Search Syntax and Parsers

BestGuess

The BestGuess parser will try to treat the query as if the Term parser was specified, unless it encounters a syntax construct it does not understand, at which point it will change into a Lucene parser and pass the query through without modification.

Terms split by whitespace (unless quoted). Constructs allowed for BestGuess Term parsing:

• alphanumeric sequence
• + and - inside a sequence (will be escaped)
• * and ? (wildcards) anywhere in a sequence (will not be escaped)
• any sequence surrounded by quotes (quotes in the term must be escaped with backslash)
• leading + (will not be escaped)
• leading - (will not be escaped)
• field:term pairs (two sequences separated by a : ) for known fields. If a second : is found it will be escaped. The known fields are:
  • content
  • title
  • path.title
  • description
  • tag
  • comments
  • path
  • id.page
  • id.parent
  • id.file
  • path.parent
  • title.parent
  • title.page
  • author
  • #{propertyname}
  • {ns}#{propertyname}
  • filename
• namespace is not recognized and should be specified per query in constraints or globally in the search/namespace-exclude config string

Boost, range query, AND, NOT, OR, parentheses are some examples of constructs that will cause the parser to shortcircuit and switch to the Lucene parser.

Filename

The Filename parser will treat the input as a single term, escaping all special characters (including whitespace) and searching both the filename field with wildcard trailing the term and the extension field with the term value. On the indexing size, filenames are tokenized as a single term with all whitespace and dashes normalized to underscores. Some sample query translations are:

jbg => filename:jpg* extension:jpg

IMG-0123 => filename:IMG_0123* extension:IMG\-0123

foo bar => filename:foo_bar* extension:foo\ bar

Term

The Term parser will escape anything that would be considered by lucene as special syntax and treat the query as a list of terms, building a lucene query per term based on the search/termquery formatting string

Lucene

The Lucene parser will pass the query unmodified through to lucene. This means that all field specifications need to be manually provided, since terms without a field prefix will only search the content field. Constraints are stil ANDed to the query, providing at the very least type filtering.

search/namespace-exclude

Default: -namespace:"template" -namespace:"template_talk" -namespace:"help" -namespace:"help_talk"

search/termquery

Default: content:{0} title:{0}^4 path.title:{0}^4 description:{0}^3 tag:{0}^2 comments:{0}