[ckan-dev] pycsw

Mon Oct 14 13:18:35 UTC 2013

Hi: I will try to setup a CKAN development environment, but in the
meantime, comments interleaved:

On Fri, Oct 11, 2013 at 7:05 AM, Adrià Mercader <adria.mercader at okfn.org> wrote:
> Hi Hilde,
>
> I'm sorry to hear that you are finding problems. The ckan-pycsw is
> relatively recent addition and may still have some rough edges.
>
> Let's take it one step at a time and see if we can make progress, see
> below for comments.
>
> On 10 October 2013 15:19, Hildegard Gerlach
> <hildegard.gerlach at jrc.ec.europa.eu> wrote:
>> Dear Angelos and Adrià,
>> I am using Redhat 6, ckan version 2.2a
>> PostgreSQL 8.4.13, Postgis 1.5.3
>>  SQLAlchemy-0.7.8   (installed by CKAN requirements.txt)
>
> I strongly recommend always using stable versions unless you are
> developing CKAN or the extensions. Use ckan 2.1, and the "stable"
> branches of ckanext-spatial and ckanext-harvest
> The rest of the versions seem fine
>
>> I am using a Postgis database for CKAN and have done harvesting from 2
>> remote sources.
> Great.
>
>> However I don't manage to load these sources into the newly
>> created database for pycsw.
>> In the pycsw.cfg file I have only changed the home and database settings.
>> The database table has been created successfully with this command
>>

As reference, can we see/you post what a few of the metadata records
look like from these sources?

>> paster ckan-pycsw setup -p /etc/ckan/default/pycsw.cfg
> That sounds fine
>
>
>> In the meantime I went back to version 2.1, but I am not sure if this made
>> any change.
>> I did
>>
>> git checkout release-v2.1
>> git pull
>>
>> pip install --upgrade -r requirements.txt
>>
>> for CKAN, the harvester and spatial extension. Is this the correct way ?
> I wouldn't use the --upgrade flag unless you find some version number
> conflicts. This will upgrade all dependencies which might have
> unexpected results
>
>
>> When harvesting before I had not specified anything in the development.ini
>> file for
>> Now I have set it to
>> ckan.spatial.validator.profiles = iso19193eden
> That is good, this is probably the best profile for generic iso
> documents (iso19193ngdc is even a bit more permissive)
>
>> and tried to reharvest, but then I get another error:
>> Error gathering the identifiers from the CSW server [Please use the updated
>> 'getrecords2' method instead of 'getrecords'. The 'getrecords' method will
>> be upgraded to use the 'getrecords2' parameters in a future version of
>> OWSLib
> That might be an issue with an incompatible version of OWSLib that got
> installed when you upgraded the requirements, we've had some issues
> with incompatibilities with owslib.
>
> What version of OWSLib do you have installed? Run the following with
> your virtualenv activated and post the result:
>

For the record, this is a Python DeprecationWarning in OWSLib.  There
have been some huge improvements to CSW in OWSLib (in
owslib.csw.getrecords2, basically improved filter support), which will
replace/become owslib.csw.getrecords).  Having said this, this is just
a warning for downstream developers to update their use of OWSLib.

> pip freeze | grep -i owslib
>
>> A workaround, but not solution, is here:
>> https://www.seegrid.csiro.au/wiki/Infosrvices/CKANHarvestingGuide
>> The workaround (not a fix) for the above error is to comment out the lines
>> (186-188) that raised the DeprecationWarning exception in the following
>> Python source file (/usr/lib/ckan/default/src/owslib/owslib/csw.py).
>> #raise DeprecationWarning("""Please use the updated 'getrecords2' method
>> instead of 'getrecords'. #The 'getrecords' method will be upgraded to use
>> the 'getrecords2' parameters #in a future version of OWSLib.""")
> Sounds like we should update our harvesters to use getrecords2. We
> have never found this, so there definitely seems to be a conflict with
> a newer version of OWSlib.

ckanext-spatial devs:

Agreed. From looking at https://github.com/okfn/ckanext-spatial and
doing `grep -R getrecords *`, the only thing you have to change is
your handling of typenames.

In owslib.csw.getrecords, qtype is a helper to define what kind of CSW
metadata to query for (default is 'dataset', others are 'service',
etc.  This is loosely tied to dc:type or apiso:Type.

In owslib.csw.getrecords2 (which will become owslib.csw.getrecords),
you pass this as an OWSLib constraint object (see
https://github.com/geopython/OWSLib/blob/master/owslib/csw.py#L284) in
a list.

I've updated my fork of ckanext-spatial per above:
https://github.com/tomkralidis/ckanext-spatial/commit/1fdf0cde4c6feed2b3f0675ff88145dd4b626e6e
(this is UNTESTED).  Should I submit this as a PR?  Do ckanext-spatial
PRs run tests? Then I can submit PR.

> We'll investigate, but we'll need to OWSLib version you are using that
> you can find with the command I mentioned before.
>
>
>> Anyway, if  I look at the harvest object like you described I don't know
>> what is wrong with it. It is too big to post on the list.
> Can you post it online somewhere and link to it? I'm curious to see
> what might be causing the ckan-pycsw command to fail.
>
>
>> I have also another problem, I cannot see the "Add harvest source" button on
>> the
>>
>> http://{my-ckan-instance}/harvest
>>
>> Website. It has disapperared.
> Make sure that you are using the latest "stable" branch on
> ckanext-harvest, we fixed some problems with this a while ago:
>
> git checkout stable
> git pull
>
> (and of course that you are logged in)
>
>
>> Sorry, but I am getting so many issues at the same time.
> Sorry about the troubles, hopefully we'll get there.
>
> Adrià
>
>
>
>> Hilde
>>
>>
>> _______________________________________________
>> ckan-dev mailing list
>> ckan-dev at lists.okfn.org
>> http://lists.okfn.org/mailman/listinfo/ckan-dev
>> Unsubscribe: http://lists.okfn.org/mailman/options/ckan-dev
>>
>
> _______________________________________________
> ckan-dev mailing list
> ckan-dev at lists.okfn.org
> http://lists.okfn.org/mailman/listinfo/ckan-dev
> Unsubscribe: http://lists.okfn.org/mailman/options/ckan-dev