[Qgis-developer] the new plugin installer

Borys Jurgiel borys at wolf.most.org.pl
Tue Jun 3 20:05:54 EDT 2008

> No, I haven't access to Mac and I boot to win just once upon a time.
> But I'd like to do some tests there.

Ok, I've found windows machine and also made a workaround of mistakes in the 
repo - it seems to be fully usable at last.

> I thought that the directory could be easily get from the file_name
> element from repository XML but it's not that case. My concern about
> not using name of plugin is that these names are likely to be
> translated inside the plugin so comparing them won't work well. 

You're right. I thought that we can't trust the zip file name match the 
directory one, but in fact it's much more trustworthy than the plugin name 
(only one doesn't match).

> So, my suggestion is to add attribute "id" to pyqgis_plugin element which
> will be identifier of the plugin and will be the same as the directory
> when the plugin resides. What do you think?

I think that would be firmly less trustworthy :) Most of us pack zip files 
with a click on a source directory, so the name is just copied. We can add 
any attributes and related methods in modules (like keywords, categories 
etc), but we must be prepared for the fact, that they both won't match in 
many cases. So it's not the way for identifying, as I think. 

If nobody disagree, I'll use the zip filename as the plugin id. 

> Nice list indeed!
> That reloading of the upgraded plugin in fact should be implemented at
> the level of PyQGIS support and not directly in the installer. It
> requires some internal actions so leaving that task for pythonqgis
> library seems to be a better option. I'll look into this and I'll let
> you know how to do reloads directly from the installer.

Great, I don't want to go beyond plugin borders until everything works. Now 
there is a notify of restart necessity (forgive me the poem). 

Are we allowed to make changes like that between 0.11 and 1.0 freeze? It's 
quite more then just a bugfix, but doesn't exceed the plugin system cleaning. 
May we?

> Exactly, I'll be glad to see that. In main repository there are
> several great plugins but the list is cluttered with others,
> experimental or not usable by wide audience.
> The server-side implementation should be fairly simple, a matter of
> new flag indicating whether the plugin is "official". The
> administrator would explicitly decide which plugins are official and
> from that the web application will generate two different lists.
> Gary, please could you look at this?

I see that 1.0 will be equipped with a tidy system of plugins :)


More information about the Qgis-developer mailing list