OpenSearch/docs/reference/analysis/tokenizers/pattern-tokenizer.asciidoc

[[analysis-pattern-tokenizer]]
=== Pattern Tokenizer

A tokenizer of type `pattern` that can flexibly separate text into terms
via a regular expression. Accepts the following settings:

[cols="<,<",options="header",]
|======================================================================
|Setting |Description
|`pattern` |The regular expression pattern, defaults to `\W+`.
|`flags` |The regular expression flags.
|`group` |Which group to extract into tokens. Defaults to `-1` (split).
|======================================================================

*IMPORTANT*: The regular expression should match the *token separators*,
not the tokens themselves.

*********************************************
Note that you may need to escape `pattern` string literal according to
your client language rules. For example, in many programming languages
a string literal for `\W+` pattern is written as `"\\W+"`.
There is nothing special about `pattern` (you may have to escape other
string literals as well); escaping `pattern` is common just because it
often contains characters that should be escaped.
*********************************************

`group` set to `-1` (the default) is equivalent to "split". Using group
>= 0 selects the matching group as the token. For example, if you have:

------------------------
pattern = '([^']+)'
group   = 0
input   = aaa 'bbb' 'ccc'
------------------------

the output will be two tokens: `'bbb'` and `'ccc'` (including the `'`
marks). With the same input but using group=1, the output would be:
`bbb` and `ccc` (no `'` marks).