testing-codegen

name	testing-codegen
description	Guide for testing workflows and code generation commands in Biome. Use when running snapshot tests for lint rules, managing insta snapshots, or regenerating analyzer/parser/formatter code after changes.
compatibility	Designed for coding agents working on the Biome codebase (github.com/biomejs/biome).

Purpose

Use this skill for testing and code generation. Covers snapshot testing with insta and code generation commands.

Prerequisites

1. Install required tools: just install-tools (installs cargo-insta)
2. Install pnpm: curl -fsSL https://get.pnpm.io/install.sh | sh - in repo root
3. Understand which changes require code generation

Common Workflows

Run Tests

# Run all tests
cargo test

# Run tests for specific crate
cd crates/biome_js_analyze
cargo test

# Run specific test
cargo test quick_test

# Show test output (for dbg! macros)
cargo test quick_test -- --show-output

# Run tests with just (uses CI test runner)
just test

# Test specific crate with just
just test-crate biome_cli

Quick Test for Rules

Fast iteration during development:

// In crates/biome_js_analyze/tests/quick_test.rs
// Modify the quick_test function:

const SOURCE: &str = r#"
const x = 1;
var y = 2;
"#;

let rule_filter = RuleFilter::Rule("nursery", "noVar");

Run:

just qt biome_js_analyze

Quick Test for Parser Development

IMPORTANT: Use this instead of building full Biome binary for syntax inspection - it's much faster!

For inspecting AST structure when implementing parsers or working with embedded languages:

// In crates/biome_html_parser/tests/quick_test.rs
// Modify the quick_test function:

#[test]
pub fn quick_test() {
  let code = r#"<button on:click={handleClick}>Click</button>"#;

  let source_type = HtmlFileSource::svelte();
  let options = HtmlParserOptions::from(&source_type);
  let root = parse_html(code, options);
  let syntax = root.syntax();

  dbg!(&syntax, root.diagnostics(), root.has_errors());
}

Run:

just qt biome_html_parser

The dbg! output shows the full AST tree structure, helping you understand:

• How directives/attributes are parsed (e.g., HtmlAttribute vs SvelteBindDirective)
• Whether values use HtmlString (quotes) or HtmlTextExpression (curly braces)
• Token ranges and offsets needed for proper snippet creation
• Node hierarchy and parent-child relationships

Snapshot Testing with Insta

Run tests and generate snapshots:

cargo test

Review generated/changed snapshots:

# Interactive review (recommended)
cargo insta review

# Accept all changes
cargo insta accept

# Reject all changes
cargo insta reject

# Review for specific test
cargo insta review --test-runner nextest

Snapshot commands:

• a - accept snapshot
• r - reject snapshot
• s - skip snapshot
• q - quit

Pruning Orphaned Snapshots

When tests are removed or renamed, their old snapshot files become orphaned. Never delete snapshot files manually with rm — always use insta's built-in pruning:

# Delete unreferenced snapshots after a successful test run
cargo insta test --unreferenced delete -p <crate_name>

# Or scoped to specific tests
cargo insta test --unreferenced delete -p biome_cli --test main -- "handle_vue"

This runs the tests first, then deletes any .snap files that no test references. It is the only safe way to clean up snapshots — manual rm risks deleting snapshots that are still needed or creating git conflicts.

Test Lint Rules

# Test specific rule by name
just test-lintrule noVar

# Run from analyzer crate
cd crates/biome_js_analyze
cargo test

Create Test Files

Single file tests - Place in tests/specs/{group}/{rule}/ under the appropriate *_analyze crate for the language:

tests/specs/nursery/noVar/
├── invalid.js           # Code that should generate diagnostics
├── valid.js             # Code that should not generate diagnostics
└── options.json         # Optional: rule configuration

File and folder naming conventions (IMPORTANT):

• Use valid or invalid in file names or parent folder names to indicate expected behaviour.
• Files/folders with valid in the name (but not invalid) are expected to produce no diagnostics.
• Files/folders with invalid in the name are expected to produce diagnostics.
• When testing cases inside a folder, prefix the name of folder using valid/invalid e.g. validResolutionReact/invalidResolutionReact

tests/specs/nursery/noShadow/
├── invalid.js                     # should generate diagnostics
├── valid.js                       # should not generate diagnostics
├── validResolutionReact/
└───── file.js              # should generate diagnostics
   └── file2.js             # should not generate diagnostics

Multiple test cases - Use .jsonc files with arrays:

// tests/specs/nursery/noVar/invalid.jsonc
[
  "var x = 1;",
  "var y = 2; var z = 3;",
  "for (var i = 0; i < 10; i++) {}"
]

Test-specific options - Create options.json:

{
  "linter": {
    "rules": {
      "nursery": {
        "noVar": {
          "level": "error",
          "options": {
            "someOption": "value"
          }
        }
      }
    }
  }
}

Top-Level Comment Convention (REQUIRED)

Every test spec file must begin with a top-level comment declaring whether it expects diagnostics. The test runner (assert_diagnostics_expectation_comment in biome_test_utils) enforces this and panics if the rules are violated.

Write the marker text using whatever comment syntax the language under test supports. For languages that do not support comments at all, rely on the file/folder naming convention (valid/invalid) instead.

For files whose name contains "valid" (but not "invalid"):

The comment is mandatory — the test panics if it is absent.

For files whose name contains "invalid" (or other names):

The comment is strongly recommended and is also enforced when present: if the comment says should generate diagnostics but no diagnostics appear, the test panics.

Rules enforced by the test runner:

File name contains	Comment present?	Behaviour
"valid" (not "invalid")	should not generate diagnostics	Passes if no diagnostics
"valid" (not "invalid")	should generate diagnostics	Passes if diagnostics present
"valid" (not "invalid")	absent	PANIC — comment is mandatory
"invalid" or neutral name	should not generate diagnostics	Passes if no diagnostics
"invalid" or neutral name	should generate diagnostics	Passes if diagnostics present
"invalid" or neutral name	absent	No enforcement (but add it anyway)

Important details:

• The comment is found by scanning the entire file's leading trivia — it does not have to be literally the first token, but putting it at the very top (line 1) is the established convention.
• Fixture/support files (e.g. foo.js, bar.ts) that don't contain "valid" or "invalid" in their name do not require a comment, since they are not considered "valid test files" by the runner.
• Files excluded from comment enforcement regardless of name: .snap, .json, .jsonc.

HTML-ish files (.vue, .svelte, .astro, .html):

These files are analyzed via the workspace-based test path (analyze_with_workspace in biome_test_utils), which checks the expectation comment by scanning the raw file content (not the parsed AST trivia). Use an HTML comment at the very top of the file:

<!-- should not generate diagnostics -->
<script setup lang="ts">
const x = 1;
</script>
<template>{{ x }}</template>

<!-- should generate diagnostics -->
<script>
debugger;
</script>

The same rules apply: valid files must have the comment, invalid files should have it. Do not place the comment inside <script> — put it at the top level of the file as an HTML comment.

Code Generation Commands

After modifying analyzers/lint rules (during development):

just gen-rules          # Updates rule registrations in *_analyze crates
just gen-configuration  # Updates configuration schemas

These lightweight commands generate enough code to compile and test without errors.

Full analyzer codegen (optional — CI autofix handles this):

just gen-analyzer

This is a composite command that runs gen-rules, gen-configuration, gen-migrate, gen-bindings, lint-rules, and format. You typically don't need to run this locally — the CI autofix job does it automatically when you open a PR.

After modifying grammar (.ungram files):

# Specific language
just gen-grammar html

# Multiple languages
just gen-grammar html css

# All languages
just gen-grammar

After modifying formatters:

just gen-formatter html

After modifying configuration:

just gen-bindings

Generates TypeScript types and JSON schema.

Full codegen (rarely needed):

just gen-all

Before committing:

just ready

Runs full codegen + format + lint (takes time).

Or run individually:

just f  # Format Rust and TOML
just l  # Lint code

Run Doctests

Test code examples in documentation comments:

just test-doc

Debugging Tests

Use dbg!() macro in Rust code:

fn some_function() -> &'static str {
  let some_variable = "debug_value";
  dbg!(&some_variable);  // Prints during test
  some_variable
}

Run with output:

cargo test test_name -- --show-output

Tips

• Snapshot organization: Group by feature/rule in separate directories
• Test both valid and invalid: Create both valid.js and invalid.js files
• Options per folder: options.json applies to all tests in that folder
• .jsonc arrays: Use for multiple quick test cases in script context (no imports/exports)
• Code generation order: Grammar → Analyzer → Formatter → Bindings
• CI compatibility: Use just commands when possible (matches CI)
• Snapshot review: Always review snapshots carefully - don't blindly accept
• Test performance: Use #[ignore] for slow tests, run with cargo test -- --ignored
• Parser inspection: Use just qt <package> to run quick_test and inspect AST, NOT full Biome builds (much faster)

For general Biome development tips (string extraction, borrow checker patterns, legacy syntax), see the biome-developer skill.

Common Test Patterns

// Snapshot test in rule file
#[test]
fn test_rule() {
  assert_lint_rule! {
        noVar,
        invalid => [
            "var x = 1;",
            "var y = 2;",
        ],
        valid => [
            "const x = 1;",
            "let y = 2;",
        ]
    }
}

// Quick test pattern
#[test]
#[ignore]  // Uncomment when using
fn quick_test() {
  const SOURCE: &str = r#"
        var x = 1;
    "#;

  let rule_filter = RuleFilter::Rule("nursery", "noVar");
  // Test runs with this configuration
}

Code Generation Dependencies

When you modify...	Run during dev...	Full (optional, CI does this)
.ungram grammar files	just gen-grammar <lang>	—
Lint rules in *_analyze	just gen-rules && just gen-configuration	just gen-analyzer
Formatter in *_formatter	just gen-formatter <lang>	—
Configuration types	just gen-bindings	—
Before committing	just f && just l	—
Full rebuild	—	just gen-all (slow)

References

• Main testing guide: CONTRIBUTING.md § Testing
• Insta documentation: https://insta.rs
• Analyzer testing: crates/biome_analyze/CONTRIBUTING.md § Testing
• Changeset guide: ../changeset/SKILL.md