On Wed, Aug 29, 2012 at 6:02 AM, G. Allegri <[email protected]> wrote: > Hi Barry, > " you can't write data providers in Python". This is the point: I wonder if > this is in absolute or it is in this moment. Given Qgis can load python and > C++ plugins, I wondered if a similar thing could be achieved for data > providers. I can imagine two ways: > > - extend the data provider registry to load python modules and then extend > the data provider class to make it call the data provider interface methods > both on Python and C++ providers (doesn't the plugins loader/registry do > something similar?) > - write a generic C++ data provider that acts as a proxy to the actual > python data provider. > > I don't know how hard it would be, but I can't find any technical barrier... > > My use case are two. > One is to have "dynamic layers", in the sense I explained in a previous > email. Let's say I have some code that does clustering on some data source. > I would like to expose it as "live", meaning that it appears as any other > static layer, but it's processed realtime.
you might be able to achieve that using "plugin layers" which are map layers that can be created and updated by plugins > The other is having the possibility to bridge, easily, any kind of data > source, using python as an adapter. > > giovanni > > > > > 2012/8/29 Barry Rowlingson <[email protected]> >> >> I have a vague memory that the python data provider idea was >> considered a while ago. I was thinking about how one could get direct >> access from Qgis to R spatial data sets stored in R's binary files. >> One method would have been to write a provider in python that used Rpy >> to bridge to R, but wasn't possible because you can't write data >> providers in Python. It could be done in C by linking to R's C >> library, but at that point I lost interest and used a looser coupling >> to get my R data into Qgis. >> >> The argument then went that if the provider was designed to access a >> file it would be a better idea to write a GDAL/OGR driver, then even >> people outside QGIS could use it. Obviously that would have to be >> C/C++ though. >> >> And then I think everyone got bored and went back to doing things. >> >> Do you have a specific use-case? >> >> On Wed, Aug 29, 2012 at 8:01 AM, G. Allegri <[email protected]> wrote: >> > Anybody interested in developing the concept and evaluating its >> > feasibility? >> > I'm considering the possibility to sponsor it. >> > >> > giovanni >> > >> > Sent from Nexus >> > >> > Il giorno 28/ago/2012 00:11, "G. Allegri" <[email protected]> ha >> > scritto: >> >> >> >> AFAIK it isn't possible to register a python script as a dataprovider. >> >> Maybe it sounds odd, but I think it would interesting if we could have >> >> it. >> >> Has it been ever considered? How hard would it be to implement it? >> >> >> >> Giovanni >> >> >> >> Sent from Nexus > > > > _______________________________________________ > Qgis-developer mailing list > [email protected] > http://lists.osgeo.org/mailman/listinfo/qgis-developer > _______________________________________________ Qgis-developer mailing list [email protected] http://lists.osgeo.org/mailman/listinfo/qgis-developer
