[Interop-dev] Python packages

Nemesis (spam-protected)
Fri May 22 12:21:15 CEST 2015


On 05/22/2015 11:46 AM, Mitar wrote:
> Hi!
>
>> Is there any particular feature you have been able to extract into a
>> separate python packagethat we can start contributing to?
...
> Improved UltraJSON:
>
> https://github.com/wlanslovenija/ultrajson

I'm also interested in ujson. Which improvements are there? Will you try
to merge them back in the official package?

...
>>   * https://github.com/djangonauts/django-rest-framework-gis (geojson api)
>>   * https://github.com/djangonauts/django-hstore (key value store in
>>     postgres)
>>   * https://github.com/djangonauts/django-rest-framework-hstore (add
>>     support for hstore in django-rest-framework)
> I see that you are using Django REST framework? How do you like it? We
> have been using Tastypie until now and made quite some code around it,
> but as it is not so actively developed, I am not sure if we should
> slowly move away from it. But Django REST framework didn't really look
> like it has all things we would need and get used to, at least the last
> time I checked.

Initially we started to tastypie too.

I don't know about you guys but our experience was terrible, I don't
remember which version it was, but it didn't feel like developing stuff
for django.

I tried django-rest-framework also on other projects and it felt much
easier so I rewrote nodeshot to use DRF and started contributing
actively to it. The community is quite active, there are many third
party plugins, and the core-devs of DRF help third-party plugins by
promoting them on their website, which helps to reduce duplication
(different packages with same purposes).
http://www.django-rest-framework.org/topics/third-party-resources/

Look, there's also a ujson renderer:
https://github.com/gizmag/drf-ujson-renderer

On 05/22/2015 11:52 AM, Jernej Kos wrote:
> On 22. 05. 2015 11:46, Mitar wrote:
>>>   * a proof of concept of storage for metrics/statistics built with
>>>     influxdb 0.9 rc:
>>>     https://github.com/ninuxorg/nodeshot/tree/master/nodeshot/core/metrics
>> So this could be made as an module for datastream. So datastream is
>> trying to server as a common API around time-series data bases. For now
>> only around MongoDB. But maybe it is time for influxdb. :-)
> I was looking into InfluxDB recently, having the same idea of making a
> datastream backend for it. It looks very nice, but the last time I
> checked it lacked certain features that we need.
>
> There were some issues with continuous queries (these are similar to
> datastream's derived streams) and storing non-integer types as values
> (as we are also storing graph topologies in our time-series database, so
> that we can see how topology changes through time).
>
> Perhaps it is time to revisit that?

Initially I considered a few options:

  * mongodb
  * graphite
  * opentsdb

Influxdb was the last one I tried, I was initially skeptical and more
inclined to use mongodb or graphite but after having tried them out I
understood that influxdb has a very high potential to become the next
successful open source timeseries storage, it has an active community
already using it in production, hundreds of contributors and a paid
group of core developers: https://github.com/influxdb/influxdb

Many common functions (min, max, average, standard deviation, ecc.) are
built into the SQL-like langauge:
http://influxdb.com/docs/v0.8/api/aggregate_functions.html

BUT (there is always a but) the 0.9 version is a rewrite, which means
some of the features that were in the 0.8 are still being reimplemented
in the 0.9.
So... we don't have a crystal ball yet, but we can try to bet :-)

Federico

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.funkfeuer.at/pipermail/interop-dev/attachments/20150522/3cddc067/attachment.html>


More information about the Interop-dev mailing list