Wraps the ArangoDB AQL function ANALYZER(expr, analyzer), which sets the Analyzer used to evaluate the wrapped search expression and all the nested functions that accept an Analyzer argument (so it does not have to be repeated). A nested function passing its own Analyzer takes precedence. Only applicable to queries against arangosearch Views — with search-alias Views and inverted indexes the Analyzer is inferred from the index definition.

The TOKENS() function is an exception: it always requires its own Analyzer argument, even when wrapped, because it is a regular string function.

Example AQL usage:

ANALYZER(PHRASE(doc.text, "foo") OR PHRASE(doc.text, "bar"), "text_en")

Wraps the ArangoDB AQL scoring function BM25(doc, k, b). The first argument must be the document variable emitted by a FOR … IN viewName operation, and the function can only be used together with a SEARCH. Sort descending by the score to get the most relevant documents first.

AQL arguments are positional: when $b is provided without $k, the helper fills $k with the official server default (1.2). The Analyzers used for indexing must have the "frequency" feature enabled (and "norm" for meaningful length normalization), otherwise the score is 0.

Example AQL usage:

FOR doc IN viewName
  SEARCH ...
  SORT BM25(doc) DESC
  RETURN doc

Wraps the ArangoDB AQL function BOOST(expr, boost). The boost value is made available to scorer functions (bm25(), tfidf()) so that matches of the wrapped sub-expression weigh more (or less) in the final score. The default boost of any search context is 1.0.

Example AQL usage:

ANALYZER(BOOST(doc.text == "foo", 2.5) OR doc.text == "bar", "text_en")

Wraps the ArangoDB AQL function EXISTS(path[, type[, analyzer]]):

• exists('doc.text') — the attribute is present;
• exists('doc.text', 'string') — present and of the given data type ("null", "bool"/"boolean", "numeric", "type", "string", "analyzer", "nested");
• exists('doc.text', analyzer: 'text_en') — present and indexed by the given Analyzer; the "analyzer" type literal is filled in automatically.

With arangosearch Views, EXISTS() only matches values if the storeValues link property is set to "id" (default "none").

Example AQL usage:

EXISTS(doc.text)
EXISTS(doc.text, "string")
EXISTS(doc.text, "analyzer", "text_en")

Wraps the ArangoDB AQL function IN_RANGE(path, low, high, includeLow, includeHigh). Inside a SEARCH operation it searches more efficiently than the equivalent pair of comparisons combined with AND; the same function also exists as a miscellaneous function outside of SEARCH (hence the constant living in MiscFunction).

low and high can be numbers or strings, but both must share the same data type. Note that string ranges in SEARCH follow byte order, not the Analyzer locale collation.

Example AQL usage:

IN_RANGE(doc.value, 3, 5, true, true)     // 3 <= value <= 5
IN_RANGE(doc.value, "a", "f", true, false) // "a" <= value < "f"

Wraps the ArangoDB AQL function LEVENSHTEIN_MATCH(path, target, distance, transpositions, maxTerms, prefix). By default a Damerau-Levenshtein distance is computed (transpositions count as one operation); pass transpositions: false for a pure Levenshtein distance. The maximum distance is 4 without transpositions and 3 with them.

AQL arguments are positional: when a later option is provided, the helper fills the earlier omitted ones with the official server defaults (transpositions = true, maxTerms = 64) so callers never need to know them. Trailing omitted options are not emitted at all. When using $prefix, the prefix must be removed from $target (the distance is computed on the remainders — see the official documentation).

Example AQL usage:

LEVENSHTEIN_MATCH(doc.text, "quikc", 2, false)        // pure Levenshtein, matches "quick"
LEVENSHTEIN_MATCH(doc.text, "kc", 1, false, 64, "qui") // prefix search

Wraps the ArangoDB AQL function MINHASH_MATCH(path, target, threshold, analyzer). The similarity is approximated with the given minhash Analyzer — an efficient first pass for entity resolution (duplicate detection) before an exact JACCARD() computation.

Argument order notice — in AQL the optional threshold sits before the mandatory analyzer; PHP forbids a required parameter after an optional one, so this helper takes the analyzer third and the optional threshold last, then re-orders the emitted AQL arguments.

Example AQL usage:

MINHASH_MATCH(doc.text, "the quick brown fox", 0.5, "myMinHash")

Wraps the variadic ArangoDB AQL function MIN_MATCH(expr1, ... exprN, minMatchCount). Inside a SEARCH operation it is index-accelerated; the same function also exists as a miscellaneous function outside of SEARCH (hence the constant living in MiscFunction).

Example AQL usage:

MIN_MATCH(doc.text == "quick", doc.text == "brown", doc.text == "fox", 2)

Wraps the ArangoDB AQL function NGRAM_MATCH(path, target, threshold, analyzer). The n-grams of both the attribute and the target are produced by the given Analyzer (use an ngram Analyzer with preserveOriginal: false and min equal to max, with the "position" and "frequency" features enabled).

Argument order notice — in AQL the optional threshold sits before the mandatory analyzer; PHP forbids a required parameter after an optional one, so this helper takes the analyzer third and the optional threshold last, then re-orders the emitted AQL arguments. When $threshold is null the three-argument AQL form is emitted (server default: 0.7).

Example AQL usage:

NGRAM_MATCH(doc.text, "quick fox", "bigram")           // threshold defaults to 0.7
NGRAM_MATCH(doc.text, "quick blue fox", 0.4, "bigram")

Wraps the ArangoDB AQL function PHRASE(path, phrasePart, analyzer). The phrase can be:

• a string — the simple form, emitted as a quoted string literal;
• an array — the advanced AQL array form, emitted with json_encode, mirroring the official syntax one-to-one: string tokens are quoted, integers act as skipTokens wildcards, and associative arrays become object tokens ({IN_RANGE: …}, {LEVENSHTEIN_MATCH: …}, {STARTS_WITH: …}, {TERM: …}, {TERMS: …}, {WILDCARD: …}).

The Analyzer must have the "position" and "frequency" features enabled, otherwise PHRASE() finds nothing. When $analyzer is omitted, the Analyzer of a wrapping analyzer() call applies (default "identity").

Example AQL usage:

PHRASE(doc.text, "quick fox", "text_en")
PHRASE(doc.text, ["ipsum", 2, "amet"], "text_en")              // 2 wildcard tokens between
PHRASE(doc.text, ["lorem", {STARTS_WITH: ["ips"]}], "text_en") // prefix object token

Wraps the ArangoDB AQL scoring function TFIDF(doc, normalize). The first argument must be the document variable emitted by a FOR … IN viewName operation, and the function can only be used together with a SEARCH. Sort descending by the score to get the most relevant documents first.

Example AQL usage:

FOR doc IN viewName
  SEARCH ...
  SORT TFIDF(doc) DESC
  RETURN doc