Bug 63713

Summary: FILEOPEN DB metadata: load lazily / in background / persistent cache file / limit by table filter
Product: LibreOffice Reporter: josef
Component: DatabaseAssignee: Not Assigned <libreoffice-bugs>
Status: NEW --- QA Contact:
Severity: enhancement    
Priority: low CC: iplaw67, jmadero.dev, lionel, lo_bugs, serval2412
Version: 4.0.2.2 release   
Hardware: Other   
OS: All   
See Also: https://bugs.freedesktop.org/show_bug.cgi?id=57872
https://bugs.freedesktop.org/show_bug.cgi?id=52027
Whiteboard: NeedsDevEval SkillSQL SkillCpp
i915 platform: i915 features:

Description josef 2013-04-19 09:10:55 UTC
Connecting to our Oracle DB takes quite long for
e.g. the first click on "Tables" or opening a query or form.

maybe this has to do with too many DB Objects.

Is it possible to reduce this loading time?

After first time connection to the DB there is no speed problem any more.
Comment 1 josef 2013-04-19 09:14:48 UTC
I've tested this behaviour under Windows with JRE 1.6 and JRE 1.7
and Linux with OpenJDK 1.6 and 1.7
Comment 2 josef 2013-04-19 09:23:05 UTC
used JDBC-Drivers:

ojdbc6.jar
ojdbc14.jar

it seems,
Libreoffice reads all metadata for all found DB-Objects.
Maybe there could be a feature that limits this function only to those objects that are listet in tables filter.
Comment 3 Julien Nabet 2013-05-04 21:05:42 UTC
Which LO version do you use? FYI, last one is 4.0.2 and 4.0.3 is gonna be released in a few days (see https://wiki.documentfoundation.org/ReleasePlan#4.0_release)
Comment 4 josef 2013-05-05 09:38:46 UTC
(In reply to comment #3)
> Which LO version do you use? FYI, last one is 4.0.2 and 4.0.3 is gonna be
> released in a few days (see
> https://wiki.documentfoundation.org/ReleasePlan#4.0_release)

Actually this has been since I've been using LibreOffice Base with Oracle DB.
Now I'm using version 4.0.2.
As soon as possible I'll change to version 4.0.3 but the change log doesn't describe any feature change.
Comment 5 Julien Nabet 2013-05-05 14:01:13 UTC
josef: thank you for your feedback. I'll put 4.0.2 for the moment for Version field.
About 4.0.3, it was just for information, I don't know if something related to your problem has been modified/fixed in this version.
Comment 6 josef 2013-05-06 05:54:59 UTC
I think, this would be a major feature change.
I have so many DB-objects, that this kind of modification would be a glance.
Comment 7 Julien Nabet 2013-05-08 07:34:54 UTC
Josef: the fdo#57872 put in "See also" may interest you.

Lionel: should we consider it as a dup?
Comment 8 Lionel Elie Mamane 2013-05-08 07:44:42 UTC
(In reply to comment #7)
> Lionel: should we consider it as a dup?

No. bug 57872 is about managing the "C++ to Java call overhead", while this bug is about less eager (more lazy) caching of metadata.
Comment 9 Julien Nabet 2013-05-08 07:49:17 UTC
Lionel: Ok I thought (wrongly now I've read your feedback) that metadata should always be completely loaded and so that it was only due to the slowliness pb.
Comment 10 Lionel Elie Mamane 2013-05-10 14:10:52 UTC
Alternatively, we could cache the metadata information in the file, and refresh it only on explicit request by the user.

That's what MS Access does (with "linked tables"). I find that rather suck and not that user-friendly, actually. So I think I'd prefer the lazy loading. Or maybe background loading?


Summary of ways improving this:

1) Load metadata lazily

2) Load metadata in background

3) Cache metadata (negative vote from me)

4) Load only metadata of tables that match the tables filter.


4) can be combined with any of 1/2/3, and should be the easiest. Setting "PropôsedEasyHack" for this point "4".
Comment 11 Terrence Enger 2013-05-15 18:14:36 UTC
The behaviour that I complained about in bug 52027 sounds similar.  I observed it with an ODBC connection but L.E.M. commented that the problem is common to all backends.

By 2012-10-07 the problem was no longer evident.  I wonder if the current bug is something different or if my older problem has returned.

( I am almost of the net, so will be slow to contribute further.  Sigh. )

Terry.
Comment 12 Terrence Enger 2013-05-16 13:15:41 UTC
(In reply to comment #11)
The problem of bug 52027 is still absent in master commit 6faa622
pulled 2013-04-19.  I conclude the the present bug is something
different, and my earlier comment is mere noise.
Comment 13 Joel Madero 2014-02-27 23:28:46 UTC
In order to limit the confusion between ProposedEasyHack and EasyHack and to make queries much easier we are changing ProposedEasyHack to NeedsDevEval.

Thank you and apologies for the noise
Comment 14 Alex Thurgood 2015-01-03 17:40:45 UTC
Adding self to CC if not already on

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.