Hey,

Great to see this conversation happening, hope to be able to
contribute. My two bits inline -

On Sun, Mar 2, 2014 at 3:15 PM, Gora Mohanty <[email protected]> wrote:
> On 2 March 2014 14:51, Sajjad Anwar <[email protected]> wrote:
> [...]
>> Hey, you don't necessarily need geodjango for these. You can write simple
>> spatial queries in PostGIS.
>> And they aren't very expensive. For example -
>> http://postgis.refractions.net/documentation/manual-2.0/ST_Within.html
>
> GeoDjango queries translate to postgis if one is using a postgresql
> back-end, so I do not think that there is much overhead there. GeoDjango
> also does a lot more, including providing an abstraction layer for GIS, so
> that I would prefer using it to coding directly in postgis. The alternatives
> that I was thinking of were along the lines of spatial search with 
> Solr/Lucene.
> These should be faster in principle, but we have not yet benchmarked
> them against postgis/GeoDjango.
>

I wouldn't worry about performance before it's a concern - I think
PostGIS is fairly performant. If one does land up hosting a lot of
data, I'd highly recommend ElasticSearch for its geo support and
scalability, but there's probably a few options. I'd also highly
recommend django-haystack [0] - in most cases, this should let one
fairly easily build out your django project, and add a performant
search indexer when one needs it. I'll be glad to help. Any-how, my
two bits would be "don't premature optimize" - I think building
something out using GeoDjango should be fairly clean and
straightforward, and I think django-haystack provides a clean
migration path forward if one does need to scale / use a specialized
indexing tool like Lucene.

Also, for Django APIs, I've been looking at django-rest-framework [1]
recently, and am quite liking it. It seemed a bit of overhead as
compared to just writing vanilla view functions for end-points, but
gives a really nice, auto-documented API frame-work with a nice
in-browser testing page, multiple formats, etc. I had to do some
tweaking to get it to output GeoJSON - any-how, if you are going down
this path and want a hand / second set of eyes, let me know.

> I would again argue that the first thing to do would be to have a stable 
> system
> that provides an API layer over raw data pulled from various sources. GIS,
> and such queries would be meta-queries over such raw data, and should be
> done later, IMHO.
>

About other sort of aggregation / meta-queries, I agree, but I think
basic GIS data can be quite fundamental to this sort of data. Its nice
to be able to query location data, and for data-sets that have
location data, I think would be nice to support? Thankfully, adding
basic geo support using GeoDjango is fairly straightfoward - I agree
that performance maybe an issue, but I think if the site is getting
that many users / has that much data, one could work out more complex
solutions then?

Anyways, let the bike-shedding not get to you, Thej :) - do whatever's
easiest / makes sense - do let me know if I can help / contribute /
when there's code up, etc.

All the best!
-Sanjay

[0] http://haystacksearch.org/
[1] http://www.django-rest-framework.org/

-- 
For more details about this list
http://datameet.org/discussions/
--- 
You received this message because you are subscribed to the Google Groups 
"datameet" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.

Reply via email to