Skip to content

Compound queries

Compound queries1 wrap other queries to either:

  • Combine their results
  • Change their behavior
  • Switch query contexts (in particular, from query to filter context)

In other words, you can use compound queries to combine any number of term-level and full-text queries (and in fact other compound queries as well).

Details

Below are the various kinds of compound queries. These are sorted with the most commonly used at the top, and cover their usual usage. Each one is linked to Elasticsearch's own documentation to provide greater details. (In most cases there are many more options for each kind of query than what is documented here.)

Bool

Bool queries combine multiple queries using must, should, must_not and filter clauses. These allow you to combine queries with logic like AND, OR and NOT.

Clause Description Context
must Query clauses must match the results, and will contribute to the score. These act like a logical AND operation. query
should Query clauses should match the results, and will contribute to the score. These act like a logical OR operation. query
must_not Query clauses must not match the results, and are used to either include or exclude results (no scoring). These act like a logical NOT operation. filter
filter Query clauses must match the results, but will not contribute to the score. filter

1.0.0 1.1.0

For example, this query would find all active (non-archived) tables with either a classification or a term assigned:

Build the query and request
1
2
3
4
5
IndexSearchRequest index = Table.select() // (1)
    .whereSome(CompoundQuery.tagged(true)) // (2)
    .whereSome(CompoundQuery.assignedTerm())
    .minSomes(1) // (3)
    .toRequest(); // (4)
  1. You can build up a compound query progressively, starting from the type of asset you want to query using the select() method. This will start a query that narrows results to only active assets of this type (Table in this example).
  2. You can add any number of conditions where some of them must match using the whereSome() helper with a query as a condition. (You can add any number of mandatory conditions using the where() helper with a query as a condition, instead.) Each query you provide can either be from a helper (like these examples) or a full-fledged Elastic Query, if you need ultimate flexibility.
  3. You can specify how many of these whereSome() conditions must match using the minSomes() helper.
  4. Finally, you can build the compound query into a search request using the toRequest() helper method.
Build the query and request
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
from pyatlan.model.fluent_search import CompoundQuery, FluentSearch
from pyatlan.model.assets import Table

index = (FluentSearch()  # (1)
     .where(CompoundQuery.asset_type(Table))  # (2)
     .where(CompoundQuery.active_assets())
     .where_some(CompoundQuery.tagged(directly=True))  # (3)
     .where_some(CompoundQuery.assigned_term())
     .min_somes(1)  # (4)
    ).to_request()  # (5)
  1. You can build up a compound query progressively by creating a FluentSearch() object and chaining conditions onto it.
  2. You can add any number of mandatory conditions using the where() helper with a query as a condition. You can use query helpers (like these examples from CompoundQuery to narrow to assets of a particular type (Table) and only active assets), or full-fledged Elastic Query's.
  3. You can add any number of conditions where some of them must match using the where_some() helper with a query as a condition.
  4. You can specify how many of these where_some() conditions must match using the min_somes() helper.
  5. Finally, you can build the compound query into a search request using the to_request() helper method.
POST /api/meta/search/indexsearch
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
{
  "dsl": {
    "query": {
      "bool": {
        "must": {
          "term": { "__state": { "value": "ACTIVE" }}
        },
        "filter": {
          "term": { "__typeName.keyword": { "value": "Table" }}
        },
        "should": [
          "exists": { "field": "__traitNames" },
          "exists": { "field": "__meanings" }
        ],
        "minimum_should_match": 1
      }
    }
  }
}

In the vast majority of cases you will use bool queries, which the SDK examples above create. There are other compound query options for influencing scores, but details for these are left to the Elasticsearch documentation linked below.

  1. This page is a summary of the details in the Elasticsearch Guide's Compound queries  ↩