Notice: this Wiki will be going read only early in 2024 and edits will no longer be possible. Please see: https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/wikis/Wiki-shutdown-plan for the plan.
Difference between revisions of "Henshin/Xtext Adapter"
| Line 5: | Line 5: | ||
When defining an Xtext language, you essentially write down an (annotated) EBNF grammar for it and Xtext will generate a meta-model etc. for you. One cool feature is that you can explicitly describe cross-references. The way to do this is by replacing a direct reference to a non-terminal (say A) with the name of the non-terminal in angular brackets, possibly followed by a token/data-type to be used to identify the relevant object. So, for example we could say: | When defining an Xtext language, you essentially write down an (annotated) EBNF grammar for it and Xtext will generate a meta-model etc. for you. One cool feature is that you can explicitly describe cross-references. The way to do this is by replacing a direct reference to a non-terminal (say A) with the name of the non-terminal in angular brackets, possibly followed by a token/data-type to be used to identify the relevant object. So, for example we could say: | ||
| − | A : /* … description of what | + | A : /* … description of what A's look like */ ; |
| − | B: | + | B: 'anA:' myA = [A|ID] ; |
This allows us to write | This allows us to write | ||
| Line 17: | Line 17: | ||
The beautiful thing about this is that it also works across models and languages. So, after importing the Henshin metamodel in your Xtext grammar, you can say things like | The beautiful thing about this is that it also works across models and languages. So, after importing the Henshin metamodel in your Xtext grammar, you can say things like | ||
| − | GTS: | + | GTS: 'gts' 'metamodel' ':' metamodel = [EPackage|QualifiedName] 'henshin' ':' module=[Module|ID] ; |
And get Xtext to do all the resolving for you. By default, any Henshin module that is visible on the classpath of my current project will be available (that’s configurable) and code completion can help you find them. | And get Xtext to do all the resolving for you. By default, any Henshin module that is visible on the classpath of my current project will be available (that’s configurable) and code completion can help you find them. | ||
Revision as of 11:06, 5 January 2020
Henshin Xtext adapter is a plugin that allows Henshin modules, rules, nodes etc to be referenced directly from any Xtext-based language.
Intended use
When defining an Xtext language, you essentially write down an (annotated) EBNF grammar for it and Xtext will generate a meta-model etc. for you. One cool feature is that you can explicitly describe cross-references. The way to do this is by replacing a direct reference to a non-terminal (say A) with the name of the non-terminal in angular brackets, possibly followed by a token/data-type to be used to identify the relevant object. So, for example we could say:
A : /* … description of what A's look like */ ; B: 'anA:' myA = [A|ID] ;
This allows us to write
anA MYA
asking Xtext to look up and find an instance of A called MYA. From the grammar above, Xtext generates a meta-class B with a non-containment reference myA pointing at an A instance. All the name resolution is done as part of parsing the textual model.
The beautiful thing about this is that it also works across models and languages. So, after importing the Henshin metamodel in your Xtext grammar, you can say things like
GTS: 'gts' 'metamodel' ':' metamodel = [EPackage|QualifiedName] 'henshin' ':' module=[Module|ID] ;
And get Xtext to do all the resolving for you. By default, any Henshin module that is visible on the classpath of my current project will be available (that’s configurable) and code completion can help you find them.
However, for all of this to work, Xtext needs some support. In particular, it needs ways of (a) finding Henshin files, and (b) giving names to elements within Henshin files. That's what this plugin provides: everything needed for using Henshin modules while parsing and manipulating Xtext files, and configuration so that the name provider can be correctly used in the context of an Xtext-generated editor.
Usage instructions
- Add org.eclipse.emf.henshin.adapters.xtext and org.eclipse.emf.henshin.model to the dependencies of your Xtext language plugin (and potentially your test plugin).
- Add org.eclipse.emf.henshin.adapters.xtext to the dependencies of your language's .ui plugin.
- Add referencedResource = "platform:/resource/org.eclipse.emf.henshin.model/model/henshin.genmodel" to the StandardLanguage block in your .mwe2 file.
- Add import "http://www.eclipse.org/emf/2011/Henshin" as henshin to your .xtext file.
- Use [henshin::Module] etc in your grammar to reference Henshin meta-classes.
By default, links have no name in Henshin. To allow them to be referenced nonetheless, the plugins generate names following the grammar rule below:
LinkName : "[" ID "->" ID ":" ID "]" ;
The first ID is the name of the source node, the second the name of the target node, while the third is the name of the EReference. This naming scheme can be imported into any class in your own code by statically importing org.eclipse.emf.henshin.adapters.xtext.NamingHelper. In an Xtend class, it is recommended to import the class as an extension.
Credits and links
- Contributed by Steffen Zschaler, see the original implementation on GitHub
- Xtext website