inspirehep / inspire-query-parser / 217

96%

Pull #80

travis-ci

es-visitor: rewrite date queries

Rewrite of date queries (closes #73):
* Converts mutli-field range queries to a chain of bool.should clauses
  so that if a field doesn't exist in a record, the query will still
  produce results. Same processing happens for date range queries.

* Date queries with equality and Value or Partial value are transformed
  into range queries, starting from the given date and up to the next
  date (according to the granularity of the date typed by the user).

* Uses PartialDate from inspire-utils, so that it drops all given date
  parts except for the year, for fields that only contain a year
  (otherwise the query in ES errors out).

* Add simple support for wildcard operator in simple and partial match
  value date queries. The policy followed is:
  * if the whole date part is a wildcard, then we ignore it and do a
    range query on the remaining parts (e.g. '2018-*' -> '2018')
  * Some number is wildcard, where again, we ignore this part. If all
    parts are ignored then we drop the date query.
  * For queries where the wildcard operator is not as a suffix (e.g.
    '2017-1*-01'), we drop the date query.

* es-visitor: date queries use ES date rounding
  This feature is used in all cases, i.e. range and equality queries.
  This is needed for supporting range queries on dates, i.e. rounding
  them up or down according to the ES range operator.

  For example, without this, a query like 'date > 2010-11', would return
  documents with date '2010-11-15', due to the date value of the query
  being interpreted by ES as '2010-11-01 01:00:00'.

  By using the suffixes for rounding up or down, the date value of the
  query is interpreted as '2010-11-30T23:59:59.999', thus not returning
  the document with date '2010-11-15', as the user would expect. See:
  https://www.elastic.co/guide/en/elasticsearch/reference/6.1/query-dsl-range-query.html#_date_math_and_rounding

* In case of malformed date the related range query is dropped.
  Consequently, ``visit_boolean_rule`` in case of both conditions being
  empty, generates an empty query (equivalent to match_all).

* parsing_driver: generate query on empty query from ES visitor
  Ensure that a query is generated at all times, if an empty query "{}"
  is returned from ES visitor, otherwise ElasticSearch DSL breaks when
  trying to wrap parser's return value in a Q object.

Signed-off-by: Chris Aslanoglou <chris.aslanoglou@gmail.com>

100 of 100 new or added lines in 3 files covered. (100.0%)

863 of 899 relevant lines covered (96.0%)

1.92 hits per line