On Feb 13, 2004, at 8:48 PM, Remi Mommsen wrote:


On Feb 13, 2004, at 9:15 AM, jfm wrote:
Similarly, to get the set of pkgs that eg still depend on gd a simple
egrep -rI '[, ]gd([, ]|$)' /sw/fink/dists/unstable|grep 'Depends:'
suffices with the 1 line convention.
(And as long as only a dozen pkgs don't follow that convention, it is still tolerable)

... Following your line of thought you will have to forbid the use of the here-doc structure for dependencies. Otherwise your simple grep reports an incomplete set of packages which is bad IMHO.

Sure I would prefer. But since I +/- know this dozen of pkgs, and since up to now I never had to
embed such uses of grep in longer scripts (running over all info files, eg to write a dependency tree
or something to be fed to tsort) which would have to be foolproof, the situation is still liveable for me.

Is it really worthwhile to complicate so much such simple uses ???

I guess a functionality for printing all packages depending on a given package could be quite easily implemented in fink itself.
And a 'f_wh' ("fink which"), to get eg the pkgs a set of given pkgs depend on
_as described in the first example in my previous msg...
And ...
Probably. But :
1) note that feature requests like #496468 and #535059 are since 2 years in the tracker ...
So I would rather wait till such things effectively exist before complicating life for substitutes.
2) you would want such such an implementation to be at least as efficient as the above-cited
'poor man" implementations that don't use the db. I mentioned this difference in the previously
cited msg; eg :


# time { fink info evolution|grep Maintainer ; }
 Maintainer: The Gnome Core Team <[EMAIL PROTECTED]>

real    0m51.066s
user    0m8.990s
sys     0m2.370s
# time grep Maintainer `f_wh evolution`
Maintainer: The Gnome Core Team <[EMAIL PROTECTED]>

real    0m23.870s
user    0m1.070s
sys     0m2.850s

(timings are quite consistent for the second, and much more variable for the first_
but differences always are in this direction, and the above 2 are rather typical, and
were obtained running simultaneously in 2 different windows).


(And to be fair, the comparison should be with such scripts for a hypothetical
fink which would really be user-friendly in this respect _ I mean, eg have
systematically info filenames of the form %n_%v_%r.info (like debs), with no other
underscores _ so "f_wh" would just have to read a couple of dirs, no files .
Similarly all variant syntaxes allowed in info files complicate such scripts _
and the users life ..)


3) What you are suggesting amounts to construct within fink a specialized
language for querying its db. It is not always easy to design well such a
language _ to be sure to cover all reasonable desiderata with a set of
primitives as small and efficient as possible.
Is it not better just to make sure the general purpose tools are easy to use ?


Best,

Jean-Francois



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Fink-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/fink-devel

Reply via email to