Zsh noob: word-splitting headache

Zsh noob: word-splitting headache

[kynn]

I'm writing this script in which, among other things, I want to do
something for each line output by some command foo.  The problem is
that the lines output by foo contain whitespace (e.g. "bar baz
frobozz"), so a construct like this

  for i in `foo`
    do something with $i
  done

fails, because i holds individual words instead of whole lines.

I've tried every trick I can think of to prevent the intra-line
word splitting while still allowing the inter-line splitting, but
nothng has worked.  What's the right way to do this?

I looked at the FAQ at http://zsh.sunsite.dk/FAQ for an answer to this
question (as well as the Zsh Manual, the Zsh User Guide, and "From
Bash to Z-Shell", but no luck.

BTW, of all the difficulties I have programming zsh (or any shell for
that matter), a great many of them are in one way or another related
to word splitting.  Is there a *programming* FAQ on zsh word
splitting?

Thanks!

kj

Re: Zsh noob: word-splitting headache

[Clint Adams]

>   for i in `foo`
>     do something with $i
>   done

  for i in ${(f)"$(foo)"}
    do something with $i
  done

Re: Zsh noob: word-splitting headache

[Bart Schaefer]

On Tue, 4 Jan 2005 kynn@panix.com wrote:

>   for i in `foo`
>     do something with $i
>   done
> 
> fails, because i holds individual words instead of whole lines.

Use a different construct.

   foo | while read i; do something with $i; done

This works nicely in zsh because zsh executes the right-hand-side of 
pipelines in the current shell when it can.  It might not work as well in 
other shells, but other shells don't have Clint's suggested ${(f)...} 
expansion either.

Re: Zsh noob: word-splitting headache

[kynn]

   Date: Tue, 4 Jan 2005 16:10:25 -0500
   From: Clint Adams <clint@zsh.org>

   >   for i in `foo`
   >     do something with $i
   >   done

     for i in ${(f)"$(foo)"}
       do something with $i
     done

What a godsend!  Thanks!

kj

Re: Zsh noob: word-splitting headache

[Philippe Troin]

Clint Adams <clint@zsh.org> writes:

> >   for i in `foo`
> >     do something with $i
> >   done
> 
>   for i in ${(f)"$(foo)"}
>     do something with $i
>   done

I tend to prefer making foo print NUL-separated records and use:

  for i in ${(ps:\0:)"$(foo)"}
    do something with $i
  done

  works with such foos as "find . -print0"

Now, I can handle ANY filename, including the ones containing
newlines.

Phil.

Re: Zsh noob: word-splitting headache

[Stephane Chazelas]

On Tue, Jan 04, 2005 at 01:30:44PM -0800, Bart Schaefer wrote:
[...]
>    foo | while read i; do something with $i; done

foo | while IFS= read -r i; do something with "$i"; done

or leading and trailing blanks would be stripped, and backslash
would be treated specially.

Note that this means that "something" stdin is modified.

The quotes around "$i" are here in case there are empty lines.

Note that in

for i in ${(f)"$(foo)"}

or

IFS=$'\n'
for i in $(foo)

The empty lines are discarded.

To avoid that, you can do:

IFS=$'\n\n'
for i in $(foo)

But still, because of the (to my mind bogus) way command
substitution works in Bourne like shells, the empty lines at the
end of foo output will still be discarded. To prevent that, you
can do:

IFS=$'\n\n'
for i in ${$(foo; echo .)[1,-2]}

$ printf '<%s>\n' ${$(echo 'a b\n\n'; echo .)[1,-2]}
<a b>
<>
<>

-- 
Stéphane