التصفية والنطاقات وسياق الاستعلام

Every clause in Elasticsearch runs in one of two contexts: query context (How well does this match? produces a score) or filter context (Does this match? yes/no, no score).

Filters do not compute relevance scores and their results are cached, so repeated filters are very cheap. Use filters whenever you do not need ranking from that clause.

Inside a bool query, the filter array runs in filter context while must runs in query context.

The range query selects values within bounds using gte, gt, lte, lt. It is commonly placed in filter context.

Range works on dates too, including date math like now-7d for relative windows.

The terms query matches any of several exact values, like a SQL IN list.

The exists query keeps only documents that have a non-null value for a field.

Stack multiple filters in the filter array; they are ANDed together and all benefit from caching.

Use must_not inside bool to exclude documents. It also runs in filter context.

When you rely on filters alone, every document scores the same, so add an explicit sort (for example by price or date) to order results.

Put the scoring text match in must and push every yes/no condition into filter. This keeps scoring focused and lets caching speed up the rest.

When should a clause go in filter context?

You learned the query vs filter context distinction, used range, terms, and exists filters inside bool, applied must_not for exclusion, and added explicit sorting when relying on filters.