[gdal-dev] GME driver for OGR

Wolf Bergenheim wolf+grass at bergenheim.net
Mon Oct 28 17:26:00 PDT 2013


Even,

>
> I don't know the GME API, so the following might be non relevant, but I'm
> wondering why you couldn't keep the current connection logic. You would "just"
> need to establish the list of vector tables from a "GetMapContent" request (or
> whatever it is called) and you would instanciate the layers just as you do
> currently by parsing the "tables" option. I'm probably missing something ?

That's actually a nice and elegant solution! :) I like it, it will
keep things simpler and should result in less shuffling of code
around.

> > Is there a precedence
> > for this type of configuration in any other driver?
>
> The boilerplate code is quite typical of the standard OGR driver. You would
> need/want a shared connection among different OGRGMEDataSource ? That would be
> the point of moving it at the driver level. Actually I've coded something like
> that in the FGDB driver since there's a bug in the underlying SDK and you
> cannot open simultaneously several connections to the same datasource.

I'm thinking that there is rarely any reason to separate the
connections, except if, maybe we'd like to support multiple accounts,
so that you can get table A using account A, and table B with account
B. But then again why don't you simply give access to A for both
tables? The way I see it we only need a single connection since the
API is RESTful.
So yeah basically trying to see if I could make the code more
re-usable, and possibly optimizing. I'm not there yet, still
implementing basic functionality, but I like to be one step ahead in
my designs.

>
> Another driver using OAuth2 or those GME concepts in a brother driver ? I have
> some troubles imagining what you have in mind, not having dug myself into the
> topic.
>

Yeah, something like that, can't get into specifics at this point. I
hope we'll announce it soon, though. Again I want to prepare for
what's coming,and looking for some advice on best practices. I hope to
avoid realizing in 2 months that I need to refactor the
OGRGMEDatasource fully, but otoh that's not going to be the end of the
world, if I have to do it that way.

--Wolf


More information about the gdal-dev mailing list