def _search_query_text(query, **kwargs):
"""Prepare the initial query, searching by text and ranking by the proximity."""
language = kwargs.get('language')
text = kwargs.get('text')
if text:
if language is not None:
to_tsquery = func.to_tsquery(SearchItem.getDictForLanguage(language), text)
else:
to_tsquery = func.to_tsquery(text)
query = query.filter(SearchItem.terms.op('@@')(to_tsquery))
return query
def build_query(params, full=False):
columns = [users.c.id, users.c.name, users.c.email_address] if full \
else [users.c.id]
query = select(columns)
query = query.order_by(users.c.id)
if 'email' in params:
query = query.where(users.c.email_address == params['email'])
if 'name' in params:
search_string = " & ".join(
"{}:*".format(word) for word in params['name'].split(' ')
)
name_vector = func.to_tsvector(users.c.name)
query = query.where(name_vector.match(search_string))
query = query.order_by(
desc(func.ts_rank(name_vector, func.to_tsquery(search_string)))
)
# if 'page_size' in params:
# query = query.limit(params['page_size'])
return query
def by_prefix(cls, prefix):
'''
Find search terms with a given prefix.
'''
# In our SQLAlchemy version the language cannot be specified in the
# `match` function, so we construct the query explicitly. See
# https://bitbucket.org/zzzeek/sqlalchemy/issues/3078
tsquery = func.to_tsquery('pg_catalog.simple', "'{}':*".format(prefix))
return cls.filter(cls.term_tsvector.op('@@')(tsquery))
def _search_query_rank(query, **kwargs):
"""
Rank query results, sorting by search rank for most content types and
integrating the rating for subjects.
"""
rank_func = None
text = kwargs.get('text')
language = kwargs.get('language')
rank_cutoff = kwargs.get('rank_cutoff')
if language is not None:
rank_func = func.ts_rank_cd(SearchItem.terms, func.to_tsquery(SearchItem.getDictForLanguage(language), text))
else:
rank_func = func.ts_rank_cd(SearchItem.terms, func.to_tsquery(text))
if rank_cutoff is not None:
query = query.filter(rank_func >= rank_cutoff)
if kwargs.get('use_rating'):
query = query.order_by((SearchItem.rating * rank_func).desc())
else:
query = query.order_by(rank_func.desc())
return query
def construct_fulltext_query_and_rank(searched):
# Text version of the text search query directly constructed from
# the searched string
simple_ts_query_text = cast(func.plainto_tsquery('english', searched),
TEXT)
# Text version of the search query using prefix search for the last
# word in the base query
# Case expression is necessary to capture the case of empty base query
# (arising e.g. when the query consists entirely of special characters
# and stop words)
prefix_ts_query_text = case([(simple_ts_query_text == "", cast("", TEXT))],
else_=simple_ts_query_text.op('||')(cast(
":*", TEXT)))
# Final text search query
ts_query = func.to_tsquery('english', prefix_ts_query_text)
# Rank for each search result
ts_rank = func.ts_rank_cd(Post.__ts_vector__, ts_query).label("rank")
return (ts_query, ts_rank)
def add_simple_text_search(query, text_columns, keywords, include_rank=True):
rank = None
keywords_j = ' & '.join(keywords)
fts_config = 'simple'
filters = [
func.to_tsvector(fts_config, column).match(keywords_j)
for column in text_columns
]
if len(filters) > 1:
filter = or_(*filters)
else:
filter = filters[0]
query = query.filter(filter)
if include_rank:
ranks = [
func.ts_rank(func.to_tsvector(fts_config, column),
func.to_tsquery(fts_config, keywords_j))
for column in text_columns
]
rank = reduce(lambda a, b: a + b, ranks)
query = query.add_column(rank.label('score'))
return query, rank
def search_query(q, search):
# works only for pg backend
search = preprocess_search_query(search)
return q.filter(model.Ebook.full_text.match(search))\
.order_by(desc(func.ts_rank_cd(model.Ebook.full_text, func.to_tsquery(text("'custom'"), search))))
def search_query(q, search):
# works only for pg backend
search = preprocess_search_query(search)
return q.filter(model.Ebook.full_text.match(search))\
.order_by(desc(func.ts_rank_cd(model.Ebook.full_text, func.to_tsquery(text("'custom'"), search))))
def search_locations(search_string):
q = db.session.query(Location).filter(db.text('location.search_column @@ to_tsquery(:terms)'))
q = q.params(terms=search_string)
q = q.order_by(db.text('ts_rank_cd(location.search_column, to_tsquery(:terms)) DESC'))
q = q.add_column(func.ts_headline('pg_catalog.english',db.text('location.body_text'),func.to_tsquery(search_string),'MaxFragments=1, StartSel = <strong>, StopSel = </strong>'))
return [{'object':location_object_to_dict(location), 'fragment':fragment} for location, fragment in q]
def add_text_search(query,
join_columns,
keywords,
locales,
include_rank=True,
lse=None):
from assembl.models.langstrings import LangStringEntry
rank = None
keywords_j = ' & '.join(keywords)
lse = lse or aliased(LangStringEntry)
join_conds = [
lse.langstring_id == join_column for join_column in join_columns
]
if len(join_conds) > 1:
join_cond = or_(*join_conds)
else:
join_cond = join_conds[0]
query = query.join(lse, join_cond)
if locales:
active_text_indices = get('active_text_indices', 'en')
locales_by_config = defaultdict(list)
any_locale = 'any' in locales
for locale in locales:
fts_config = postgres_language_configurations.get(locale, 'simple')
if fts_config not in active_text_indices:
fts_config = 'simple'
locales_by_config[fts_config].append(locale)
conds = {}
# TODO: to_tsquery vs plainto_tsquery vs phraseto_tsquery
for fts_config, locales in locales_by_config.items():
conds[fts_config] = (or_(
*[((lse.locale == locale) | lse.locale.like(locale + "_%"))
for locale in locales]) if 'any' not in locales else None,
func.to_tsvector(fts_config, lse.value))
filter = [
cond & v.match(keywords_j, postgresql_regconfig=conf)
for (conf, (cond, v)) in conds.items() if cond is not None
]
if any_locale:
(_, v) = conds['simple']
filter.append(v.match(keywords_j, postgresql_regconfig='simple'))
query = query.filter(or_(*filter))
if include_rank:
if len(conds) > 1:
if any_locale:
(_, v) = conds['simple']
else_case = func.ts_rank(
v, func.to_tsquery('simple', keywords_j))
else:
else_case = 0
rank = case(
[(cond, func.ts_rank(v, func.to_tsquery(conf, keywords_j)))
for (conf,
(cond, v)) in conds.items() if cond is not None],
else_=else_case).label('score')
else:
(conf, (cond, v)) = next(iter(conds.items()))
rank = func.ts_rank(v,
func.to_tsquery(conf,
keywords_j)).label('score')
query = query.add_column(rank)
else:
fts_config = 'simple'
query = query.filter(
func.to_tsvector(fts_config,
lse.value).match(keywords_j,
postgresql_regconfig=fts_config))
if include_rank:
rank = func.ts_rank(func.to_tsvector(fts_config, lse.value),
func.to_tsquery(fts_config,
keywords_j)).label('score')
query = query.add_column(rank)
return query, rank
def do_search(search_query, offset, result_count, new_domains, framework_slug):
try:
sort_dir = list(search_query['sort'][0].values())[0]['order']
except (KeyError, IndexError):
sort_dir = 'asc'
try:
sort_by = list(search_query['sort'][0].values())[0]['sort_by']
except (KeyError, IndexError):
sort_by = None
try:
terms = search_query['query']['filtered']['filter']['terms']
except (KeyError, IndexError):
terms = {}
roles_list = None
seller_types_list = None
if terms:
new_domains = 'prices.serviceRole.role' not in terms
try:
if new_domains:
roles_list = terms['domains.assessed']
else:
roles = terms['prices.serviceRole.role']
roles_list = set(_['role'][7:] for _ in roles)
except KeyError:
pass
try:
seller_types_list = terms['seller_types']
except: # noqa
pass
try:
search_term = search_query['query']['match_phrase_prefix']['name']
except KeyError:
search_term = ''
EXCLUDE_LEGACY_ROLES = not current_app.config['LEGACY_ROLE_MAPPING']
if new_domains:
q = db.session.query(Supplier).outerjoin(SupplierDomain).outerjoin(Domain) \
.outerjoin(SupplierFramework).outerjoin(Framework)
else:
q = db.session.query(Supplier).outerjoin(PriceSchedule).outerjoin(ServiceRole) \
.outerjoin(SupplierFramework).outerjoin(Framework)
q = q.filter(
Supplier.status != 'deleted', Supplier.abn != Supplier.DUMMY_ABN,
or_(Framework.slug == framework_slug, ~Supplier.frameworks.any()))
tsquery = None
if search_term:
if any(c in search_term for c in ['#', '-', '_', '/', '\\']):
tsquery = func.phraseto_tsquery(search_term)
elif ' ' in search_term:
tsquery = func.plainto_tsquery(search_term)
else:
tsquery = func.to_tsquery(search_term + ":*")
q = q.add_column(
func.ts_headline(
'english',
func.concat(Supplier.summary, ' ',
Supplier.data['tools'].astext, ' ',
Supplier.data['methodologies'].astext, ' ',
Supplier.data['technologies'].astext, ''), tsquery,
'MaxWords=25, MinWords=20, ShortWord=3, HighlightAll=FALSE, MaxFragments=1'
))
q = q.group_by(Supplier.id)
try:
code = search_query['query']['term']['code']
q = q.filter(Supplier.code == code)
except KeyError:
pass
if roles_list is not None:
if new_domains:
if EXCLUDE_LEGACY_ROLES:
d_agg = postgres.array_agg(cast(Domain.name, TEXT))
q = q.filter(SupplierDomain.status == 'assessed')
q = q.having(d_agg.contains(array(roles_list)))
else:
sr_agg = postgres.array_agg(
cast(func.substring(ServiceRole.name, 8), TEXT))
q = q.having(sr_agg.contains(array(roles_list)))
if seller_types_list is not None and 'recruiter' in seller_types_list:
q = q.filter(Supplier.is_recruiter == 'true')
seller_types_list.remove('recruiter')
if len(seller_types_list) == 0:
seller_types_list = None
if seller_types_list is not None:
selected_seller_types = select(
[postgres.array_agg(column('key'))],
from_obj=func.json_each_text(Supplier.data[('seller_type', )]),
whereclause=cast(column('value'), Boolean)).as_scalar()
q = q.filter(selected_seller_types.contains(array(seller_types_list)))
if sort_by:
if sort_by == 'latest':
ob = [desc(Supplier.last_update_time)]
else:
ob = [asc(Supplier.name)]
else:
if sort_dir == 'desc':
ob = [desc(Supplier.name)]
else:
ob = [asc(Supplier.name)]
if search_term:
ob = [desc(func.ts_rank_cd(Supplier.text_vector, tsquery))] + ob
q = q.filter(Supplier.text_vector.op('@@')(tsquery))
q = q.order_by(*ob)
raw_results = list(q)
results = []
for x in range(len(raw_results)):
if type(raw_results[x]) is Supplier:
result = raw_results[x]
else:
result = raw_results[x][0]
if raw_results[x][1] is not None and raw_results[x][1] != '':
result.summary = raw_results[x][1]
results.append(result)
sliced_results = results[offset:(offset + result_count)]
q = db.session.query(Supplier.code, Supplier.name, Supplier.summary, Supplier.is_recruiter,
Supplier.data, Domain.name.label('domain_name'),
SupplierDomain.status.label('domain_status'))\
.outerjoin(SupplierDomain, Domain)\
.filter(Supplier.id.in_([sr.id for sr in sliced_results]))\
.order_by(Supplier.name)
suppliers = [r._asdict() for r in q]
sliced_results = []
for key, group in groupby(suppliers, key=itemgetter('code')):
supplier = group.next()
supplier['seller_type'] = supplier.get(
'data') and supplier['data'].get('seller_type')
supplier['domains'] = {'assessed': [], 'unassessed': []}
for s in chain([supplier], group):
domain, status = s['domain_name'], s['domain_status']
if domain:
if status == 'assessed':
supplier['domains']['assessed'].append(domain)
else:
supplier['domains']['unassessed'].append(domain)
for e in ['domain_name', 'domain_status', 'data']:
supplier.pop(e, None)
sliced_results.append(supplier)
return sliced_results, len(results)
def casestudies_search():
search_query = get_json_from_request()
offset = get_nonnegative_int_or_400(request.args, 'from', 0)
result_count = get_positive_int_or_400(
request.args, 'size', current_app.config['DM_API_SUPPLIERS_PAGE_SIZE'])
sort_dir = search_query.get('sort_dir', 'asc')
sort_by = search_query.get('sort_by', None)
domains = search_query.get('domains', None)
seller_types = search_query.get('seller_types', None)
search_term = search_query.get('search_term', None)
framework_slug = request.args.get('framework', 'digital-marketplace')
q = db.session.query(CaseStudy).join(Supplier).outerjoin(SupplierDomain).outerjoin(Domain) \
.outerjoin(SupplierFramework).outerjoin(Framework)
q = q.filter(
Supplier.status != 'deleted',
or_(Framework.slug == framework_slug, ~Supplier.frameworks.any()))
tsquery = None
if search_term:
if ' ' in search_term:
tsquery = func.plainto_tsquery(search_term)
else:
tsquery = func.to_tsquery(search_term + ":*")
q = q.add_column(
func.ts_headline(
'english',
func.concat(CaseStudy.data['approach'].astext, ' ',
CaseStudy.data['role'].astext), tsquery,
'MaxWords=150, MinWords=75, ShortWord=3, HighlightAll=FALSE, FragmentDelimiter=" ... " '
))
else:
q = q.add_column("''")
q = q.add_column(Supplier.name)
q = q.add_column(postgres.array_agg(Supplier.data))
q = q.group_by(CaseStudy.id, Supplier.name)
if domains:
d_agg = postgres.array_agg(cast(Domain.name, TEXT))
q = q.having(d_agg.contains(array(domains)))
if seller_types:
selected_seller_types = select(
[postgres.array_agg(column('key'))],
from_obj=func.json_each_text(Supplier.data[('seller_type', )]),
whereclause=cast(column('value'), Boolean)).as_scalar()
q = q.filter(selected_seller_types.contains(array(seller_types)))
if sort_dir in ('desc', 'z-a'):
ob = [desc(CaseStudy.data['title'].astext)]
else:
ob = [asc(CaseStudy.data['title'].astext)]
if search_term:
ob = [
desc(
func.ts_rank_cd(
func.to_tsvector(
func.concat(
Supplier.name, CaseStudy.data['title'].astext,
CaseStudy.data['approach'].astext)), tsquery))
] + ob
condition = func.to_tsvector(
func.concat(Supplier.name, CaseStudy.data['title'].astext,
CaseStudy.data['approach'].astext)).op('@@')(tsquery)
q = q.filter(condition)
q = q.order_by(*ob)
raw_results = list(q)
results = []
for x in range(len(raw_results)):
result = raw_results[x][0].serialize()
if raw_results[x][1] is not None and raw_results[x][1] != '':
result['approach'] = raw_results[x][1]
if raw_results[x][2] is not None:
result['supplierName'] = raw_results[x][2]
if raw_results[x][3] is not None and raw_results[x][3][0] is not None:
result['seller_type'] = raw_results[x][3][0].get('seller_type')
results.append(result)
total_results = len(results)
sliced_results = results[offset:(offset + result_count)]
result = {
'hits': {
'total': total_results,
'hits': [{
'_source': r
} for r in sliced_results]
}
}
try:
response = jsonify(result), 200
except Exception as e:
response = jsonify(message=str(e)), 500
return response
def do_search(search_query, offset, result_count, new_domains, framework_slug):
try:
sort_dir = list(search_query['sort'][0].values())[0]['order']
except (KeyError, IndexError):
sort_dir = 'asc'
try:
sort_by = list(search_query['sort'][0].values())[0]['sort_by']
except (KeyError, IndexError):
sort_by = None
try:
terms = search_query['query']['filtered']['filter']['terms']
except (KeyError, IndexError):
terms = {}
roles_list = None
seller_types_list = None
if terms:
new_domains = 'prices.serviceRole.role' not in terms
try:
if new_domains:
roles_list = terms['domains.assessed']
else:
roles = terms['prices.serviceRole.role']
roles_list = set(_['role'][7:] for _ in roles)
except KeyError:
pass
try:
seller_types_list = terms['seller_types']
except: # noqa
pass
try:
search_term = search_query['query']['match_phrase_prefix']['name']
except KeyError:
search_term = ''
EXCLUDE_LEGACY_ROLES = not current_app.config['LEGACY_ROLE_MAPPING']
if new_domains:
q = db.session.query(Supplier).outerjoin(SupplierDomain).outerjoin(Domain) \
.outerjoin(SupplierFramework).outerjoin(Framework)
else:
q = db.session.query(Supplier).outerjoin(PriceSchedule).outerjoin(ServiceRole) \
.outerjoin(SupplierFramework).outerjoin(Framework)
q = q.filter(Supplier.status != 'deleted', Supplier.abn != Supplier.DUMMY_ABN,
or_(Framework.slug == framework_slug, ~Supplier.frameworks.any()))
tsquery = None
if search_term:
if any(c in search_term for c in ['#', '-', '_', '/', '\\']):
tsquery = func.phraseto_tsquery(search_term)
elif ' ' in search_term:
tsquery = func.plainto_tsquery(search_term)
else:
tsquery = func.to_tsquery(search_term + ":*")
q = q.add_column(func.ts_headline(
'english',
func.concat(Supplier.summary,
' ',
Supplier.data['tools'].astext,
' ',
Supplier.data['methodologies'].astext,
' ',
Supplier.data['technologies'].astext, ''),
tsquery,
'MaxWords=25, MinWords=20, ShortWord=3, HighlightAll=FALSE, MaxFragments=1'
))
q = q.group_by(Supplier.id)
try:
code = search_query['query']['term']['code']
q = q.filter(Supplier.code == code)
except KeyError:
pass
if roles_list is not None:
if new_domains:
if EXCLUDE_LEGACY_ROLES:
d_agg = postgres.array_agg(cast(Domain.name, TEXT))
q = q.having(d_agg.contains(array(roles_list)))
else:
sr_agg = postgres.array_agg(cast(func.substring(ServiceRole.name, 8), TEXT))
q = q.having(sr_agg.contains(array(roles_list)))
if seller_types_list is not None and 'recruiter' in seller_types_list:
q = q.filter(Supplier.is_recruiter == 'true')
seller_types_list.remove('recruiter')
if len(seller_types_list) == 0:
seller_types_list = None
if seller_types_list is not None:
selected_seller_types = select(
[postgres.array_agg(column('key'))],
from_obj=func.json_each_text(Supplier.data[('seller_type',)]),
whereclause=cast(column('value'), Boolean)
).as_scalar()
q = q.filter(selected_seller_types.contains(array(seller_types_list)))
if sort_by:
if sort_by == 'latest':
ob = [desc(Supplier.last_update_time)]
else:
ob = [asc(Supplier.name)]
else:
if sort_dir == 'desc':
ob = [desc(Supplier.name)]
else:
ob = [asc(Supplier.name)]
if search_term:
ob = [desc(func.ts_rank_cd(Supplier.text_vector, tsquery))] + ob
q = q.filter(Supplier.text_vector.op('@@')(tsquery))
q = q.order_by(*ob)
raw_results = list(q)
results = []
for x in range(len(raw_results)):
if type(raw_results[x]) is Supplier:
result = raw_results[x]
else:
result = raw_results[x][0]
if raw_results[x][1] is not None and raw_results[x][1] != '':
result.summary = raw_results[x][1]
results.append(result)
sliced_results = results[offset:(offset + result_count)]
q = db.session.query(Supplier.code, Supplier.name, Supplier.summary, Supplier.is_recruiter,
Supplier.data, Domain.name.label('domain_name'),
SupplierDomain.status.label('domain_status'))\
.outerjoin(SupplierDomain, Domain)\
.filter(Supplier.id.in_([sr.id for sr in sliced_results]))\
.order_by(Supplier.name)
suppliers = [r._asdict() for r in q]
sliced_results = []
for key, group in groupby(suppliers, key=itemgetter('code')):
supplier = group.next()
supplier['seller_type'] = supplier.get('data') and supplier['data'].get('seller_type')
supplier['domains'] = {'assessed': [], 'unassessed': []}
for s in chain([supplier], group):
domain, status = s['domain_name'], s['domain_status']
if domain:
if status == 'assessed':
supplier['domains']['assessed'].append(domain)
else:
supplier['domains']['unassessed'].append(domain)
for e in ['domain_name', 'domain_status', 'data']:
supplier.pop(e, None)
sliced_results.append(supplier)
return sliced_results, len(results)
def casestudies_search():
search_query = get_json_from_request()
offset = get_nonnegative_int_or_400(request.args, 'from', 0)
result_count = get_positive_int_or_400(request.args, 'size', current_app.config['DM_API_SUPPLIERS_PAGE_SIZE'])
sort_dir = search_query.get('sort_dir', 'asc')
sort_by = search_query.get('sort_by', None)
domains = search_query.get('domains', None)
seller_types = search_query.get('seller_types', None)
search_term = search_query.get('search_term', None)
framework_slug = request.args.get('framework', 'digital-marketplace')
q = db.session.query(CaseStudy).join(Supplier).outerjoin(SupplierDomain).outerjoin(Domain) \
.outerjoin(SupplierFramework).outerjoin(Framework)
q = q.filter(Supplier.status != 'deleted', or_(Framework.slug == framework_slug, ~Supplier.frameworks.any()))
tsquery = None
if search_term:
if ' ' in search_term:
tsquery = func.plainto_tsquery(search_term)
else:
tsquery = func.to_tsquery(search_term + ":*")
q = q.add_column(func.ts_headline(
'english',
func.concat(
CaseStudy.data['approach'].astext,
' ',
CaseStudy.data['role'].astext),
tsquery,
'MaxWords=150, MinWords=75, ShortWord=3, HighlightAll=FALSE, FragmentDelimiter=" ... " '
))
else:
q = q.add_column("''")
q = q.add_column(Supplier.name)
q = q.add_column(postgres.array_agg(Supplier.data))
q = q.group_by(CaseStudy.id, Supplier.name)
if domains:
d_agg = postgres.array_agg(cast(Domain.name, TEXT))
q = q.having(d_agg.contains(array(domains)))
if seller_types:
selected_seller_types = select(
[postgres.array_agg(column('key'))],
from_obj=func.json_each_text(Supplier.data[('seller_type',)]),
whereclause=cast(column('value'), Boolean)
).as_scalar()
q = q.filter(selected_seller_types.contains(array(seller_types)))
if sort_dir in ('desc', 'z-a'):
ob = [desc(CaseStudy.data['title'].astext)]
else:
ob = [asc(CaseStudy.data['title'].astext)]
if search_term:
ob = [desc(func.ts_rank_cd(func.to_tsvector(
func.concat(Supplier.name, CaseStudy.data['title'].astext,
CaseStudy.data['approach'].astext)), tsquery))] + ob
condition = func.to_tsvector(func.concat(Supplier.name,
CaseStudy.data['title'].astext,
CaseStudy.data['approach'].astext)).op('@@')(tsquery)
q = q.filter(condition)
q = q.order_by(*ob)
raw_results = list(q)
results = []
for x in range(len(raw_results)):
result = raw_results[x][0].serialize()
if raw_results[x][1] is not None and raw_results[x][1] != '':
result['approach'] = raw_results[x][1]
if raw_results[x][2] is not None:
result['supplierName'] = raw_results[x][2]
if raw_results[x][3] is not None and raw_results[x][3][0] is not None:
result['seller_type'] = raw_results[x][3][0].get('seller_type')
results.append(result)
total_results = len(results)
sliced_results = results[offset:(offset + result_count)]
result = {
'hits': {
'total': total_results,
'hits': [{'_source': r} for r in sliced_results]
}
}
try:
response = jsonify(result), 200
except Exception as e:
response = jsonify(message=str(e)), 500
return response
