Commit Graph

29 Commits

Author SHA1 Message Date
Sarah Hoffmann
4a00a3c0f5 penalize name token splitting when phrases are used 2023-07-17 16:27:25 +02:00
Sarah Hoffmann
8366e4ca83 penalize search with frequent partials
Avoid search against frequent partials if we have already looked for
the full name equivalents.
2023-07-17 16:27:25 +02:00
Sarah Hoffmann
283db76e45 avoid splitting of first token when a housenumber is present
This only covers the case of <poi name> <street name> <housenumber>
which is exceedingly rare.
2023-07-17 16:27:25 +02:00
Sarah Hoffmann
8a36ed4f6f increase threshold for full name searches
They still should be preferrred over expensive partial name searches.
2023-07-17 16:27:25 +02:00
Sarah Hoffmann
d0f45155c8 fix search for housenumber names
The search still included a lookup of housenumbers in children which is
wrong.
2023-07-17 16:27:25 +02:00
Sarah Hoffmann
7932b1849b selected lambdas for search 2023-07-14 15:43:29 +02:00
Sarah Hoffmann
cc45930ef9 avoid lookup via partials on frequent words
Drops expensive searches via partials on terms like 'rue de'.

See #2979.
2023-07-06 12:16:57 +02:00
Sarah Hoffmann
3266daa8fd add a small penalty to lookups in address vectors 2023-07-04 16:54:42 +02:00
Sarah Hoffmann
49e0d83d5d fix linting issues 2023-07-01 20:18:59 +02:00
Sarah Hoffmann
673c3c7a55 replace regexp_match with generic op() functions
Works around a bug in SQLAlchemy where regexp_match creates an
unstable cache key.
2023-07-01 18:15:22 +02:00
Sarah Hoffmann
5135041405 replace CASE construct with plpgsql function 2023-07-01 18:15:22 +02:00
Sarah Hoffmann
9f6f12cfeb move search to bind parameters 2023-07-01 18:03:07 +02:00
Sarah Hoffmann
3a21999a17 move text normalization into extra function 2023-06-22 10:48:05 +02:00
Sarah Hoffmann
9bc5be837b remove useless check
Found by new mypy version.
2023-06-21 11:56:39 +02:00
Sarah Hoffmann
4ad8818809 avoid fallback country lookup when places are excluded 2023-06-20 12:22:08 +02:00
Sarah Hoffmann
d0a1e8e311 tweak postcode search
Give a preference to left-right reading, i.e <postcode>,<address>
prefers a postcode search while <address>,<postcode> rather does
an address search.

Also exclude non-addressables, countries and state from results when a
postcode is contained in the query.
2023-06-20 11:56:43 +02:00
Sarah Hoffmann
3443d2c129 fix debug output for searches
Fields have been renamed.
2023-06-20 10:12:19 +02:00
Sarah Hoffmann
317cc5c544 make code backwards-compatible with older Python versions 2023-05-26 23:33:35 +02:00
Sarah Hoffmann
43c27dffd2 fix new linting warning 2023-05-26 16:31:43 +02:00
Sarah Hoffmann
146a0b29c0 add support for search by houenumber 2023-05-26 14:10:57 +02:00
Sarah Hoffmann
964bc7fbe0 ignore name rankings without variants 2023-05-26 13:53:34 +02:00
Sarah Hoffmann
dcfb228c9a add API functions for search functions
Search is now split into three functions: for free-text search,
for structured search and for search by category. Note that the
free-text search does not have as many hidden features like
coordinate search. Use the search parameters for that.
2023-05-24 18:05:43 +02:00
Sarah Hoffmann
dc99bbb0af implement actual database searches 2023-05-24 13:52:31 +02:00
Sarah Hoffmann
c42273a4db implement search builder 2023-05-23 11:23:44 +02:00
Sarah Hoffmann
3bf489cd7c implement token assignment 2023-05-22 15:49:03 +02:00
Sarah Hoffmann
d8240f9ee4 add query analyser for legacy tokenizer 2023-05-22 11:07:14 +02:00
Sarah Hoffmann
2448cf2a14 add factory for query analyzer 2023-05-22 09:23:19 +02:00
Sarah Hoffmann
004883bdb1 query analyzer for ICU tokenizer 2023-05-22 08:46:19 +02:00
Sarah Hoffmann
ff66595f7a add data structure for tokenized query 2023-05-21 09:30:57 +02:00