesome in some cases?) in practice.
>
> speaking of which, i now got this to work in cygwin and under linux
>
> echo "/a/|/b/" | tr '|' '\n' | awk -v base="a" '($1 ~ "/"base"/")
> {print}'
>
> so I'm a happy camper.
>
> > (BTW, since it hasn't been mentioned yet; to avoid problems
> > with quoting you can put the awk program in a file and call
> > it using awk's -f option. As a side effect you are forced to
> > use a cleaner and more robust way to pass shell variables.)
>
> good to know, thanks. i have occasionally used awk that way in the past,
but
> get a lot of great mileage with one-liners in the middle of pipes. the
> challenge becomes how to file away the useful ones for future recycling.
:)
>
> cheers, Tim
If I need to get pattern-matching info from the shell, I generally
write in a C-like way, use "match" and "[g]sub" to do the work, and
have my patterns as strings in quotes instead of patterns in slashes.
But then, I also tend to write long (1000+ lines) gawk programs using
both getline (oh the horror!) and patterns from standard input.
To show how far behind the times I am (using awk for over 15 years
now), I have started learning perl. One of these days, I may move from
fortran 77 to fortran 20xx.
|
