Gelatin Syntax

The following functions and types may be used within a Gelatin syntax.



A string is any series of characters, delimited by the ' character. Escaping is done using the backslash character. Examples:

'test me'
'test \'escaped\' strings'


VARNAMEs are variable names. They may contain the following set of characters:



The output that is generated by Gelatin is represented by a tree consisting of nodes. The NODE type is used to describe a single node in a tree. It is a URL notated string consisting of the node name, optionally followed by attributes. Examples for NODE include:



A PATH addresses a node in the tree. Addressing is relative to the currently selected node. A PATH is a string with the following syntax:





This type describes a Python regular expression. The expression MUST NOT extract any subgroups. In other words, when using bracket expressions, always use (?:). Example:

/^(test|foo|bar)$/         # invalid!
/^(?:test|foo|bar)$/       # valid

If you are trying to extract a substring, use a match statement with multiple fields instead.



define statements assign a value to a variable. Examples:

define my_test /(?:foo|bar)/
define my_test2 'foobar'
define my_test3 my_test2


Match statements are lists of tokens that are applied against the current input document. They parse the input stream by matching at the current position. On a match, the matching string is consumed from the input such that the next match statement may be applied. In other words, the current position in the document is advanced only when a match is found.

A match statement must be followed by an indented block. In this block, each matching token may be accessed using the $X variables, where X is the number of the match, starting with $0.


define digit /[0-9]/
define number /[0-9]+/

grammar input:
    match 'foobar':
        do.say('Match was: $0!')
    match 'foo' 'bar' /[\r\n]/:
        do.say('Match was: $0!')
    match 'foobar' digit /\s+/ number /[\r\n]/:
        do.say('Matches: $1 and $3')

You may also use multiple matches resulting in a logical OR:

match 'foo' '[0-9]' /[\r\n]/
    | 'bar' /[a-z]/ /[\r\n]/
    | 'foobar' /[A-Z]/ /[\r\n]/:
    do.say('Match was: $1!')


imatch statements are like match statements, except that matching is case-insensitive.


when statements are like match statements, with the difference that upon a match, the string is not consumed from the input stream. In other words, the current position in the document is not advanced, even when a match is found. when statements are generally used in places where you want to “bail out” of a grammar without consuming the token.


grammar user:
    match 'Name:' /\s+/ /\S+/ /\n/:
        do.say('Name was: $2!')
    when 'User:':

grammar input:
    match 'User:' /\s+/ /\S+/ /\n/:
        out.enter('user/name', '$2')


skip statements are like match statements without any actions. They also do not support lists of tokens, but only one single expression.


grammar user:
    skip /#.*?[\r\n]+/
    match 'Name: ' /\s+/ /\n/:
        do.say('Name was: $2!')
    when 'User:':

Output Generating Functions

out.create(PATH[, STRING])

Creates the leaf node (and attributes) in the given path, regardless of whether or not it already exists. In other words, using this function twice will lead to duplicates. If the given path contains multiple elements, the parent nodes are only created if the do not yet exist. If the STRING argument is given, the new node is also assigned the string as data. In other words, the following function call:

out.create('parent/child?name="test"', 'hello world')

leads to the following XML output:

    <child name="test">hello world</child>

Using the same call again, like so:

out.create('parent/child?name="test"', 'hello world')
out.create('parent/child?name="test"', 'hello world')

the resulting XML would look like this:

    <child name="test">hello world</child>
    <child name="test">hello world</child>

out.replace(PATH[, STRING])

Like out.create(), but replaces the nodes in the given path if they already exist.

out.add(PATH[, STRING])

Like out.create(), but appends the string to the text of the existing node if it already exists.

out.add_attribute(PATH, NAME, STRING)

Adds the attribute with the given name and value to the node with the given path.[, STRING])

Like out.create(), but also selects the addressed node, such that the PATH of all subsequent function calls is relative to the selected node until the end of the match block is reached.

out.enter(PATH[, STRING])

Like, but only creates the nodes in the given path if they do not already exist.

out.enqueue_before(REGEX, PATH[, STRING])

Like out.add(), but is not immediately executed. Instead, it is executed as soon as the given regular expression matches the input, regardless of the grammar in which the match occurs.

out.enqueue_after(REGEX, PATH[, STRING])

Like out.enqueue_before(), but is executed after the given regular expression matches the input and the next match statement was processed.

out.enqueue_on_add(REGEX, PATH[, STRING])

Like out.enqueue_before(), but is executed after the given regular expression matches the input and the next node is added to the output.


Removes any items from the queue that were previously queued using the out.enqueue_*() functions.


Specifies the name of the root tag, if the output is XML. Has no effect on JSON and YAML output.

Control Functions


Skip the current match and jump back to the top of the current grammar block.

Skip the current match and continue with the next match statement without jumping back to the top of the current grammar block.
This function is rarely used and probably not what you want. Instead, use do.skip() in almost all cases, unless it is for some performance-specific hacks.


Immediately leave the current grammar block and return to the calling function. When used at the top level (i.e. in the input grammar), stop parsing.


Prints the given string to stdout, with additional debug information.

Like do.say(), but immediately terminates with an error.