Add support for Faceting

[rhblind]

We should support faceting.
This should probably be implemented as a FacetingFilter, which picks up and filters on faceting queries like narrowing.

https://django-haystack.readthedocs.org/en/latest/faceting.html#faceting

[rhblind]

Suggestions for solution

HaystackFacetFilter

• Read fields, exclude and a special field_options from view.serializer_class.Meta
• Read and parse query parameters to override field_options.
• From field_options; figure out which is date_facets, field_facets and query_facets.
• Apply facets to SearchQuerySet and return it as a queryset. Filters needs to be chainable, therefore it must return a regular SearchQuerySet

Example field_options:

field_options = {
    "firstname": {},
    "lastname": {},
    "created": {
        "start_date": datetime.now() - timedelta(days=3 * 365),
        "end_date": datetime.now(),
        "gap_by": "day",
        "gap_amount": 10
    }
}

field_options can be overridden by adding to query string in the following format:

fieldname=option1:value1,option2:value2

/api/v1/search/?firstname=limit:100&created=start_date:21. Jan 2015,gap_by:day

FacetSerializer (or FacetSerializerMixin)

• Have dates, fields and queries as default fields.
• Needs to have a SearchQuerySet().facet_counts() as its initial data
• Each field should serialize like below

{
    "dates": [
        "created": [
            {
                "text": "2015-05-15T21:08:35",
                "count": 76,
                "narrow": "https://example.com/api/v1/search/dates/created/?text=2015-05-15T21:08:35"
            },
            {
                "text": "2015-06-12T23:14:10",
                "count": 29,
                "narrow": "https://example.com/api/v1/search/dates/created/?text=2015-06-12T23:14:10"
            }
        ]
    ],
    "fields": [
        "firstname": [
            {
                "text": "John",
                "count": 26,
                "narrow": "https://example.com/api/v1/search/fields/firstname/?text=John"
            },
            {
                "text": "Randy",
                "count": 13,
                "narrow": "https://example.com/api/v1/search/fields/firstname/?text=Randy"
            }
        ]
    ],
    "queries": [
        ...
    ]
}

The narrow url must be injected per result. The format however is open for debate.

FacetViewMixin

• Can override list() method in order to feed the SearchQuerySet().facet_counts() as initial data to the serializer_class.
• Provide a @list_route action for narrowing results
• Should take care of parsing the selected_facets from query parameter

[rhblind]

This is basically done and pushed to development branch. Some missing features such as narrowing, but will open a dedicated Issue for that

[hubert10]

Hello Sir?
I have been working with drf haystack elasticsearch and according to the haystack latest documentation, it seems for faceting and more-like-this working, but for me it doesn't.
I am using

elasticsearch==2.4.0
Django==1.10
djangorestframework==3.4.6
django-haystack==2.5.dev1
drf-haystack==1.6.0rc2

serializers and views:
for faceting

class PersonFacetSerializer(HaystackFacetSerializer):
    serialize_objects = True  
    class Meta:
        index_classes = [PersonIndex]
        fields = ["firstname", "lastname", "created"]
        field_options = {
            "firstname": {},
            "lastname": {},
            "created": {
                "start_date": datetime.datetime.now() - timedelta(days=3 * 365),
                "end_date": datetime.datetime.now(),
                "gap_by": "month",
                "gap_amount": 3
            }
        }
class PersonSearchViewSet(FacetMixin, HaystackViewSet):
    index_classes = [PersonIndex]
    serializer_class = PersonSerializer
    facet_filter_backends = [HaystackFacetFilter]
    filter_backends = [HaystackFilter]
    facet_serializer_class = PersonFacetSerializer

for more_like_this

class PersonSerializer(HaystackSerializer):
    more_like_this = serializers.HyperlinkedIdentityField(view_name="search-person-mlt-more-like-this", read_only=True)
    class Meta:
        index_classes = [PersonIndex]
        fields = ["firstname", "lastname"]
class PersonViewSet(MoreLikeThisMixin, HaystackViewSet):
    index_models = [Person]
    serializer_class = PersonSerializer

urls:

routerV2 = routers.DefaultRouter()
router = routers.DefaultRouter()

routerV2.register("person", viewset=PersonSearchView, base_name="Person-facet-search")
router.register("person", viewset=PersonViewSet, base_name="search-more-like-this")
urlpatterns = [
    url(r'^admin/', admin.site.urls),
    url(r'^', include('backend.urls')),
    url(r'^api-auth/', include('rest_framework.urls', namespace='rest_framework')),
    url(r'^api/v2/', include(routerV2.urls)),
    url(r'^search', include('haystack.urls')),
    url(r'^', include(routerV2.urls)),
    url(r'^', include('haystack.urls')),
    url(r'^facets', include(routerV2.urls))
]+static(settings.MEDIA_URL, document_root=settings.MEDIA_ROOT)

queries

http://localhost:8000/api/v2/person/?firstname=hubert
GET /api/v2/person/?firstname=hubert
HTTP 200 OK
  Allow: GET, HEAD, OPTIONS
  Content-Type: application/json
  Vary: Accept

  {
    "count": 3,
    "next": null,
    "previous": null,
    "results": [
        {
            "text": "\nDiallo\nhubert\nOukam\n\n",
            "firstname": "hubert",
            "lastname": "Diallo",
            "city": "Oukam",
            "highlighted": "...<em class=\"highlighted\">hubert</em>\nOukam\n\n"
        },
        {
            "text": "\nkanyamahanga\nhubert\nOukam\n\n",
            "firstname": "hubert",
            "lastname": "kanyamahanga",
            "city": "Oukam",
            "highlighted": "...<em class=\"highlighted\">hubert</em>\nOukam\n\n"
        },
        {
            "text": "\nNiang\nkeba hubert\nFoar\n\n",
            "firstname": "keba hubert",
            "lastname": "Niang",
            "city": "Foar",
            "highlighted": "...<em class=\"highlighted\">hubert</em>\nFoar\n\n"
        }
    ]
}

GET /person/?firstname=hubert
HTTP 200 OK
  Allow: GET, HEAD, OPTIONS
  Content-Type: application/json
  Vary: Accept

  {
    "count": 3,
    "next": null,
    "previous": null,
    "results": [
        {
            "firstname": "hubert",
            "lastname": "Diallo",
            "more_like_this": null
        },
        {
            "firstname": "hubert",
            "lastname": "kanyamahanga",
            "more_like_this": null
        },
        {
            "firstname": "keba hubert",
            "lastname": "Niang",
            "more_like_this": null
        }
    ]
}

Please if there is what I what I am missing, any help, guide or suggestion will be welcomed!

Thank you!