multios stripping colors?

multios stripping colors?

[Ray Andrews]

All:

     $   eval "$@"

That line in a function of mine can successfully eat some very complex 
input including various pipes and color codes.  However, when I try this:

$ eval "$@" >&2 >! /tmp/_output

Although I do get both outputs, the screen looses all colorized output 
that was previously there.  As I understand the docs, this might be 
expected at lest in so far as it seems to say that both outputs must be 
created at the same time, thus if there is some reason that color codes 
can't be sent to the file, then they won't be sent to the screen either. 
Is there anything I can do to get my color back?  On the screen at 
least.  I can send color escapes to a file and thecolors show upwhen the 
file is cat-ed, but I'm not sure that the colors as above can be sent to 
a file.  Can they? Seems to me I remember that they can't.

Re: multios stripping colors?

[Oliver Kiddle]

Ray Andrews wrote:
>
>      $   eval "$@"
>
> That line in a function of mine can successfully eat some very complex 
> input including various pipes and color codes.  However, when I try this:

eval "eats" only shell syntax. Colour codes are eaten by your
terminal. Perhaps you contrived to prevent them being generated or
constructed something to filter them out.

> $ eval "$@" >&2 >! /tmp/_output
>
> Although I do get both outputs, the screen looses all colorized output 

There is nothing intrinsic to your construct that loses colour escape
sequences. Try the following:

  eval "print -P '%K{blue}Hello'" >&2 >! /tmp/_output      
  cat /tmp/_output 

This should give you a blue background from both commands.

So you might need to take a closer look at exactly what argument you are
passing to eval.

Some commands, make coloured output conditional on whether their
standard output points to a terminal. For an example, try:
  git log | cat

Oliver

Re: multios stripping colors?

[Ray Andrews]

On 31/01/18 11:30 PM, Oliver Kiddle wrote:
> Ray Andrews wrote:
>>       $   eval "$@"
>>
>> That line in a function of mine can successfully eat some very complex
>> input including various pipes and color codes.  However, when I try this:
> eval "eats" only shell syntax. Colour codes are eaten by your
> terminal. Perhaps you contrived to prevent them being generated or
> constructed something to filter them out.
Of course.  Here's a current example:

             _execute aptitude search \'$iinst ?description($1)\' '|' 
egrep --color \'^\|$1\ $ggrep_oneline\'

'_execute' is a function that contains the above 'eval' plus a few other 
things.  As I run it with the line just above, I get 'aptitude' output 
colored via egrep ....
>
>> $ eval "$@" >&2 >! /tmp/_output
>>
>> Although I do get both outputs, the screen looses all colorized output
... but again, if I do it this way the output to the screen goes B&W.
> There is nothing intrinsic to your construct that loses colour escape
> sequences. Try the following:
>
>    eval "print -P '%K{blue}Hello'" >&2 >! /tmp/_output
>    cat /tmp/_output
>
> This should give you a blue background from both commands.
Right, so it does.
>
> So you might need to take a closer look at exactly what argument you are
> passing to eval.
>
> Some commands, make coloured output conditional on whether their
> standard output points to a terminal. For an example, try:
>    git log | cat

Maybe that's it.  So is 'aptitude' or 'egrep' itself determining that 
color is not to be had due to the double output?  Sounds likely, I know 
that the greps all behave differently when receiving piped input than 
when not.  And I'm correct that if one output is modified the other will 
be too?

Thanks, I'll play with it a bit more.
=======================
>
> Oliver
>

Re: multios stripping colors?

[Ray Andrews]

On 31/01/18 11:30 PM, Oliver Kiddle wrote:
>
> Some commands, make coloured output conditional on whether their
> standard output points to a terminal. For an example, try:
>    git log | cat
>
> Oliver
>
By gum, that's it.

     egrep --color=always

... solves the problem.  As you say, egrep was trying to be clever 
there, detecting the output to the file and killing the color, but you 
can force the issue with 'color=always'.  Thanks Oliver, there is always 
a way.