Separate search for unique identifiers out from text search

This commit is contained in:
Mouse Reeve 2021-01-02 15:48:59 -08:00
parent a2e8cf1993
commit a413c87963

View file

@ -1,6 +1,9 @@
''' using a bookwyrm instance as a source of book data '''
from functools import reduce
import operator
from django.contrib.postgres.search import SearchRank, SearchVector
from django.db.models import Count, F
from django.db.models import Count, F, Q
from bookwyrm import models
from .abstract_connector import AbstractConnector, SearchResult
@ -10,36 +13,14 @@ class Connector(AbstractConnector):
''' instantiate a connector '''
def search(self, query, min_confidence=0.1):
''' search your local database '''
vector = SearchVector('title', weight='A') +\
SearchVector('subtitle', weight='B') +\
SearchVector('authors__name', weight='C')
results = models.Edition.objects.annotate(
search=vector
).annotate(
rank=SearchRank(vector, query)
).filter(
rank__gt=min_confidence
).order_by('-rank')
# when there are multiple editions of the same work, pick the closest
editions_of_work = results.values(
'parent_work'
).annotate(
Count('parent_work')
).values_list('parent_work')
# first, try searching unqiue identifiers
results = search_identifiers(query)
if not results:
# then try searching title/author
results = search_title_author(query, min_confidence)
search_results = []
for work_id in set(editions_of_work):
editions = results.filter(parent_work=work_id)
default = editions.filter(parent_work__default_edition=F('id'))
default_rank = default.first().rank if default.exists() else 0
# if mutliple books have the top rank, pick the default edition
if default_rank == editions.first().rank:
selected = default.first()
else:
selected = editions.first()
search_results.append(self.format_search_result(selected))
for result in results:
search_results.append(self.format_search_result(result))
if len(search_results) >= 10:
break
return search_results
@ -53,7 +34,8 @@ class Connector(AbstractConnector):
year=search_result.published_date.year if \
search_result.published_date else None,
connector=self,
confidence=search_result.rank,
confidence=search_result.rank if \
hasattr(search_result, 'rank') else 1,
)
@ -75,3 +57,50 @@ class Connector(AbstractConnector):
def expand_book_data(self, book):
pass
def search_identifiers(query):
''' tries remote_id, isbn; defined as dedupe fields on the model '''
filters = [{f.name: query} for f in models.Edition._meta.get_fields() \
if hasattr(f, 'deduplication_field') and f.deduplication_field]
results = models.Edition.objects.filter(
reduce(operator.or_, (Q(**f) for f in filters))
).distinct()
# when there are multiple editions of the same work, pick the default.
# it would be odd for this to happen.
return results.filter(parent_work__default_edition__id=F('id')) \
or results
def search_title_author(query, min_confidence):
''' searches for title and author '''
print('DON"T BOTHER')
vector = SearchVector('title', weight='A') +\
SearchVector('subtitle', weight='B') +\
SearchVector('authors__name', weight='C')
results = models.Edition.objects.annotate(
search=vector
).annotate(
rank=SearchRank(vector, query)
).filter(
rank__gt=min_confidence
).order_by('-rank')
# when there are multiple editions of the same work, pick the closest
editions_of_work = results.values(
'parent_work'
).annotate(
Count('parent_work')
).values_list('parent_work')
for work_id in set(editions_of_work):
editions = results.filter(parent_work=work_id)
default = editions.filter(parent_work__default_edition=F('id'))
default_rank = default.first().rank if default.exists() else 0
# if mutliple books have the top rank, pick the default edition
if default_rank == editions.first().rank:
yield default.first()
else:
yield editions.first()