jorgecarleitao / django-sphinxql Goto Github PK

At least I faced with problem that in Django 1.10 removed add_aggregate func, I tryed replace by func add_annotation according to docs 1.8, but it raised another errror.Roll back to Django 1.9(

The readme file indicates INDEXES should contain:

'path': os.path.join(BASE_DIR, '_index'),

But the code is looking for it in caps, so like this:

'PATH': os.path.join(BASE_DIR, '_index'),

Hello! I use UUID as PK in my project. I used to solve this problem in self-written config:

sql_query = SELECT ('x' || translate("id"::text, '-', ''))::bit(64)::bigint AS "id", "id" AS "uid", "text" FROM "documents_document"

But django-sphinxql doesn't allow using such a solution. There are two problems:

1. It generates sql_query from Django queryset, so I can't redefine "id" field and cast it to integer in this way.
2. It matches Sphinx id to DB id.

Is there any way to solve my problem? I didn't find it.

If there no way to do this, I propose to:

1. Add "sql_query" option to "source_params". If "sql_query" is present, "query" attribute is ignored.
2. Add method "get_pk" to Index. It will be used in _annotated_models and _get_query. Like this:

indexes = OrderedDict([(index_obj.get_pk(), index_obj)
                       for index_obj in
                       self._sphinx_queryset[:self.max_search_count]])

This method can be overriden in Index subclass.

If you like this proposal, I will do the job and make pull request soon.

sphinx returns the indexed data from its covering indexes as you know, removing the need for a subsequent mysql call.

i've been able to use this functionality to do pagination via RawQuery but unfortunately the built in paginator isn't able to do limits/offsets or cache the RawQuerySet fetching the entire result for each page.

do you think you can add the functionality to do proper cached pagination with limits?

Omg, please help, I cannot understand how I can override settings for multiple models with multiple indexes, example of settings would be great, thank you!

We also need to check if this also happens on the Sphinx side.

Hi!
My project and the Sphinx indices are on different servers. I set parameters for sphinx:

INDEXES = {
     'searchd_params':{
         'listen': 'example.com:9306:mysql41',
     }
 }

But when i try to search

MyModelIndex.objects.search("searchstring")

i get an error from sphinxql/configuration/connection.py , line 45

port = int(port_string.split(':')[0])
ValueError: invalid literal for int() with base 10: 'example.com'

Sphinx documentation says that the hostname can be found before the first delimiter ":", but not necessarily.
Please correct parameters parsing.

At the moment, installing the app as per documentation fails because the module indexes.py is not loaded and thus the indexes are not registered.

One current fix to this is to import the indexes.py by a module in the project.

Using Django 1.7 app config API may be the way to go here.

Currently we only support Python 3. Given that Django supports both, either we can try to support Python 2.7, or wait for Django to drop support on Python 2.7.

Sphinx indexes letters outside ascii as space separators. For instance, "Frühstück" is indexed was three words "Fr hst ck", see Sphinx docs.

I've just submitted this issue to Sphinx. Meanwhile, we have to cover this situation.