This plugin aims to deliver a feature-rich, high quality, high performance IDE experience for Jsonnet. It is designed to work in large repositories with deeply nested imports.
Features
Syntax Highliting and Basic Editing features
Snippets
Custom linting code that is able to deal with large codebases
The analysis code is optimized for real-time linting, and can return in <5ms when the normal linter could take minutes.
Formatting
Delta text update support for efficient editing
Designed to remain performant in large repos with many files open
Automatic detection of bazel-bin for generated files
Type and Value Deduction
Supports imported files
Able to follow variables, function return values, and array/object indexing
Autocomplete
Stdlib support with documentation and typed signatures
Scoped variable completion
Dotted autocomplete
Template object field completion
Import path completion for files
Go to Definition
Can follow definitions in other files, including json files
Hover Information
Function Signature Help
AST Recovery
The LSP is able recover common syntax issues while typing (like a missing semicolon) for a smoother experience
Missing Features
These are features I consider pretty important that are still missing:
More type deduction logic and analysis
More IDE options (options for linting, jpath, etc)
First class multi-dimension jsonnet support
Development
To develop the LSP, change the jsonnet.lsp.binaryPath setting to the runlsp.sh script in the root. Reloading the LSP in vscode (shift+cmd+p -> jsonnet: reload language server) will rebuild the server.
To develop the client, open editor/code in vscode, and hit F5 to open a debug build of the client. Generally developing the LSP does not need a debug version of the client.
Release
Github actions are setup to publish to OpenVSX automatically when a release is created
The LSP binaries are bundled into the .vsix extension, to help reduce bugs from version skew and simplify installation.
