(grr it happened again >.<, my apologies to the guy who received this mail, it
was intended for the list)
From: Petr Jakeš <petr.ja...@tpc.cz>
To: vicariou...@gmail.com
Date: Thursday 01 April 2010
> Some examples of the code, so we all can learn from your experiences, will
> be nice.
>
> Regards
>
> Petr
>
Well, for this particular case, I was using a factory class that was in charge
of creating the several objects that were required. It worked more or less
like so: there is an indexer class, which takes care of crawling the
filesystem. For every file or directory it encounters, it passes a series of
parameters to the factory class (or object ;)), which the factory uses to
instantiate the object that represents said file/dir. Yes, you may say that
has too much coupling, but given the handling that one must do regarding
connections and transactions (and taking care not to use them in a different
thread than the one they were created on), it seemed the easiest and more or
less "right" way to do it.
So, let's say that there's one method in this Factory class/object to
instantiate each of the objects that can be scanned from the filesystem
(regular file, directory, image/video/audio file). Instantiation in itself
isn't too expensive (or at least there isn't much that you can do about it).
But when it comes to extracting the metadata for some particular files
(namely, audio, video and image files) I found that I was setting fields on
the object one at a time. This caused several UPDATE statements, which slowed
everything down.
So, in order to alleviate this, I did the following:
def new_photo(self, parent, name, relpath, mimetype, atime, mtime,
root, size, strsize, strabs):
trans = self._conn.transaction()
atime, mtime = self._check_stat(atime, mtime)
mimetype = mimetype[0]
photo = Photo(parent = parent, name = name, relpath = relpath,
mimetype = mimetype, atime = atime, mtime = mtime,
size = size, strsize = strsize, root = root,
isdir = False, strabs = strabs,
connection = trans)
self._get_photo_metadata(photo)
trans.commit(close = True)
return photo
The photo (I didn't use "Image" as a name so I wouldn't overwrite some other
class already defined within Python) was the heaviest class in this regard,
since it extracted two scaled images from the original (that's just something
the pogram does, you can try it out if you want to). This extraction is done
in the _get_photo_metadata method. So I got a transaction from the regular
SQLObject connection, and used it as a parameter at object creation time,
instead of the connection. After that I commit it with the close parameter on
True, since then the method ends and I don't want that transaction lying
around so that after working I still have a journal file for the SQLite db :P.
Hope it is useful for somebody in the future. Don't forget that the
transaction object you get from the connection.transaction() method, is
already begun. You just have to work directly with it and DON'T FORGET to
close it :) (via commit or rollback).
Thanks for the help people.
Cheers
Juan Manuel Santos
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
sqlobject-discuss mailing list
sqlobject-discuss@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sqlobject-discuss
