XMP/CC in Tracker, first patch

Hey Jamie,

I'm done with school now and have lots of time to devote to
this.

For quick questions, is IRC a good place to ask?  Is there a
particular
time that you're more likely to be on?

Attached is work I've done so far, diff'ed against trunk. 
It looks for
.filename.xmp in the current directory, and if it finds it
runs the
extractor on it.  At the moment, only the cc:license field
is recognized.

It also updates the pdf extractor to read xmp from the pdf's
metadata
field.  The current Poppler glib bindings don't allow
reading the
metadata field, so that had to be patched to make it work. 
I've emailed
the Poppler mailing list about this.

Configure.ac is updated to check for the exempi library
(http://www.f
iguiere.net/hub/blog/?Exempi) and conditionally compile
in
support for xmp.

Overall, I've spend the last few weeks poking around in the
code and
trying to get familiar with how Tracker works.  I imagine
I'll focus
down on one thing at a time as I get comfortable with the
overall picture.

How does the patch look?  Style?  Approach? etc...  Also,
how do I go
about getting commit access?

Cheers,
Jason
> On Tue, 2007-06-05 at 17:09 -0700, Jason Kivlighn
wrote:
>   
>> Hey Jamie,
>>
>> I'm in the middle of finals week, so there hasn't
been much hacking yet,
>> but I've got two questions.  Where is my code going
to be developed?  A
>> branch or trunk or...?
>>     
>
> trunk
>
>   
>> Also, I'm wondering how to go about automatically
reading sidecar
>> files.  You mentioned putting the code in
tracker-extract.  It's
>> straightforward to pass along any XMP to that code
for any format that
>> can include it.  However, I'm not sure what the
best method of reading
>> sidecar files is.  Do I spawn tracker-extract to do
the work, making a
>> new mimetype for XMP?  At the moment I'm looking
at
>> tracker_db_index_file() calling
tracker_metadata_get_embedded() on the
>> sidecar file if it exists.
>>     
>
>
> a mime type of xmp would make sense I think. I guess we
need to do the
> following:
>
> For each file being indexed:
>
> 1) check if a corresponding xmp file exists (I assume
they will start
> with "." and be hidden and end with
".xmp")
>
> 2) call tracker-extract if it exists and merge with any
other metadata
> tracker tracker extracts for that file
>
> I guess adding the code to
tracker_metadata_get_embedded would seem
> right
>
> hope that answers your question?
>
> jamie.
>
>
>
>   


_______________________________________________
cc-devel mailing list
cc-devellists.ibiblio.org
ht
tp://lists.ibiblio.org/mailman/listinfo/cc-devel

Jason Kivlighn wrote:

> Attached is work I've done so far, diff'ed against
trunk.  It looks for
> .filename.xmp in the current directory, and if it finds
it runs the
> extractor on it.  At the moment, only the cc:license
field is recognized.

Why .filename.xmp ? It should be basename.xmp (no dot
prepended)
I don't know where this convention comes from but definitely
not the one
from the current (non-free) implementations.
I'd rather wish we don't reinvent the wheel.

> How does the patch look?  Style?  Approach? etc... 
Also, how do I go
> about getting commit access?

xmp_init() cost a lot to call. Same as xmp_terminate(). You
should only
call then once. Actually xmp_terminate() isn't mandatory as
when the
programm quits, things will be cleaned up.

And the CC namespace of XMP is defined in the library. It is
NS_CC. This
is cosmetic, but ensure that we use the right string.


Hub
_______________________________________________
cc-devel mailing list
cc-devellists.ibiblio.org
ht
tp://lists.ibiblio.org/mailman/listinfo/cc-devel