def do_filter(self, engine_slug, queryset, search_text):
"""Performs the full text filter."""
model = queryset.model
content_type = ContentType.objects.get_for_model(model)
pk = model._meta.pk
if has_int_pk(model):
ref_name = "object_id_int"
else:
ref_name = "object_id"
return queryset.extra(
tables = ("watson_searchentry",),
where = (
"watson_searchentry.engine_slug = %s",
"watson_searchentry.search_tsv @@ to_tsquery('{search_config}', %s)".format(
search_config = self.search_config
),
"watson_searchentry.{ref_name} = {table_name}.{pk_name}".format(
ref_name = ref_name,
table_name = connection.ops.quote_name(model._meta.db_table),
pk_name = connection.ops.quote_name(pk.db_column or pk.attname),
),
"watson_searchentry.content_type_id = %s"
),
params = (engine_slug, self.escape_postgres_query(search_text), content_type.id),
)
def _create_model_filter(self, models):
"""Creates a filter for the given model/queryset list."""
filters = Q()
for model in models:
filter = Q()
# Process querysets.
if isinstance(model, QuerySet):
sub_queryset = model
model = model.model
queryset = sub_queryset.values_list("pk", flat=True)
if has_int_pk(model):
filter &= Q(
object_id_int__in = queryset,
)
else:
live_ids = list(queryset)
if live_ids:
filter &= Q(
object_id__in = live_ids,
)
else:
# HACK: There is a bug in Django (https://code.djangoproject.com/ticket/15145) that messes up __in queries when the iterable is empty.
# This bit of nonsense ensures that this aspect of the query will be impossible to fulfill.
filter &= Q(
content_type = ContentType.objects.get_for_model(model).id + 1,
)
# Add the model to the filter.
content_type = ContentType.objects.get_for_model(model)
filter &= Q(
content_type = content_type,
)
# Combine with the other filters.
filters |= filter
return filters
def _create_model_filter(self, models, backend):
"""Creates a filter for the given model/queryset list."""
from django.contrib.contenttypes.models import ContentType
from watson.models import has_int_pk
filters = Q()
for model in models:
filter = Q()
# Process querysets.
if isinstance(model, QuerySet):
sub_queryset = model
model = model.model
queryset = sub_queryset.values_list("pk", flat=True)
if has_int_pk(model):
filter &= Q(object_id_int__in=queryset, )
else:
queryset = queryset.annotate(watson_pk_str=RawSQL(
backend.do_string_cast(
connections[queryset.db],
model._meta.pk.db_column or model._meta.pk.attname,
), ()), ).values_list("watson_pk_str", flat=True)
filter &= Q(object_id__in=queryset, )
# Add the model to the filter.
content_type = ContentType.objects.get_for_model(model)
filter &= Q(content_type=content_type, )
# Combine with the other filters.
filters |= filter
return filters
def do_filter(self, engine_slug, queryset, search_text):
"""Performs the full text filter."""
model = queryset.model
content_type = ContentType.objects.get_for_model(model)
pk = model._meta.pk
if has_int_pk(model):
ref_name = "object_id_int"
else:
ref_name = "object_id"
return queryset.extra(
tables=("watson_searchentry", ),
where=(
"watson_searchentry.engine_slug = %s",
"MATCH (watson_searchentry.title, watson_searchentry.description, watson_searchentry.content) AGAINST (%s IN BOOLEAN MODE)",
"watson_searchentry.{ref_name} = {table_name}.{pk_name}".
format(
ref_name=ref_name,
table_name=connection.ops.quote_name(model._meta.db_table),
pk_name=connection.ops.quote_name(pk.db_column
or pk.attname),
),
"watson_searchentry.content_type_id = %s",
),
params=(engine_slug, self._format_query(search_text),
content_type.id),
)
def _create_model_filter(self, models):
"""Creates a filter for the given model/queryset list."""
from django.contrib.contenttypes.models import ContentType
from watson.models import has_int_pk
filters = Q()
for model in models:
filter = Q()
# Process querysets.
if isinstance(model, QuerySet):
sub_queryset = model
model = model.model
queryset = sub_queryset.values_list("pk", flat=True)
if has_int_pk(model):
filter &= Q(object_id_int__in=queryset, )
else:
live_ids = list(queryset)
if live_ids:
filter &= Q(object_id__in=live_ids, )
else:
# HACK: There is a bug in Django
# (https://code.djangoproject.com/ticket/15145)
# that messes up __in queries when the iterable is empty.
# This bit of nonsense ensures that this aspect of the query
# will be impossible to fulfill.
filter &= Q(
content_type=ContentType.objects.get_for_model(
model).id + 1, )
# Add the model to the filter.
content_type = ContentType.objects.get_for_model(model)
filter &= Q(content_type=content_type, )
# Combine with the other filters.
filters |= filter
return filters
def do_filter(self, engine_slug, queryset, search_text):
"""Performs the full text filter."""
model = queryset.model
content_type = ContentType.objects.get_for_model(model)
connection = connections[queryset.db]
pk = model._meta.pk
if has_int_pk(model):
ref_name = "object_id_int"
ref_name_typecast = ""
else:
ref_name = "object_id"
# Cast to text to make join work with uuid columns
ref_name_typecast = "::text"
return queryset.extra(
tables=("watson_searchentry",),
where=(
"watson_searchentry.engine_slug = %s",
"watson_searchentry.search_tsv @@ to_tsquery('{search_config}', %s)".format(
search_config=self.search_config
),
"watson_searchentry.{ref_name} = {table_name}.{pk_name}{ref_name_typecast}".format(
ref_name=ref_name,
table_name=connection.ops.quote_name(model._meta.db_table),
pk_name=connection.ops.quote_name(pk.db_column or pk.attname),
ref_name_typecast=ref_name_typecast
),
"watson_searchentry.content_type_id = %s"
),
params=(engine_slug, self.escape_postgres_query(search_text), content_type.id),
)
def _get_entries_for_obj(self, obj):
"""Returns a queryset of entries associate with the given obj."""
from django.contrib.contenttypes.models import ContentType
model = obj.__class__
content_type = ContentType.objects.get_for_model(model)
object_id = force_text(obj.pk)
# Get the basic list of search entries.
search_entries = SearchEntry.objects.filter(
content_type = content_type,
engine_slug = self._engine_slug,
)
if has_int_pk(model):
# Do a fast indexed lookup.
object_id_int = int(obj.pk)
search_entries = search_entries.filter(
object_id_int = object_id_int,
)
else:
# Alas, have to do a slow unindexed lookup.
object_id_int = None
search_entries = search_entries.filter(
object_id = object_id,
)
return object_id_int, search_entries
def _create_model_filter(self, models, backend):
"""Creates a filter for the given model/queryset list."""
from django.contrib.contenttypes.models import ContentType
from watson.models import has_int_pk
filters = Q()
for model in models:
filter = Q()
# Process querysets.
if isinstance(model, QuerySet):
sub_queryset = model
model = model.model
queryset = sub_queryset.values_list("pk", flat=True)
if has_int_pk(model):
filter &= Q(
object_id_int__in=queryset,
)
else:
queryset = queryset.annotate(
watson_pk_str=RawSQL(backend.do_string_cast(
connections[queryset.db],
model._meta.pk.db_column or model._meta.pk.attname,
), ()),
).values_list("watson_pk_str", flat=True)
filter &= Q(
object_id__in=queryset,
)
# Add the model to the filter.
content_type = ContentType.objects.get_for_model(model)
filter &= Q(
content_type=content_type,
)
# Combine with the other filters.
filters |= filter
return filters
def do_filter(self, engine_slug, queryset, search_text):
"""Performs the full text filter."""
model = queryset.model
content_type = ContentType.objects.get_for_model(model)
connection = connections[queryset.db]
pk = model._meta.pk
if has_int_pk(model):
ref_name = "object_id_int"
else:
ref_name = "object_id"
return queryset.extra(
tables=("watson_searchentry",),
where=(
"watson_searchentry.engine_slug = %s",
"MATCH (watson_searchentry.title, watson_searchentry.description, watson_searchentry.content) "
"AGAINST (%s IN BOOLEAN MODE)",
"watson_searchentry.{ref_name} = {table_name}.{pk_name}".format(
ref_name=ref_name,
table_name=connection.ops.quote_name(model._meta.db_table),
pk_name=connection.ops.quote_name(pk.db_column or pk.attname),
),
"watson_searchentry.content_type_id = %s",
),
params=(engine_slug, self._format_query(search_text), content_type.id),
)
def _create_model_filter(self, models):
"""Creates a filter for the given model/queryset list."""
from django.contrib.contenttypes.models import ContentType
from watson.models import has_int_pk
filters = Q()
for model in models:
filter = Q()
# Process querysets.
if isinstance(model, QuerySet):
sub_queryset = model
model = model.model
queryset = sub_queryset.values_list("pk", flat=True)
if has_int_pk(model):
filter &= Q(
object_id_int__in=queryset,
)
else:
filter &= Q(
object_id__in=queryset,
)
# Add the model to the filter.
content_type = ContentType.objects.get_for_model(model)
filter &= Q(
content_type=content_type,
)
# Combine with the other filters.
filters |= filter
return filters
def do_filter(self, engine_slug, queryset, search_text):
"""Filters the given queryset according the the search logic for this backend."""
model = queryset.model
connection = connections[queryset.db]
db_table = connection.ops.quote_name(SearchEntry._meta.db_table)
model_db_table = connection.ops.quote_name(model._meta.db_table)
pk = model._meta.pk
id = connection.ops.quote_name(pk.db_column or pk.attname)
# Add in basic filters.
word_query = ["""
({db_table}.{engine_slug} = %s)
""", """
({db_table}.{content_type_id} = %s)
"""]
word_kwargs = {
"db_table": db_table,
"model_db_table": model_db_table,
"engine_slug": connection.ops.quote_name("engine_slug"),
"title": connection.ops.quote_name("title"),
"description": connection.ops.quote_name("description"),
"content": connection.ops.quote_name("content"),
"content_type_id": connection.ops.quote_name("content_type_id"),
"object_id": connection.ops.quote_name("object_id"),
"object_id_int": connection.ops.quote_name("object_id_int"),
"id": id,
"iregex_operator": connection.operators["iregex"],
}
word_args = [
engine_slug,
ContentType.objects.get_for_model(model).id,
]
# Add in join.
if has_int_pk(model):
word_query.append("""
({db_table}.{object_id_int} = {model_db_table}.{id})
""")
else:
word_query.append("""
({db_table}.{object_id} = {model_db_table}.{id})
""")
# Add in all words.
for word in search_text.split():
regex = regex_from_word(word)
word_query.append(
"""
({db_table}.{title} {iregex_operator}
OR {db_table}.{description} {iregex_operator}
OR {db_table}.{content} {iregex_operator})
"""
)
word_args.extend((regex, regex, regex))
# Compile the query.
full_word_query = " AND ".join(word_query).format(**word_kwargs)
return queryset.extra(
tables=(db_table,),
where=(full_word_query,),
params=word_args,
)
def do_filter(self, engine_slug, queryset, search_text):
"""Filters the given queryset according the the search logic for this backend."""
model = queryset.model
connection = connections[queryset.db]
db_table = connection.ops.quote_name(SearchEntry._meta.db_table)
model_db_table = connection.ops.quote_name(model._meta.db_table)
pk = model._meta.pk
id = connection.ops.quote_name(pk.db_column or pk.attname)
# Add in basic filters.
word_query = ["""
({db_table}.{engine_slug} = %s)
""", """
({db_table}.{content_type_id} = %s)
"""]
word_kwargs = {
"db_table": db_table,
"model_db_table": model_db_table,
"engine_slug": connection.ops.quote_name("engine_slug"),
"title": connection.ops.quote_name("title"),
"description": connection.ops.quote_name("description"),
"content": connection.ops.quote_name("content"),
"content_type_id": connection.ops.quote_name("content_type_id"),
"object_id": connection.ops.quote_name("object_id"),
"object_id_int": connection.ops.quote_name("object_id_int"),
"id": id,
"iregex_operator": connection.operators["iregex"],
}
word_args = [
engine_slug,
ContentType.objects.get_for_model(model).id,
]
# Add in join.
if has_int_pk(model):
word_query.append("""
({db_table}.{object_id_int} = {model_db_table}.{id})
""")
else:
word_query.append("""
({db_table}.{object_id} = {model_db_table}.{id})
""")
# Add in all words.
for word in search_text.split():
regex = regex_from_word(word)
word_query.append(
"""
({db_table}.{title} {iregex_operator}
OR {db_table}.{description} {iregex_operator}
OR {db_table}.{content} {iregex_operator})
"""
)
word_args.extend((regex, regex, regex))
# Compile the query.
full_word_query = " AND ".join(word_query).format(**word_kwargs)
return queryset.extra(
tables=(db_table,),
where=(full_word_query,),
params=word_args,
)
def _create_model_filter(self, models):
"""Creates a filter for the given model/queryset list."""
from django.contrib.contenttypes.models import ContentType
from watson.models import has_int_pk
filters = Q()
for model in models:
filter = Q()
# Process querysets.
if isinstance(model, QuerySet):
sub_queryset = model
model = model.model
queryset = sub_queryset.values_list("pk", flat=True)
if has_int_pk(model):
filter &= Q(object_id_int__in=queryset, )
else:
filter &= Q(object_id__in=queryset, )
# Add the model to the filter.
content_type = ContentType.objects.get_for_model(model)
filter &= Q(content_type=content_type, )
# Combine with the other filters.
filters |= filter
return filters
def register(self, model, adapter_cls=SearchAdapter, **field_overrides):
"""
Registers the given model with this search engine.
If the given model is already registered with this search engine, a
RegistrationError will be raised.
"""
# Add in custom live filters.
if isinstance(model, QuerySet):
live_queryset = model
model = model.model
field_overrides["get_live_queryset"] = lambda self_: live_queryset.all()
# Check for existing registration.
if self.is_registered(model):
raise RegistrationError("{model!r} is already registered with this search engine".format(
model = model,
))
# Perform any customization.
if field_overrides:
adapter_cls = type("Custom" + adapter_cls.__name__, (adapter_cls,), field_overrides)
# Perform the registration.
adapter_obj = adapter_cls(model)
self._registered_models[model] = adapter_obj
# Add in a generic relation, if not exists.
if not hasattr(model, "searchentry_set"):
if has_int_pk(model):
object_id_field = "object_id_int"
else:
object_id_field = "object_id"
generic_relation = generic.GenericRelation(
SearchEntry,
object_id_field = object_id_field,
)
model.searchentry_set = generic_relation
generic_relation.contribute_to_class(model, "searchentry_set")
# Connect to the signalling framework.
post_save.connect(self._post_save_receiver, model)
pre_delete.connect(self._pre_delete_receiver, model)