widget

[wolnelektury.git] / src / search / utils.py

diff --git a/src/search/utils.py b/src/search/utils.py
index b2cbe8d..77ff1ae 100644 (file)
--- a/src/search/utils.py
+++ b/src/search/utils.py
@@ -1,79 +1,31 @@
+from django.conf import settings
 from django.db.models import Func
-from django.contrib.postgres.search import SearchVector, SearchQuery, SearchQueryField, SearchHeadline as SH
-
-
-
-class UnaccentTSVector(Func):
-    function = 'UNACCENT'
-    template = '%(function)s(%(expressions)s::text)::tsvector'
-
-
-class Unaccent(Func):
-    function = 'UNACCENT'
-
-    
-class ConcatTSVector(Func):
-    function = 'CONCAT'
-    template = '%(function)s(%(expressions)s)::tsvector'    
-
-
-class UnaccentTSQuery(Func):
-    function = 'UNACCENT'
-    template = '%(function)s(%(expressions)s::text)::tsquery'
-    output_field = SearchQueryField()
-
-
-class TSV(Func):
+from django.contrib.postgres.search import SearchQuery, SearchVectorField
+
+
+class UnaccentSearchQuery(SearchQuery):
+    '''
+    The idea is to run unaccent *after* the query is already passed through the language dictionary.
+    '''
+    def as_sql(self, *args, **kwargs):
+        sql, params = super().as_sql(*args, **kwargs)
+        if settings.SEARCH_USE_UNACCENT:
+            sql = f'unaccent({sql}::text)::tsquery'
+        return sql, params
+
+
+class UnaccentSearchVector(Func):
+    '''
+    We do the indexing twice, to account for non-diacritic versions.
+    For example: user enters 'róże' -> stem to 'róża' -> unaccent to 'roza'.
+    But user enters 'roze' -> stem leaves it as is, so we need original form in the vector.
+    '''
     function='to_tsvector'
-    template = '''unaccent(
-      %(function)s('polish', %(expressions)s)::text)::tsvector ||
-     to_tsvector(
-       'polish_simple', 
-       unaccent(%(expressions)s)
-     )'''
-
-
-def build_search_vector(*fields):
-    return TSV(*fields)
-
-
-def build_search_query(*fields, **kwargs):
-    return UnaccentTSQuery(SearchQuery(*fields, **kwargs))
-
-
-
-class SearchHeadline(SH):
-
-    def __init__(
-        self,
-        expression,
-        query,
-        *,
-        config=None,
-        start_sel=None,
-        stop_sel=None,
-        max_words=None,
-        min_words=None,
-        short_word=None,
-        highlight_all=None,
-        max_fragments=None,
-        fragment_delimiter=None,
-    ):
-        options = {
-            "StartSel": start_sel,
-            "StopSel": stop_sel,
-            "MaxWords": max_words,
-            "MinWords": min_words,
-            "ShortWord": short_word,
-            "HighlightAll": highlight_all,
-            "MaxFragments": max_fragments,
-            "FragmentDelimiter": fragment_delimiter,
-        }
-        self.options = {
-            option: value for option, value in options.items() if value is not None
-        }
-        expressions = (expression, query)
-        if config is not None:
-            config = SearchConfig.from_parameter(config)
-            expressions = (config,) + expressions
-        Func.__init__(self, *expressions)
+    if settings.SEARCH_USE_UNACCENT:
+        template = f'''unaccent(
+        %(function)s('{settings.SEARCH_CONFIG}', %(expressions)s)::text)::tsvector ||
+        to_tsvector(
+        '{settings.SEARCH_CONFIG_SIMPLE}', 
+        unaccent(%(expressions)s)
+        )'''
+    output_field = SearchVectorField()