class ForumDocument(SumoDocument):
"""
ES document for forum posts. Thread information is duplicated across all posts in that thread.
"""
thread_title = field.Text()
thread_forum_id = field.Keyword()
thread_created = field.Date()
thread_creator_id = field.Keyword()
thread_is_locked = field.Boolean()
thread_is_sticky = field.Boolean()
content = field.Text()
author_id = field.Keyword()
created = field.Date()
updated = field.Date()
updated_by_id = field.Keyword()
class Index:
name = config.FORUM_INDEX_NAME
using = config.DEFAULT_ES7_CONNECTION
def get_field_value(self, field, instance, *args):
if field.startswith("thread_"):
instance = instance.thread
field = field[len("thread_"):]
return super().get_field_value(field, instance, *args)
@classmethod
def get_model(cls):
return Post
@classmethod
def get_queryset(cls):
return Post.objects.select_related("thread")
class Ssn_trace(InnerDoc):
is_valid = field.Boolean()
is_deceased = field.Boolean()
ssn = field.Keyword()
human_message = field.Text()
issued = field.Object(Ssn_issued)
def test_boolean_deserialization():
bf = field.Boolean()
assert not bf.deserialize("false")
assert not bf.deserialize(False)
assert not bf.deserialize("")
assert not bf.deserialize(0)
assert bf.deserialize(True)
assert bf.deserialize("true")
assert bf.deserialize(1)
def _schema2doc_map(self):
_map = {
'integer':
dsl_field.Long(),
'number':
dsl_field.ScaledFloat(scaling_factor=100),
'string':
dsl_field.Text(analyzer=polish_analyzer,
fields={
'raw': dsl_field.Text(),
'keyword': dsl_field.Keyword(),
}),
'any':
dsl_field.Text(analyzer=polish_analyzer,
fields={
'raw': dsl_field.Text(),
'keyword': dsl_field.Keyword(),
}),
'boolean':
dsl_field.Boolean(),
'time':
dsl_field.Text(
fields={
'text': dsl_field.Text(),
'time': dsl_field.Date(
format=constance_config.TIME_FORMATS),
}),
'duration':
dsl_field.DateRange(),
'default':
dsl_field.Text(),
'date':
dsl_field.Text(
fields={
'text': dsl_field.Text(),
'date': dsl_field.Date(
format=constance_config.DATE_FORMATS),
}),
'datetime':
dsl_field.Text(
fields={
'text':
dsl_field.Text(),
'datetime':
dsl_field.Date(format=constance_config.DATE_FORMATS),
})
}
for key, val in _map.items():
_map[key] = CustomObject(properties={
'val': val,
'repr': dsl_field.Keyword(),
})
return _map
class ForumDocument(SumoDocument):
"""
ES document for forum posts. Thread information is duplicated across all posts in that thread.
"""
thread_title = field.Text()
thread_forum_id = field.Keyword()
forum_slug = field.Keyword()
thread_id = field.Keyword()
thread_created = field.Date()
thread_creator_id = field.Keyword()
thread_is_locked = field.Boolean()
thread_is_sticky = field.Boolean()
content = field.Text()
author_id = field.Keyword()
created = field.Date()
updated = field.Date()
updated_by_id = field.Keyword()
class Index:
pass
def prepare_forum_slug(self, instance):
return instance.thread.forum.slug
def get_field_value(self, field, instance, *args):
if field.startswith("thread_"):
instance = instance.thread
field = field[len("thread_"):]
return super().get_field_value(field, instance, *args)
@classmethod
def get_model(cls):
return Post
@classmethod
def get_queryset(cls):
return Post.objects.prefetch_related("thread", "thread__forum")
class CompanyDocType(DocType):
date_of_creation = FormattedDate(date_format='%Y-%m-%d')
description = field.Text()
employees = field.Text()
facebook_url = field.Text()
pk = field.Integer()
keywords = field.Text()
linkedin_url = field.Text()
logo = field.Text()
has_single_sector = field.Boolean()
modified = FormattedDate(date_format='%Y-%m-%dT%H:%M:%S.%fZ')
name = field.Text()
number = field.Text()
sectors = field.Text(multi=True)
sectors_label = field.Text(multi=True)
slug = field.Text()
summary = field.Text()
twitter_url = field.Text()
website = field.Text()
supplier_case_studies = field.Nested(
properties={
'pk': field.Integer(),
'title': field.Text(),
'short_summary': field.Text(),
'description': field.Text(),
'sector': field.Text(),
'keywords': field.Text(),
'image_one_caption': field.Text(),
'image_two_caption': field.Text(),
'image_three_caption': field.Text(),
'testimonial': field.Text(),
'slug': field.Text(),
})
class Meta:
index = 'company'
POS: 'NOUN',
'OP': "+"
}]), ('pnp', [{
POS: "ADJ",
"OP": "*"
}, {
POS: "PROPN",
"OP": "+"
}])]
PD2ES_TYPES = {
np.dtype('O'): field.Text(),
np.dtype('int64'): field.Integer(),
np.dtype('float64'): field.Double(),
np.dtype('<M8[ns]'): field.Date(),
np.dtype('bool'): field.Boolean()
}
def load_spacy_model():
nlp = spacy.load('en_core_web_lg')
nlp.tokenizer.infix_finditer = re.compile(r'[~\-_]').finditer
return nlp
def make_matcher(nlp, patterns=PATTERNS):
matcher = spacy.matcher.Matcher(nlp.vocab)
for _id, pattern in PATTERNS:
matcher.add(_id, None, pattern)
return matcher
class AnswerDocument(QuestionDocument):
"""
ES document for Answers. Every Answer in DB gets an AnswerDocument in ES.
Child class to QuestionDocument, with fields here un-prefixed.
This document defines the answer-specific fields which are included in an AnswerDocument
in addition to the de-normalized fields of an Answer's Question which are defined in
QuestionDocument. Since QuestionDocument and AnswerDocument are stored in the same index,
ES sees QuestionDocuments and AnswerDocuments the same, just with some documents missing
certain fields.
Enables aggregations on answers, such as when creating contribution metrics, and enables
searching within an AAQ thread, or on Answer-specific properties like being a solution.
"""
creator_id = field.Keyword()
created = field.Date()
content = SumoLocaleAwareTextField(term_vector="with_positions_offsets")
updated = field.Date()
updated_by_id = field.Keyword()
num_helpful_votes = field.Integer()
num_unhelpful_votes = field.Integer()
is_solution = field.Boolean()
@classmethod
def prepare(cls, instance, **kwargs):
"""Override super method to exclude certain docs."""
# Add a discard field in the document if the following conditions are met
# Answer document is spam
if instance.is_spam or instance.question.is_spam:
instance.es_discard_doc = "unindex_me"
obj = super().prepare(instance, **kwargs)
# add a prefix to the id so we don't clash with QuestionDocuments
obj.meta.id = "a_{}".format(obj.meta.id)
return obj
def prepare_is_solution(self, instance):
solution_id = instance.question.solution_id
return solution_id is not None and solution_id == instance.id
def prepare_locale(self, instance):
return instance.question.locale
def prepare_num_helpful_votes(self, instance):
if hasattr(instance, "es_num_helpful_votes"):
return instance.es_num_helpful_votes
return instance.num_helpful_votes
def prepare_num_unhelpful_votes(self, instance):
if hasattr(instance, "es_num_unhelpful_votes"):
return instance.es_num_unhelpful_votes
return instance.num_unhelpful_votes
def prepare_answer_content(self, instance):
# clear answer_content field from QuestionDocument,
# as we don't need the content of sibling answers in an AnswerDocument
return None
def get_field_value(self, field, instance, *args):
if field.startswith("question_"):
instance = instance.question
return super().get_field_value(field, instance, *args)
def to_action(self, *args, **kwargs):
# if the id is un-prefixed, add it
if not str(self.meta.id).startswith("a_"):
self.meta.id = f"a_{self.meta.id}"
return super().to_action(*args, **kwargs)
@classmethod
def get(cls, id, **kwargs):
# if the id is un-prefixed, add it
if not str(id).startswith("a_"):
id = f"a_{id}"
return super().get(id, **kwargs)
@classmethod
def get_model(cls):
return Answer
@classmethod
def get_queryset(cls):
return (
Answer.objects
# prefetch each answer's question,
# applying the same optimizations as in the QuestionDocument
.prefetch_related(
Prefetch("question", queryset=QuestionDocument.get_queryset()))
# count votes in db to improve performance
.annotate(
es_num_helpful_votes=Count("votes",
filter=Q(votes__helpful=True)),
es_num_unhelpful_votes=Count("votes",
filter=Q(votes__helpful=False)),
))
class QuestionDocument(SumoDocument):
"""
ES document for Questions. Every Question in DB gets a QuestionDocument in ES.
Parent class to AnswerDocument, with most fields here prefixed with "question_".
This document defines the question-specific fields (most of) which are de-normalized
in the AnswerDocument. Since QuestionDocument and AnswerDocument are stored in the
same index, ES sees QuestionDocuments and AnswerDocuments the same, just with some
documents missing certain fields.
Enables searching for AAQ threads as a unit.
"""
question_id = field.Keyword()
question_title = SumoLocaleAwareTextField()
question_creator_id = field.Keyword()
question_content = SumoLocaleAwareTextField(
term_vector="with_positions_offsets")
question_created = field.Date()
question_updated = field.Date()
question_updated_by_id = field.Keyword()
question_has_solution = field.Boolean()
question_is_locked = field.Boolean()
question_is_archived = field.Boolean()
question_product_id = field.Keyword()
question_topic_id = field.Keyword()
question_taken_by_id = field.Keyword()
question_taken_until = field.Date()
question_tag_ids = field.Keyword(multi=True)
question_num_votes = field.Integer()
# store answer content to optimise searching for AAQ threads as a unit
answer_content = SumoLocaleAwareTextField(
multi=True, term_vector="with_positions_offsets")
locale = field.Keyword()
class Index:
name = config.QUESTION_INDEX_NAME
using = config.DEFAULT_ES7_CONNECTION
@classmethod
def prepare(cls, instance):
"""Override super method to exclude certain docs."""
# Add a discard field in the document if the following conditions are met
# Question document is spam
if instance.is_spam:
instance.es_discard_doc = "unindex_me"
return super(QuestionDocument, cls).prepare(instance)
def prepare_question_tag_ids(self, instance):
return [tag.id for tag in instance.tags.all()]
def prepare_question_has_solution(self, instance):
return instance.solution_id is not None
def prepare_question_num_votes(self, instance):
if hasattr(instance, "es_question_num_votes"):
return instance.es_question_num_votes
return instance.num_votes
def prepare_answer_content(self, instance):
return [
answer.content for answer in (
# when bulk indexing use answer queryset prefetched in `get_queryset` method
# this is to avoid running an extra query for each question in the chunk
instance.es_question_answers_not_spam if hasattr(
instance, "es_question_answers_not_spam")
# fallback if non-spam answers haven't been prefetched
else instance.answers.filter(is_spam=False))
]
def get_field_value(self, field, *args):
if field.startswith("question_"):
field = field[len("question_"):]
return super().get_field_value(field, *args)
@classmethod
def get_model(cls):
return Question
@classmethod
def get_queryset(cls):
return (
Question.objects
# prefetch answers which aren't spam to avoid extra queries when iterating over them
.prefetch_related(
Prefetch(
"answers",
queryset=Answer.objects.filter(is_spam=False),
to_attr="es_question_answers_not_spam",
))
# prefetch tags to avoid extra queries when iterating over them
.prefetch_related("tags")
# count votes in db to improve performance
.annotate(es_question_num_votes=Count("votes")))
_schema2doc_map = {
'integer': dsl_field.Float(),
'number': dsl_field.ScaledFloat(scaling_factor=100),
'string': dsl_field.Text(
analyzer=polish_analyzer,
fields={
'raw': dsl_field.Text(),
}
),
'any': dsl_field.Text(
analyzer=polish_analyzer,
fields={
'raw': dsl_field.Text(),
}
),
'boolean': dsl_field.Boolean(),
'date': dsl_field.Date(),
'datetime': dsl_field.Date(),
'time': dsl_field.Date()
}
_schema_to_api_field = {
'integer': api_fields.Number,
'number': api_fields.Number,
'string': api_fields.String,
'any': api_fields.String,
'boolean': api_fields.Boolean,
'date': api_fields.DateTime,
'datetime': api_fields.DateTime,
'time': api_fields.Time
}
class ShpData(IndexedData):
_type = 'geo'
_schema2doc_map = {
'C':
dsl_field.Text(
analyzer=polish_analyzer,
fields={
'raw': dsl_field.Text(),
'keyword': dsl_field.Keyword(),
},
),
'D':
dsl_field.Date(),
'N':
dsl_field.ScaledFloat(scaling_factor=100),
'L':
dsl_field.Boolean(),
'@':
dsl_field.Date(),
'I':
dsl_field.Long(),
'+':
dsl_field.Long(),
'F':
dsl_field.Float(),
'O':
dsl_field.Double(),
}
_schema_to_api_field = {
'C': api_fields.String,
'D': api_fields.DateTime,
'N': api_fields.Number,
'L': api_fields.Boolean,
'@': api_fields.DateTime,
'I': api_fields.Number,
'+': api_fields.Number,
'F': api_fields.Number,
'O': api_fields.Number,
}
_schema_long_names = {
'C': 'string',
'D': 'datetime',
'N': 'number',
'L': 'boolean',
'@': 'datetime',
'I': 'integer',
'+': 'integer',
'F': 'number',
'O': 'number',
}
_source = None
_schema = None
_transformer = None
def __init__(self, resource, from_table_index=False):
super().__init__(resource)
self.from_table_index = from_table_index
@property
def has_geo_data(self):
return True
@property
def is_chartable(self):
fields = self.schema
return len(fields) > 1 and any(
(field.type in ('N', 'I', '+', 'F', 'O') for field in fields))
@property
def source(self):
if not self._source:
with ArchiveReader(self.resource.main_file.path) as extracted:
shp_path = next(
iter(f for f in extracted if f.endswith('.shp')))
self._source = shapefile.Reader(shp_path)
self._transformer = ShapeTransformer(extracted)
return self._source
def get_schema(self, **kwargs):
use_aliases = kwargs.get('use_aliases', False)
headers = self.reversed_headers_map
return {
'fields': [{
'name': headers[item.name] if use_aliases else item.name,
'type': self._schema_long_names[item.type],
'format': 'default'
} for item in self.schema]
}
@property
def schema(self):
if not self._schema:
self._schema = [
DBSchemaField(*_f) for _f in self.source.fields[1:]
]
return self._schema
def prepare_doc(self):
_fields = {
'shape':
dsl_field.GeoShape(),
'point':
dsl_field.GeoPoint(),
'shape_type':
dsl_field.Integer(),
'label':
dsl_field.Text(),
'resource':
dsl_field.Nested(
properties={
'id':
dsl_field.Integer(),
'title':
dsl_field.Text(analyzer=polish_analyzer,
fields={'raw': dsl_field.Keyword()})
}),
'updated_at':
dsl_field.Date(),
'row_no':
dsl_field.Long()
}
_map = {}
for idx, _f in enumerate(self.schema, 1):
if _f.type not in self._schema2doc_map:
continue
alias_name = _f.name
field_name = f'col{idx}'
_field = self._schema2doc_map[_f.type]
_map[field_name] = alias_name
_fields[field_name] = _field
_fields['Index'] = type('Index', (type, ), {'name': self.idx_name})
doc = type(self.idx_name, (Document, ), _fields)
doc._doc_type.mapping._meta['_meta'] = {'headers': _map}
return doc
def get_api_fields(self):
record_fields = {}
for f in self.schema:
field_name = self.reversed_headers_map[f.name]
field_cls = self._schema_to_api_field[f.type]
record_fields[field_name] = field_cls(is_tabular_data_field=True)
return record_fields
@staticmethod
def _get_row_id(row):
return str(
uuid.uuid5(uuid.NAMESPACE_DNS,
'+|+'.join(str(i)[:10000] for i in row)))
def _docs_iter(self, doc):
for row_no, sr in enumerate(self.source.shapeRecords(), 1):
geojson = self._transformer.transform(sr.shape)
v = {
'shape': geojson,
'updated_at': datetime.now(),
'row_no': row_no,
'resource': {
'id': self.resource.id,
'title': self.resource.title
},
}
for i, val in enumerate(sr.record, 1):
v[f'col{i}'] = val if val != b'' else None
v['shape_type'] = sr.shape.shapeType
v['point'] = median_point(geojson)
tds = self.resource.tabular_data_schema
if tds is not None and 'geo' in tds and 'label' in tds['geo']:
v['label'] = sr.record[tds['geo']['label'].get('col_name')]
d = doc(**v)
d.meta.id = self._get_row_id(sr.record)
yield d
def __init__(self, *args, **kwargs):
super(ContributorField, self).__init__(*args, **kwargs)
self.properties['username'] = field.String(index='not_analyzed')
self.properties['is_freelance'] = field.Boolean()
class CompanyDocument(Document):
wildcard = field.Text(analyzer=american_english_analyzer)
casestudy_wildcard = field.Text(analyzer=american_english_analyzer)
keyword_wildcard = field.Keyword()
case_study_count = field.Integer()
date_of_creation = field.Date(index=False)
description = field.Text(
copy_to='wildcard', analyzer=american_english_analyzer
)
has_description = field.Boolean()
employees = field.Keyword(index=False, store=True)
facebook_url = field.Keyword(index=False, store=True)
pk = field.Integer(index=False)
keywords = field.Text(copy_to='wildcard')
linkedin_url = field.Keyword(index=False, store=True)
logo = field.Keyword(index=False, store=True)
has_single_sector = field.Boolean()
modified = field.Date(index=False)
ordering_name = field.Keyword()
name = field.Text(copy_to=['wildcard', 'ordering_name'])
number = field.Keyword(copy_to='keyword_wildcard',)
sectors = field.Keyword(multi=True, copy_to='keyword_wildcard', store=True)
sectors_label = field.Keyword(
multi=True, copy_to='keyword_wildcard', store=True
)
expertise_industries = field.Keyword(
multi=True, copy_to='keyword_wildcard', store=True
)
expertise_regions = field.Keyword(
multi=True, copy_to='keyword_wildcard', store=True
)
expertise_languages = field.Keyword(
multi=True, copy_to='keyword_wildcard', store=True
)
expertise_countries = field.Keyword(
multi=True, copy_to='keyword_wildcard', store=True
)
# Represents Dict as it's the primitive datatype for this field
expertise_products_services = field.Object()
expertise_products_services_labels = field.Keyword(
multi=True, copy_to='keyword_wildcard', store=True
)
expertise_labels = field.Keyword(
multi=True, copy_to='keyword_wildcard', store=True
)
slug = field.Keyword(copy_to='keyword_wildcard', store=True)
summary = field.Text(
copy_to='wildcard', analyzer=american_english_analyzer
)
twitter_url = field.Keyword(index=False, store=True)
website = field.Keyword(copy_to='keyword_wildcard', store=True)
supplier_case_studies = field.Nested(
properties={
'pk': field.Integer(index=False),
'title': field.Text(copy_to='casestudy_wildcard'),
'short_summary': field.Text(copy_to='casestudy_wildcard'),
'description': field.Text(copy_to='casestudy_wildcard'),
'sector': field.Keyword(copy_to='keyword_wildcard', store=True),
'keywords': field.Text(copy_to='casestudy_wildcard'),
'image_one_caption': field.Text(copy_to='casestudy_wildcard'),
'image_two_caption': field.Text(copy_to='casestudy_wildcard'),
'image_three_caption': field.Text(copy_to='casestudy_wildcard'),
'testimonial': field.Text(copy_to='casestudy_wildcard'),
'website': field.Keyword(copy_to='casestudy_wildcard', store=True),
'slug': field.Keyword(copy_to='keyword_wildcard', store=True),
'testimonial_name': field.Keyword(
copy_to='casestudy_wildcard', store=True
),
'testimonial_company': field.Text(copy_to='casestudy_wildcard'),
'testimonial_job_title': field.Text(copy_to='casestudy_wildcard'),
}
)
is_showcase_company = field.Boolean()
is_published_investment_support_directory = field.Boolean()
is_published_find_a_supplier = field.Boolean()
class Meta:
index = settings.ELASTICSEARCH_COMPANY_INDEX_ALIAS
class QuestionDocument(SumoDocument):
"""
ES document for Questions. Every Question in DB gets a QuestionDocument in ES.
Parent class to AnswerDocument, with most fields here prefixed with "question_".
This document defines the question-specific fields (most of) which are de-normalized
in the AnswerDocument. Since QuestionDocument and AnswerDocument are stored in the
same index, ES sees QuestionDocuments and AnswerDocuments the same, just with some
documents missing certain fields.
Enables searching for AAQ threads as a unit.
"""
question_id = field.Keyword()
question_title = SumoLocaleAwareTextField()
question_creator_id = field.Keyword()
question_content = SumoLocaleAwareTextField(
term_vector="with_positions_offsets")
question_created = field.Date()
question_updated = field.Date()
question_updated_by_id = field.Keyword()
question_has_solution = field.Boolean()
question_is_locked = field.Boolean()
question_is_archived = field.Boolean()
question_is_spam = field.Boolean()
question_marked_as_spam = field.Date()
question_marked_as_spam_by_id = field.Keyword()
question_product_id = field.Keyword()
question_topic_id = field.Keyword()
question_taken_by_id = field.Keyword()
question_taken_until = field.Date()
question_tag_ids = field.Keyword(multi=True)
question_num_votes = field.Integer()
# store answer content to optimise searching for AAQ threads as a unit
answer_content = SumoLocaleAwareTextField(
multi=True, term_vector="with_positions_offsets")
locale = field.Keyword()
class Index:
name = config.QUESTION_INDEX_NAME
using = config.DEFAULT_ES7_CONNECTION
def prepare_question_tag_ids(self, instance):
return [tag.id for tag in instance.tags.all()]
def prepare_question_has_solution(self, instance):
return instance.solution_id is not None
def prepare_question_num_votes(self, instance):
if hasattr(instance, "es_question_num_votes"):
return instance.es_question_num_votes
return instance.num_votes
def prepare_answer_content(self, instance):
return [answer.content for answer in instance.answers.all()]
def get_field_value(self, field, *args):
if field.startswith("question_"):
field = field[len("question_"):]
return super().get_field_value(field, *args)
@classmethod
def get_model(cls):
return Question
@classmethod
def get_queryset(cls):
return (Question.objects.prefetch_related("answers")
# prefetch tags to avoid extra queries when iterating over them
.prefetch_related("tags")
# count votes in db to improve performance
.annotate(es_question_num_votes=Count("votes")))
