Always save first N bytes of output from all commands?

Always save first N bytes of output from all commands?

[Francisco Borges]

[-- Attachment #1: Type: text/plain, Size: 1023 bytes --]

Hi,

Is there an easy easy way to *always* save the (first N bytes) of the
output all commands? Or least save the output of the very last command?

Something like always having

 | tee ~/.output_history/$TTY/${HISTCMD}.out

but ideally with a control over the number of bytes written. Is something
like that possible?

Right now I resort to insert-last-command-output function (original by pws
at http://www.zsh.org/mla/users/2005/msg00550.html while answering a
question of mine...). To re-execute a command and insert its output in the
command line. Sometimes it is fast enough and sometimes it is slow. I know
we have "keep" (./share/zsh/functions/keeper) distributed with Zsh itself,
but keep much like tee has to be manually inserted into the command itself
and that it becomes less awesome.

I work on OSX and I bet I can write AppleScript code to... somehow fetch
the output and insert in the buffer but it I'd rather a (portable)
Zsh-based solution.

Any ideas other than going the AppleScript route?
-- 
Francisco

Re: Always save first N bytes of output from all commands?

[Daniel Shahaf]

Francisco Borges wrote on Tue, 08 Jan 2019 15:53 +0100:
> Is there an easy easy way to *always* save the (first N bytes) of the
> output all commands? Or least save the output of the very last command?
> 
> Something like always having
> 
>  | tee ~/.output_history/$TTY/${HISTCMD}.out
> 
> but ideally with a control over the number of bytes written. Is something
> like that possible?
> 
> Right now I resort to insert-last-command-output function (original by pws
> at http://www.zsh.org/mla/users/2005/msg00550.html while answering a
> question of mine...). To re-execute a command and insert its output in the
> command line. Sometimes it is fast enough and sometimes it is slow. I know
> we have "keep" (./share/zsh/functions/keeper) distributed with Zsh itself,
> but keep much like tee has to be manually inserted into the command itself
> and that it becomes less awesome.

You could wrap the accept-line widget with something along these lines:

accept-line() {
  PREBUFFER="{ $PREBUFFER"; BUFFER+=" }"
  BUFFER+="| tee >(head -c 1024 > /path/to/file)"
  zle .accept-word -- "$@"
}
zle -N accept-line

(untested)

However, commands that check if their stdout isatty() (the C equivalent
of «[[ -t 1 ]]») will with this change think it's not a tty.  ls(1)
without arguments is the standard example of this.  You could probably
fix that by creating a pty, but I don't have an example of that.

> I work on OSX and I bet I can write AppleScript code to... somehow fetch
> the output and insert in the buffer but it I'd rather a (portable)
> Zsh-based solution.
> 
> Any ideas other than going the AppleScript route?

Cheers,

Daniel