I regularly spend time working on FHIR ImplementationGuides, and I use the Visual Studio Code IDE to develop them. It is an IDE I used to work with, but as of lately, my preference goes to the Jetbrains IDEs.
The FHIR ImplementationGuides are written in a mix of Markdown, HTML, XML, JSON, INI and a specific language called FHIR Shorthand (FSH). While the former languages are properly supported in many IDEs, the latest is only supported in Visual Studio Code through a dedicated extension.
In this article, I will investigate the possibilities to add syntax highlighting for FHIR Shorthand in the JetBrains IDEs.
Existing resources#
We can investigate how the Visual Studio Code extension implemented the syntax highlighting for FHIR Shorthand. Its sources are available in a GitHub repository, and we can quickly find a promising syntaxes folder that contains a single file: fsh.tmLanguage.json. This file contains the language grammar in a TextMate definition in JSON format, mostly made of keywords and regular expressions (RegExp). That's a good resource in our quest, but may prove itself limited because not all languages can be properly parsed by RegExps (only Type-3 in the Chomsky hierarchy).
FSH files are transpiled to FHIR resources by SUSHI, a Node.js application. The application uses a parser to read the FSH files, which may be useful to us. The sources are also available on GitHub, and after a quick search, we can find two useful files: FSH.g4 and FSHLexer.g4. Those are ANTLR4 grammar files, and they contain the grammar to lex and parse the FSH language. This is a very powerful tools, but it may be overkill for syntax highlighting only.
Doing it the official way#
JetBrains documentation is quite extensive, and after a quick search, we can find a whole documentation section dedicated to Custom Language Support, including information about syntax highlighting. It indicates that it requires the creation of a plugin, the definition of a lexing grammar with JFlex, and the definition of a parsing grammar in an extended NBF syntax (Grammar-Kit).
While JFlex and the Grammar-Kit BNF syntax are not far from the ANTLR4 syntax, it can become quite complex to convert the existing grammar to the JetBrains format because they do not share the same features. I once did such an attempt, and it revealed itself quite inconclusive ( lexer, parser).
Doing it with the ANTLR grammar#
The documentation contains a single sentence about using an ANTLR grammar for the parsing, referring to a project led by ANTLR maintainers: the antlr4-intellij-adaptor library. While the repository is still maintained, there was no new release since 2019, and the documentation is quite scarce. It is not clear if this parser can properly substitute the Grammar-Kit parser, and if it can be used for syntax highlighting.
The ANTLR4 grammar would also need to be modified for a proper parser generation, because the lexer currently discards
some tokens that are not needed for the parser (the comments and whitespaces), whereas JetBrains IDEs require a continuous
token stream. While it should be a minor modification (replacing -> skip by -> channel(HIDDEN)),
the advantage to not maintain a separate grammar than the official project is lost.
Doing it with the TextMate grammar#
JetBrains now supports TextMate grammars through the bundled
TextMate bundles plugin. It does not
require the creation of a plugin, but can be configured in the IDE settings. The immediate issue is that we only
have a .tmLanguage file, but a .tmBundle is needed. There are a
few leads
to convert the former to the latter, but nothing too simple.
Doing it with an LSP integration#
The last possibility I found to implement syntax highlighting for a custom language in JetBrains IDEs is to use the Language Server Protocol (LSP) with a application that implements the protocol for the custom language. I'm not aware of any existing LSP server for FSH, but the current SUSHI application could be a good candidate for developing such a server. The LSP offers a lot of features, but only a small subset is currently supported. It is nonetheless the most future-proof and helpful solution, as the LSP standard is widely adopted and in active development. Other IDEs (including Visual Studio Code) would profit from the LSP server, implementing the FSH logic once and providing its results to all the IDEs.
Personal decision#
While I might try to implement the TextMate grammar in the JetBrains IDEs without creating a plugin to profit faster from syntax highlighting, I will also investigate the ANTLR4-based parser. A proper parser would allow implementing a lot of other features, such as code completion, code navigation, and refactoring thanks to the impressive APIs accessible to the plugin developers.
The LSP solution is also very appealing, but it requires a lot of work and most probably a lot of coordination with the SUSHI developers. It is a long-term solution, and I will keep it in mind for the future.
You can check the progress of the implementation in the GitHub repository.