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 tag_search(text, count=5):
"""Search in the tag_search_items table (for location tags)."""
QTag = aliased(LocationTag)
QParent = aliased(LocationTag)
text = text.lower().strip()
query = meta.Session.query(TagSearchItem)\
.join(QTag)\
.outerjoin((QParent, QParent.id==QTag.parent_id))\
.filter(TagSearchItem.terms.op('@@')(func.plainto_tsquery(text)))\
.order_by(or_(func.lower(func.btrim(QParent.title)) == text, func.lower(func.btrim(QParent.title_short)) == text).desc())\
.order_by(or_(func.lower(func.btrim(QTag.title)) == text, func.lower(func.btrim(QTag.title_short)) == text).desc())\
.order_by(func.ts_rank_cd(TagSearchItem.terms, func.plainto_tsquery(text)))
if count is not None:
query = query.limit(count)
return query.all()
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 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 _gen_search_elements(query,
title_only,
next_rank,
page_size,
A,
B=[],
C=[],
D=[]):
"""
Given a query and four sets of fields, (A, B, C, D), generates a bunch of postgres expressions for full text search.
The four sets are in decreasing order of "importance" for ranking.
A should be the "title", the others can be anything.
If title_only=True, we only perform a trigram search against A only
"""
if not title_only:
# a postgres tsquery object that can be used to match against a tsvector
tsq = func.websearch_to_tsquery(REGCONFIG, query)
# the tsvector object that we want to search against with our tsquery
tsv = _build_tsv(A, B, C, D)
# document to generate snippet from
doc = _build_doc(A, B, C, D)
title = _build_doc(A)
# trigram based text similarity between title and query string
sim = _similarity(query, title)
# ranking algo, weigh the similarity a lot, the text-based ranking less
rank = (TRI_SIMILARITY_WEIGHT * sim +
func.ts_rank_cd(tsv, tsq)).label("rank")
# the snippet with results highlighted
snippet = func.ts_headline(REGCONFIG, doc, tsq,
"StartSel=**,StopSel=**").label("snippet")
def do_search_query(orig_query):
"""
Does the right search filtering, limiting, and ordering for the query
"""
return (orig_query.filter(
or_(tsv.op("@@")(tsq), sim > TRI_SIMILARITY_THRESHOLD)).filter(
rank <= next_rank if next_rank is not None else True).
order_by(rank.desc()).limit(page_size + 1).all())
else:
title = _build_doc(A)
# trigram based text similarity between title and query string
sim = _similarity(query, title)
# ranking algo, weigh the similarity a lot, the text-based ranking less
rank = sim.label("rank")
# used only for headline
tsq = func.websearch_to_tsquery(REGCONFIG, query)
doc = _build_doc(A, B, C, D)
# the snippet with results highlighted
snippet = func.ts_headline(REGCONFIG, doc, tsq,
"StartSel=**,StopSel=**").label("snippet")
def do_search_query(orig_query):
"""
Does the right search filtering, limiting, and ordering for the query
"""
return (orig_query.filter(sim > TRI_SIMILARITY_THRESHOLD).filter(
rank <= next_rank if next_rank is not None else True).order_by(
rank.desc()).limit(page_size + 1).all())
return rank, snippet, do_search_query
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
