ElasticSearch::QueryParser - Check or filter query strings
Passing an illegal query string to ElasticSearch, the request will fail. When using a query string from an external source, eg the keywords field from a web search form, it is important to filter it to avoid these failures.
You may also want to allow or disallow certain query string features, eg the ability to search on a particular field.
The ElasticSearch::QueryParser takes care of this for you.
See http://lucene.apache.org/java/3_0_3/queryparsersyntax.html for more information about the Lucene Query String syntax, and http://www.elasticsearch.org/guide/reference/query-dsl/query-string-query.html#Syntax_Extension for custom ElasticSearch extensions to the query string syntax.
use ElasticSearch; my $es = ElasticSearch->new(servers=>'127.0.0.1:9200'); my $qp = $es->query_parser(%opts); my $filtered_query_string = $qp->filter($unchecked_query_string) my $results = $es->search( query=> { query_string=>{ query => $filtered_query_string } });
For example:
my $qs = 'foo NOT AND -bar - baz * foo* secret_field:SIKRIT "quote'; print $qp->filter($qs); # foo AND -bar baz foo* "quote"
my $qp = ElasticSearch::QueryParser->new(%opts); my $qp = $es->query_parser(%opts);
Creates a new ElasticSearch::QueryParser object, and sets the passed in options (see "OPTIONS").
$filtered_query_string = $qp->filter($unchecked_query_string, %opts)
Checks a passed in query string and returns a filtered version which is suitable to pass to ElasticSearch.
Note: filter() can still return an empty string, which is not considered a valid query string, so you should still check for that before passing to ElasticSearch.
filter()
If any %opts are passed in to filter(), these are added to the default %opts as set by "new()", and apply only for the current run.
%opts
"filter()" does not promise to parse the query string in exactly the same way as Lucene, just to clear it up so that it won't throw an error when passed to ElasticSearch.
$filtered_query_string = $qp->check($unchecked_query_string, %opts)
Checks a passed in query string and throws an error if it is not valid. This is useful for debugging your own query strings.
If any %opts are passed in to check(), these are added to the default %opts as set by "new()", and apply only for the current run.
check()
You can set various options to control how your query strings are filtered.
The defaults (if no options are passed in) are:
escape_reserved => 0 fields => 0 boost => 1 allow_bool => 1 allow_boost => 1 allow_fuzzy => 1 allow_slop => 1 allow_ranges => 0 wildcard_prefix => 1
Any options passed in to "new()" are merged with these defaults. These options apply for the life of the QueryParser instance.
Any options passed in to "filter()" or "check()" are merged with the options set in "new()" and apply only for the current run.
For instance:
$qp = ElasticSearch::QueryParser->new(allow_fuzzy => 0); $qs = "foo~0.5 bar^2 foo:baz"; print $qp->filter($qs, allow_fuzzy => 1, allow_boost => 0); # foo~0.5 bar baz print $qp->filter($qs, fields => 1 ); # foo bar^2 foo:baz
Reserved characters must be escaped to be used in the query string. By default, "filter()" will remove these characters. Set escape_reserved to true if you want them to be escaped instead.
escape_reserved
Reserved characters: + - && || ! ( ) { } [ ] ^ " ~ * ? : \
+ - && || ! ( ) { } [ ] ^ " ~ * ? : \
Normally, you don't want to allow your users to specify which fields to search. By default, "filter()" removes any field prefixes, eg:
$qp->filter('foo:bar secret_field:SIKRIT') # bar SIKRIT
You can set fields to 1 to allow all fields, or pass in a hashref with a list of approved fieldnames, eg:
fields
1
$qp->filter('foo:bar secret_field:SIKRIT', fields => 1); # foo:bar secret_field:SIKRIT $qp->filter('foo:bar secret_field:SIKRIT', fields => {foo => 1}); # foo:bar SIKRIT
ElasticSearch extends the standard Lucene syntax to include:
_exists_:fieldname and _missing_:fieldname
The fields option applies to these fieldnames as well.
Query strings can use boolean operators like:
foo AND bar NOT baz OR ! (foo && bar)
By default, boolean operators are allowed. Set allow_bool to false to disable them.
allow_bool
false
Note: This doesn't affect the + or - operators, which are always allowed. eg:
+
-
+apple -crab
Boost allows you to give a more importance to a particular word, group of words or phrase, eg:
foo^2 (bar baz)^3 "this exact phrase"^5
By default, boost is enabled. Setting allow_boost to false would convert the above example to:
allow_boost
foo (bar baz) "this exact phrase"
Lucene supports fuzzy searches based on the Levenshtein Distance, eg:
supercalifragilisticexpialidocious~0.5
To disable these, set allow_fuzzy to false.
allow_fuzzy
While a phrase search (eg "this exact phrase") looks for the exact phrase, in the same order, you can use phrase slop to find all the words in the phrase, in any order, within a certain number of words, eg:
phrase search
"this exact phrase"
For the phrase: "The quick brown fox jumped over the lazy dog." Query string: Matches: "quick brown" Yes "brown quick" No "quick fox" No "brown quick"~2 Yes # within 2 words of each other "fox dog"~6 Yes # within 6 words of each other
To disable this "phrase slop", set allow_slop to false
allow_slop
Lucene can accept ranges, eg:
date:[2001 TO 2010] name:[alan TO john]
To enable these, set allow_ranges to true.
allow_ranges
true
Lucene can accept wildcard searches such as:
jo*n smith?
Lucene takes these wildcards and expands the search to include all matching terms, eg jo*n could be expanded to jon, john, jonathan etc
jo*n
jon
john
jonathan
This can result in a huge number of terms, so it is advisable to require that the first $min characters of the word are not wildcards.
$min
By default, the wildcard_prefix requires that at least the first character is not a wildcard, ie * is not acceptable, but s* is.
wildcard_prefix
*
s*
You can change the minimum length of the non-wildcard prefix by setting wildcard_prefix, eg:
$qp->filter("foo* foobar*", wildcard_prefix=>4) # "foo foobar*"
This is a new module, so it is likely that there will be bugs, and the list of options and how "filter()" cleans up the query string may well change.
If you have any suggestions for improvements, or find any bugs, please report them to http://github.com/clintongormley/ElasticSearch.pm/issues.
Patches welcome!
Copyright 2010 Clinton Gormley.
This program is free software; you can redistribute it and/or modify it under the terms of either: the GNU General Public License as published by the Free Software Foundation; or the Artistic License.
See http://dev.perl.org/licenses/ for more information.
To install ElasticSearch, copy and paste the appropriate command in to your terminal.
cpanm
cpanm ElasticSearch
CPAN shell
perl -MCPAN -e shell install ElasticSearch
For more information on module installation, please visit the detailed CPAN module installation guide.