Dne pátek 27 duben 2007 13:30 Klaus Kaempf napsal(a):

* Jiri Srain [Apr 27. 2007 13:26]:

...

I'm not sure I understand the question properly, but is the possiblity that the packages are in different files in different order the answer for you?

Another quetsion: How will libxml behave if you parse three files simulatenously? (primary.xml, other.xml, filelist.xml)

Huh ? Why should one do that ?

See below.

And why is order important ?

Parse primary first, read one package, write one package. Then other.

Find each package in the database, modify it. I thought that was what Jano wanted to avoid. Jiri

Then filelist.

If there is a package in primary with no data in other or filelist, it will have incomplete attributes. This shouldn't harm.

If there is a package in either other or filelist, but without representation in primary, log a warning and continue.

Klaus

-- Regards, Jiri Srain YaST Team Leader --------------------------------------------------------------------- SUSE LINUX, s.r.o. e-mail: jsrain@suse.cz Lihovarska 1060/12 tel: +420 284 028 959 190 00 Praha 9 fax: +420 284 028 951 Czech Republic http://www.suse.cz

Duncan Mac-Vicar Prett wrote:

On Friday 27 April 2007 13:41:46 Jiri Srain wrote:

...

Find each package in the database, modify it. I thought that was what Jano wanted to avoid.

I meant to avoid the memory consumption that would be necessary otherwise. filelists.xml.gz for factory has 13 megs currently. If done like read package into mem, insert into db, delete from mem, memory consumption is zero...

When you insert a package, you get the id. You can get a map with the ids.

... if done this way, the memory overhead would be only this map with the ids. Still much better than before.

Those optmizations could also be introduced in the cache store.

I would go for this, if it would mean good memory vs. speed-up trade-off. jano -- To unsubscribe, e-mail: zypp-devel+unsubscribe@opensuse.org For additional commands, e-mail: zypp-devel+help@opensuse.org

* Jiri Srain [Apr 27. 2007 13:42]:

...

And why is order important ?

Parse primary first, read one package, write one package. Then other.

Find each package in the database, modify it. I thought that was what Jano wanted to avoid.

sqlite is reasonably fast and does quite good caching. It might be helpful to have different tables for (the attributes contained in) primary, other, and filelists. So you don't have to modify a table entry, but just to query the 'master' table for the package id and do 'stream' writing to the other tables. With an index, this should be cheap. Klaus -- To unsubscribe, e-mail: zypp-devel+unsubscribe@opensuse.org For additional commands, e-mail: zypp-devel+help@opensuse.org