Query parsers

eDisMax lets a field alias stand in for several real fields: f.<alias>.qf=<fields> (and the equivalent for pf) makes <alias>:value in q expand to a search across <fields>, which is handy for exposing a friendly name like text over several concrete columns without a copy field. uf (user fields) is the allow/deny list of field names an end user is permitted to search explicitly inside q (title:foo) — it defaults to uf=* -query (every field except the raw sub-query escape hatch); set it to something narrower, or to uf=-*, when q comes from untrusted input and fielded search should be disabled entirely.

\{!frange} turns a function query into a range filter, keeping only documents whose function value falls within l (lower bound) / u (upper bound), with incl / incu (both default true) controlling whether each bound is inclusive. \{!func} instead runs the function purely as the query value/score, without the surrounding function-query braces val requires. Both are the idiomatic way to filter or sort on a computed value — distance, a boosted combination of fields, age of a document — rather than a stored one.

\{!term f=<field>}<value> builds a single exact-term query without running <value> through the field’s analyzer — the counterpart to a term filter in other search engines, and the idiom for turning one facet value straight into a filter query. \{!terms f=<field>} takes the same idea for a list, matching any document whose field equals one of several comma-separated (or separator- delimited) values, with a method local param (termsFilter, booleanQuery, automaton, docValuesTermsFilter, …​) to pick the underlying implementation for very large lists.

Solr has no native SQL-style join across collections, but two parsers cover the common cases without denormalizing data into a single flat document. \{!join from=<field> to=<field>} matches documents whose to field equals the from field of any document matched by the subordinate query — an "IN (subquery)" over two independently-indexed sets of documents, optionally against another core or collection via fromIndex, with method (index, dvWithScore, topLevelDV, or crossCollection) choosing the join strategy for cardinality and scoring needs. See Join Query Parser.

Block-join is the alternative for documents that are genuinely parent/child (a product and its variants, an invoice and its line items) indexed together as one nested block via nested/child documents. \{!parent which=<parentFilter>} takes a query matched against children and returns their parents; \{!child of=<parentFilter>} runs the other direction, returning children of matched parents. Both accept filters to apply additional fq-style filters only within the block-join evaluation, and \{!parent} accepts score (none/avg/max/min/total) to aggregate child scores onto the parent. See Block Join Query Parser.

\{!collapse field=<field>} groups the result set by field, keeping one representative document per group — the same outcome as field collapsing/grouping, but implemented as a post- filter that composes with faceting and is generally the faster choice for a single collapse field. It takes min / max (a function or field deciding which group member survives) and sort (an explicit sort for the representative).

\{!knn f=<denseVectorField> topK=<k>}[v1,v2,…​] finds the topK documents whose stored dense vector is closest to the query vector, using the field’s configured similarity function (cosine/dot-product/Euclidean). Because it wraps a vector search rather than a term lookup, it is typically combined with a filter (fq) or with \{!knn …​} inside a bq/boost clause rather than run alone. Full coverage, including indexing dense-vector fields, is in Dense vector search.

A handful of narrower parsers round out the set:

See Other Parsers for the complete parameter reference of each, including the less common \{!field} and \{!complexphrase} parsers.