postgresql - Django psql full text search not matching un-stemmed word -

i'm running django 1.10.1 against postgres 9.4. staging server , dev environments have psql servers @ version 9.4.9 , production rds instance @ 9.4.7.

it seems searchvectorfield not storing search configuration given in production, though in staging , dev, , seems either version thing (unlikely, given version difference , worked on 9.3 in staging/dev) or fact production on rds instead of local on server.

i'm using custom configuration full-text search called unaccent, looks this:

      token      |     dictionaries       -----------------+-----------------------  asciihword      | english_stem  asciiword       | english_stem  email           | simple  file            | simple  float           | simple  host            | simple  hword           | unaccent,english_stem  hword_asciipart | english_stem  hword_numpart   | simple  hword_part      | unaccent,english_stem  int             | simple  numhword        | simple  numword         | simple  sfloat          | simple  uint            | simple  url             | simple  url_path        | simple  version         | simple  word            | unaccent,english_stem 

unaccent installed in both environments, , works in both environments.

i'm storing search data in django.contrib.postgres.search.searchvectorfield on writer model:

class writer(models.model):     #...     search = searchvectorfield(blank=true) 

that column updated following search vector:

writer_search_vector = (searchvector('first_name', 'last_name', 'display_name',                                      config='unaccent', weight='a') +                         searchvector('raw_search_data', config='unaccent', weight='b')) 

by following statement, runs periodically:

writer.objects.update(search=search_utils.writer_search_vector) 

and, reason, configuration storing on staging server , in dev, not in production. e.g., code returns same results in environments:

in [3]: writer.objects.annotate(searchy=searchvector('last_name')).filter(searchy='kostenberger') out[3]: <queryset []> in [4]: writer.objects.annotate(searchy=searchvector('last_name', config='unaccent')).filter(searchy='kostenberger') out[4]: <queryset [<writer: andreas j. köstenberger>, <writer: margaret elizabeth köstenberger>]> 

but in staging, following correct result if use stored vector:

in [5]: writer.objects.filter(search='kostenberger') out[5]: <queryset [<writer: andreas j. köstenberger>, <writer: margaret elizabeth köstenberger>]> 

while in production, against rds instance, following, incorrect result:

in [5]: writer.objects.filter(search='kostenberger') out[5]: <queryset []> 

and yet, in production still, unaccent works english_stem not, in match stemmed version of text (below), not original version (above):

in [6]: writer.objects.filter(search='kostenberg') out[6]: <queryset [<writer: margaret elizabeth köstenberger>, <writer: andreas j. köstenberger>]> 

note database tables writer in 2 environments identical test.

any ideas why stored vector isn't working in production correct config, while if create vector on fly work?

on rds postgres, aren't allowed change default_text_search_config parameter. so, have configure text search each query:

from django.contrib.postgres.search import searchrank, searchquery … search_query = searchquery(value='kostenberger', config='unaccent') writer.objects.filter(search=search_query)