[pacman-dev] [PATCH 6/7] pacsearch: using pacman color theme

[Andrew Gregory]

On 01/23/14 at 02:58pm, Pierre Neidhardt wrote:
> On 14-01-23 08:38:22, Andrew Gregory wrote:
> > On 01/23/14 at 12:07am, Pierre Neidhardt wrote:
> > > No more per-repo coloring: this was not Arch-agnostic, and there is no
> > > reasonable, simple way to color repos in a consistant manner with only 6 colors.
> > > 
> > > 'local' is in red: this way we benefit from the pacman -Ss && pacman -Qs combo.
> > > 
> > > to_color subroutine: it takes an array instead of a string, this is faster and
> > > simpler.
> > > 
> > > Signed-off-by: Pierre Neidhardt <ambrevar at gmail.com>
> > > ---
> > >  contrib/pacsearch.in | 32 ++++++++++++++------------------
> > >  1 file changed, 14 insertions(+), 18 deletions(-)
> > 
> > If we use pacman's color theme and remove per-repo coloring does this script
> > still do anything worth keeping it around and fixing its bugs (try `pacsearch
> > pacman mirrorlist`)?  All that's left is searching both -Ss and -Qs, recoloring
> > "local/" red, adding "[installed]" to -Qs entries, and hiding -Qs entries that
> > are also in -Ss.  All but the last can be accomplished in a few lines of bash
> > using sed.
> 
> That was what I suggested in the first place a few days ago in my 'request for
> removal' mail. But Dan finds the combo of -Ss and -Qs quite useful and pointed
> quite good reasons to keep it.

I'm not disputing the usefulness of searching -Qs and -Ss at the same time, but
that alone is a one line bash function.  There's no point in keeping this
script if that's all we care about.

> This script is tiny, right, but not enough to be a one-liner (the "duplicate
> entry removal" needs a hash table and some parsing). Besides when you do it with
> sed/awk, you lose the coloring. And with sed/awk it's hard to sort without
> another call to 'sort', which does not save time. All these reasons are good
> enough to have this script.

pacsearch does not sort its results and you can still have color with sed/awk
with --color=always.

> If I understood you correctly, `pacsearch pacman mirrorlist` is not a
> bug. pacsearch's argument is a pattern, not package names. See pacsearch -h.

pacman mirrorlist is not the name of a package, those are two separate search
terms that pacsearch needlessly combines into a single term when it passes them
to pacman.

apg