Developing Domain-Specific Languages in Python with textX

Igor Dejanović (igord at uns ac rs)

Strumenta Meetups, 24 February 2022

Created 2022-02-24 Thu 19:18, press ESC for overview and use arrow keys for movement, press Ctrl+Shift+F for search

Table of Content

1. Overview


2. A brief history

3. Setup

To create Python environment

python -m venv venv

then, activate the environment and install textX.

source venv/bin/activate
pip install textx[cli]

4. Hello World

model = """
hello You, Me; Everybody
mm_str = r"""
Hello: 'hello' to_greet+=Who[/,|;/];
Who: name=ID;
from textx import metamodel_from_str
mm = metamodel_from_str(mm_str)
m = mm.model_from_str(model)
print([ for who in m.to_greet])
['You', 'Me', 'Everybody']

See: matches, assignments, and rule types.

5. Hello from file

In file hello.tx:

Hello: 'hello' to_greet+=Who[/,|;/];
Who: name=/[^,;]+/;

In file model.hello:

hello World, Solar System; Universe

In file

from textx import metamodel_from_file
def who_processor(who): =
mm = metamodel_from_file('hello.tx')
mm.register_obj_processors({'Who': who_processor})
m = mm.model_from_file('model.hello')
print([ for who in m.to_greet])
<textx.metamodel.TextXMetaModel object at 0x7f5add1b7dc0>
['World', 'Solar System', 'Universe']

6. textx command

Used for checking the grammar, investigating languages and generators and running generators.

$ textx --help
Usage: textx [OPTIONS] COMMAND [ARGS]...

  --debug  Debug/trace output.
  --help   Show this message and exit.

  check            Check/validate model given its file path.
  generate         Run code generator on a provided model(s).
  list-generators  List all registered generators
  list-languages   List all registered languages
  version          Print version info.

7. Grammar check

textx check --grammar hello.tx model.hello
Error: None:1:1: error: Expected 'hello' at position /home/igor/repos/, 1) => '*hi World, '.

8. Running example - Workflow DSL

A tiny workflow DSL will be used in the rest of the slides.

8.1. Model/Program

An example in this language might look like:

package BuildHouse {
    task buyLand {
    searchAds, chooseLand, buyLand
    next makePlan
    task makePlan {
    searchforArchitect, giveInstructions, choosePlan
    next buildHouse
    task buildHouse {
package BuildFence {
  task buildFence {
    chooseMaterial, buildFence

8.2. Metamodel/Grammar

Model: elements*=Element;
Element: Package | Task;
Package: 'package' name=ID '{'
Task: 'task' name=ID '{'
         ('next' next=[Task])?

8.3. Check/use the language

from textx import metamodel_from_file
mm = metamodel_from_file('workflow.tx')
model = mm.model_from_file('example.workflow')
<textx:workflow.Model instance at 0x7f655fbd4340>
textx check --grammar workflow.tx example.workflow
/home/igor/repos/ OK.

9. Visualize (meta-)model

Visualization is done using textX’s generator framework.

$ textx list-generators
any -> dot         textX[2.3.0]  Generating dot visual...
textX -> dot       textX[2.3.0]  Generating dot visual...
textX -> PlantUML  textX[2.3.0]  Generating PlantUML v...

To visualize (meta-)model we use generators that produce dot or plantuml outputs.

textx generate workflow.tx --target dot --overwrite

dot file can be visualized either by transforming to an image using dot tool (part of GraphViz):

dot -Tpng -O

Or opening it in some dot visualizer, e.g. xdot.

Similarly, we can produce PlantUML diagram by specifying plantuml target.

textx generate workflow.tx --target plantuml --overwrite
Generating plantuml target from models:
-> /home/igor/repos/
    To convert to png run "plantuml workflow.pu"

And then convert it to png image using plantuml:

plantuml workflow.pu


You get a nice UML diagrams directly from your grammars.

10. Some differences to xText

A few notes for those familiar with xText.

10.1. Lexical grammar

textX doesn’t have a separate lexical grammar. There are Match rules that resembles something close to lexical grammar but not quite.

10.2. Assignments

textX integrates repetition and assignments:

In xText you would write:

Domainmodel :

While in textX it would be:

Domainmodel :

Optional assignment in xText:


In textX:


10.3. Regex matches

textX has simple string matches (like 'something to match') and regex matches where you can use a full power of Python regex engine inside /.../.

For example:

    name=/[a-zA-Z]+/ age=INT;

10.4. Repetition modifiers

textX provides a syntactic construct called repetition modifier which enables parser to be altered during parsing of a specific repetition expression.


list_of_ints+=INT (',' list_of_ints+=INT)*



Modifier can also be a regex match:


Repetition modifier can be applied to any repetition (zero or more, one or more, optional, unordered group).

(First /\d+/ Second)*[',']

Besides matches there are other modifiers. For example EOL terminator:

STRING*[',', eolterm]

would match the first line of:

"first", "second", "third"
, "fourth"

10.5. Unordered groups

Xtext support unordered groups using the & operator.

    static?='static'? & final?='final'? & visibility=Visibility;

enum Visibility:
    PUBLIC='public' | PRIVATE='private' | PROTECTED='protected';

In textX unordered groups are specified as a special kind of repetitions. Thus, repetition modifiers can be applied also:

    (static?='static' final?='final' visibility=Visibility)#[',']

    'public' | 'private' | 'protected';


private, static, final
static, private, final
public, static

10.6. Scoping

Scoping in textX is done either by using Python through registration of scope providers, or declaratively using Reference Resolving Expression Language.

Xtext provides a Scoping API which can be used by the Xtend code to specify scoping rules.

10.7. More differences

For more differences please see this page.

11. Language/generator registration

11.1. Create language description

from textx import LanguageDesc

def entity_metamodel():
    # Construct and configure the meta-model
    # e.g. by calling metamodel_from_file

entity_lang = LanguageDesc(
    description='Entity-relationship language',

11.2. Programmatic registration

LanguageDesc instance can be registered programmatically by the register_language function:

from textx import register_language

The meta-model can be accessed from any Python program like this:

from textx import metamodel_for_language
lang_mm = metamodel_for_language('entity')

11.3. Declarative registration

  • Registration can be done declaratively using or setup.cfg.
        'textx_languages': [
            'entity = entity.metamodel:entity_lang',

11.4. Using a decorator

There is a convenient language decorator to make registration easier.

from textx import language

@language('entity', '*.ent')
def entity_lang():
    Entity-relationship language
    # Create, configure and return an instance of the meta-model

12. Scoping and RREL

  • In a link rule reference, the name matched at the location must be resolved to the referenced object.

    For example

        Attribute: 'attr' ref=[Class] name=ID ';';
  • Global search by default.
  • Programmatic scoping providers may be registered to resolve references.

12.1. RREL

Declarative specification of reference resolving strategy.


Attribute: 'attr' ref=[Class|FQN|^packages*.classes] name=ID ';';

12.2. RREL operators and markers

  • . - dot navigation. Searches for the attribute in the current AST context.
    • e.g. . is this object, .. is parent, ... is a parent of a parent
    • relative lookup. Example: .a.b
  • ~ - do not consume name.
    • ~extends*.methods - search for method name in the inheritance hierarchy.
  • * - repeat/expand - .~extends*.methods expands to: .methods, .~extends.methods, .~extends.~extends.methods
  • ^ - bottom-up search. Example ^packages*.classes expands to .classes, ..packages.classes, ...packages.packages.classes

12.3. Extending example to use FQN

12.3.1. Model

package BuildHouse {
        task feasibility DONE {
             next buyLand
        task buyLand DONE {
             searchAds, findLand, buyLand
             next makePlan
        task makePlan DOING {
             chooseArchitect, giveInstructions, choosePlan
             next buildHouse
        task buildHouse TODO {
             next BuildFence.feasibility
        task moveIn {}
package BuildFence {
        task feasibility TODO {}
        task buildFence {
             chooseCompany, giveInstructions, buildFence
             next BuildHouse.moveIn

12.3.2. Meta-model

Model: elements+=Element;
Package: 'package' name=ID '{'
Element: Package | Task;
Task: 'task' name=ID (state=State)? '{'
            ('next' next+=[Task|FQN|^elements*.elements][','])?
State: 'TODO' | 'DOING' | 'DONE';
Step: !'next' ID;
FQN: ID+['.'];
Comment: /\/\/.*/;

13. Project scaffolding

$ textx startroject <folder>

Command asks a few questions and generates the project files. To install the generated project in developers mode (editable):

$ pip install -e <folder>

After installation the language (or generator) is visible to the textx list-languages or textx list-generators commands.

startproject command is not defined in the base textX library but in textX-dev package. Thus to have it registered we must install this project:

pip install textX-dev

Alternatively, we can install all dev dependencies:

pip install textX[dev]

14. textX-LS

  • textX-LS is a language server that provides smartness for all domain specific languages based on textX.
  • Consists of three parts: core, server and VS Code client.
  • Uses two textX generators:

15. A more complex example - PyFlies

  • PyFlies is Domain-Specific Language (DSL) for designing experiments in psychology

16. Where to go next?