FS#41525 - [pacsysclean] unnecessary slow

Attached to Project: Pacman
Opened by Bjoern Hiller (Zaphod) - Monday, 11 August 2014, 12:16 GMT
Last edited by Allan McRae (Allan) - Thursday, 27 November 2014, 01:14 GMT
Task Type General Gripe
Category Scripts & Tools
Status Closed
Assigned To No-one
Architecture All
Severity Very Low
Priority Normal
Reported Version git
Due in Version 4.2.0
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Summary and Info: pacsysclean is slow due to the use of bash for filtering.

A possible patch is attached which uses gawk for the whole processing. Explicitely stating gawk as sorted for loops are a GNU-extension. Also, the output format is slightly changed as all packages with the same size are printed on one line.

Timing goes down from 8.9s to 1.0s on a Raspberry Pi and 1.0s to 0.2s on a i7-860.
   pacsysclean_faster.patch (1.2 KiB)
This task depends upon

Closed by  Allan McRae (Allan)
Thursday, 27 November 2014, 01:14 GMT
Reason for closing:  Fixed
Additional comments about closing:  pacsysclean removed in commit bb878217cd
Comment by Dave Reisner (falconindy) - Monday, 11 August 2014, 12:49 GMT
> Explicitely stating gawk as sorted for loops are a GNU-extension.
This is problematic because we explicitly do not use gawk elsewhere in the codebase -- usage of awk is portable is nawk and mawk, as well.
Comment by Bjoern Hiller (Zaphod) - Wednesday, 13 August 2014, 12:57 GMT
A new patch is attached which should work with other awk-implementations (tested using awk -W posix) and has the same benefits as the old one (also same timings).
   pacsysclean_faster_posix.patch (1.1 KiB)
Comment by Allan McRae (Allan) - Wednesday, 13 August 2014, 13:48 GMT
pacsysclean is broken in current git because when using the -Qi output it does not look at the suffix of the Installed Size. This patch does not fix that.

Note there is a replacement script available that is fast and fixes this issue.
Comment by Allan McRae (Allan) - Wednesday, 19 November 2014, 07:03 GMT
Flagged for 4.2.0 - not because it will be fixed, but because pacsysclean is probably going to be removed.

Loading...