Re: multiple inputs / outputs through descriptors

[textutils1]

 Hi Bob,

   thanks for your response. Comments below ...

>   diff <(process1) <(process2)

  That works great for most purposes ! Thanks for reminding 
of this syntax.

  But how would I make the open FDs available to a 
sub-shell? e.g. how would I say :

  ( read 63>&1 ; ... ; read 62>&1 ; ... ) <(...) <(...)

 e.g. this gives me syntax error :
  ( read 63>&1 ; read 62>&1 ) <(ls) <(ls -L)

   Thx,

     John


aside: Do you know whether there is a way to *switch* two
descriptors (like stdout / stderr) ? I.e. the intuitive
'( ls -d / /asdfasdf 1>&2 2>&1 ) >err 2>out' doesn't work.


On Sat, 30 Aug 2003, Bob Proulx wrote:

> Date: Sat, 30 Aug 2003 21:14:54 -0600
> From: Bob Proulx <address@hidden>
> To: address@hidden
> Cc: address@hidden
> Subject: Re: multiple inputs / outputs through descriptors
> 
> address@hidden wrote:
> >   Since I didn't find textutils / coreutils discussion list
> > (only coreutils-announce), and this is more of a technical
> > implementation / feature issue, I'm writing this into
> > bug-textutils. Please forward / repost where appropriate or
> > let me know.
> 
> You have sent your note go a reasonable place.  But for the future
> here is the documentation you were looking for.
> 
>   http://www.gnu.org/software/coreutils/
> 
> >   Standardize throughout GNU utils (just like long options 
> > have been standardized) a descriptor input / output syntax. 
> 
> In the old days '-' was considered the standard for stdin and stdout.
> Look at 'tar' for one example.  (e.g. 'gunzip file.tar.gz | tar tvf -')
> But that is "in-band control" and creates the possibility of name
> conflicts.
> 
> > e.g. I could write (in bash) something like this :
> > 
> >  ( process1 --output 3>&1 ; process2 --output 4>&1 ) | diff --input-file 
> > %3% --input-file %4%
> 
> Yes, that is ugly.  How about using bash's process substitution?
> 
>    Process Substitution
>        Process  substitution is supported on systems that support
>        named pipes (FIFOs) or the /dev/fd method of  naming  open
>        files.  It takes the form of <(list) or >(list).  The pro-
>        cess list is run with its input or output connected  to  a
>        FIFO  or  some  file in /dev/fd.  The name of this file is
>        passed as an argument to the current command as the result
>        of the expansion.  If the >(list) form is used, writing to
>        the file will provide input for list.  If the <(list) form
>        is  used, the file passed as an argument should be read to
>        obtain the output of list.
> 
> Then your above example looks like this.  This looks much cleaner to
> me.
> 
>   diff <(process1) <(process2)
> 
> Here is a real example you can try.
> 
>   diff -u <(printf "foo\nbar\n") <(printf "baz\nbar\n")
> 
>   --- /tmp/sh-np-a20909   Sat Aug 30 20:59:55 2003
>   +++ /tmp/sh-np-b20909   Sat Aug 30 20:59:55 2003
>   @@ -1,2 +1,2 @@
>   -foo
>   +baz
>    bar
> 
> This works even on my ancient HP-UX 10.20 machine upon which I tested
> this example.  As long as the system supports named pipes this is
> available.  Newer systems with /dev/fd are even better.  But I don't
> believe this is specified by POSIX and so should be considered a bash
> only feature.  But I say that not knowing if other shells support it
> or not.  Just that it is not a POSIX /bin/sh feature.
> 
> Bob
> 

-- 
-- Gospel of Jesus' kingdom = saving power of God for all who believe --
                 ## To some, nothing is impossible. ##