NLP++ TextMate Grammars

TextMate grammars for NLP++, the programming language built for natural language processing, and for the supporting VisualText file formats.

These are the canonical, license-clean source grammars. They are extracted from the VS Code NLP++ extension so that downstream projects — GitHub Linguist, Shiki, bat, Sublime Text, JetBrains IDEs, and anything else that consumes TextMate grammars — can vendor them directly without pulling in the whole extension.

Grammars

File	Scope name	Extensions	Description
syntaxes/nlp.tmLanguage.json	source.nlp	.nlp, .pat	NLP++ language — rules, regions, and code
syntaxes/dict.tmLanguage.json	source.dict	.dict	Dictionary files
syntaxes/kb.tmLanguage.json	source.kb	.kb	Knowledge base dumps
syntaxes/kbb.tmLanguage.json	source.kbb	.kbb	Knowledge base build files
syntaxes/tree.tmLanguage.json	source.tree	.tree, .log	Parse tree output
syntaxes/txxt.tmLanguage.json	source.txxt	.txxt	Highlighted text output
syntaxes/seq.tmLanguage.json	source.seq	.seq	Analyzer pass sequence files

source.nlp is the one most integrations want. The others describe VisualText's data and output formats rather than the language itself.

Matching VS Code language configurations (brackets, comments, auto-closing pairs) are in language-configuration/.

Scope names

Every scope follows the TextMate naming conventions (keyword.control, constant.language, entity.other.attribute-name, string.quoted.double, ...) and ends in the language's own suffix (.nlp, .tree, ...), so any theme colors NLP++ without custom rules. Older versions used invented names that only colored correctly with the VS Code extension's own color settings. If you have color customizations keyed to those names, rename them:

Old	New
keyword.region.nlp (@RULES, @CODE, ...)	keyword.control.directive.region.nlp
keyword.constants.nlp (_xWILD, _ROOT, ...)	constant.language.nlp
keyword.attribute.nlp (match, opt, trig, ...)	entity.other.attribute-name.nlp
keyword.equals.nlp	keyword.operator.assignment.nlp
variable.rewrite.nlp (the suggested concept)	entity.name.type.rewrite.nlp
variable.rewrite-symbol.nlp	meta.attribute.rewrite.nlp
literal.nlp, literal.seq, literal.tree	meta.literal.*
string.double.*	string.quoted.double.*
constant.character.escape, punctuation.paren.*, punctuation.square.* (.nlp)	same, with .nlp
keyword.rewrite.tree (_NP, ...)	entity.name.type.tree
keyword.node.tree (token text)	string.unquoted.token.tree
keyword.attribute.tree (alpha, num, ...)	storage.type.tree
keyword.fired.tree	keyword.other.fired.tree
keyword.other.kbb, keyword.concept.kbb, keyword.concept.kbb1, keyword.concept.kbb2	entity.name.type.level0.kbb ... level3.kbb
comment.line.dict (headword), keyword.region.dict (key), entity.name.function.dict (value)	entity.name.tag.dict, support.type.property-name.dict, string.unquoted.value.dict
comment.line.dict in .kb, .kbb, .tree; constant.numeric.tree in .kbb; meta.expression.tree in .txxt	the file's own suffix
variable.parameter.nlp in .kb (add, bind, ...)	keyword.control.kb

Comments

# runs to end of line. C-style /* ... */ block comments are supported by the NLP++ engine as of version 3.7.14, in the pass language (.nlp, .pat) and in the four line-oriented data formats (.seq, .kb, .dict, .kbb). They do not nest — the first */ closes — and delimiters inside a string or a # comment are ordinary text. .tree files are engine output and use * line comments only.

Samples

samples/ contains real-world NLP++ from the parse-en-us English parser (MIT licensed, same authors) — useful for grammar regression testing and for submissions that require non-trivial example code.

Using these grammars

Shiki / VS Code / any TextMate host — point at the raw JSON:

const nlp = await fetch(
  'https://raw.githubusercontent.com/VisualText/nlpplus-tmbundle/main/syntaxes/nlp.tmLanguage.json'
).then(r => r.json())

Vendoring — add this repository as a git submodule. It is intentionally small and has no build step, no dependencies, and no generated files.

About NLP++

NLP++ is a programming language designed specifically for natural language processing, created by Amnon Meyers and David de Hilster. Analyzers are written as sequences of passes, each applying pattern rules to a parse tree. See visualtext.org.

License

MIT — see LICENSE. Copyright 1998- Amnon Meyers, David de Hilster.