[pacman-dev] sync_prepare clean-up III, TODO

[ngaba at bibl.u-szeged.hu]

> I'm not sure I see what's the performance problem with this type, but I'm
> curious, so could you elaborate a bit ? :)
> If you see a better way that would also clean up the code, it could be
> interesting.
Hi!
1. Almost all alpm functions need a list of pmpkg_ts; and the pmpkg_t <-> 
pmsyncpkg_t conversion is costly (see _alpm_sync_prepare: after topo sorting 
the list, you need to do an O(n^2) conversion); a kernel list-like 
implemetation would be more efficient (or merging pmsyncpkt_t to pmpkg_t)
2. The implementation of replaces is very ugly.
3. alpm_sync_free doesn't free (or what is FREE?!) the pkg member of spkg 
struct, however other parts of the code assume that it does it: huge memleak 
(?!)
4. in more general: this pmpkg_t of alpm_list implementation is not efficient, 
internally graph-like structures would be better, now you need to do at least O
(n) computation to find the "edges" (and we should do this only once) etc.
Bye, ngaba






----------------------------------------------------
SZTE Egyetemi Könyvtár - http://www.bibl.u-szeged.hu
This mail sent through IMP: http://horde.org/imp/