MLF Language Tests (mlf-lang crate)#

This directory contains crate-specific integration tests for the mlf-lang crate only. These tests verify parsing, validation, and workspace resolution using real MLF files.

For workspace-level integration tests that span multiple crates (codegen, CLI, etc.), see /tests/README.md at the workspace root.

Test Structure#

tests/
├── integration_test.rs      # Test runner that discovers and executes tests
└── lang/                     # Language-level tests
    ├── namespace_imports/    # Import statement behavior
    ├── namespace_exports/    # Export and main resolution
    └── constraints/          # Type constraint validation

Writing Tests#

Each test is a directory containing:

One or more .mlf files - The MLF code to test
expected.json - Expected test outcome

Expected Result Format#

{
  "status": "success"  // or "error"
}

For error tests:

{
  "status": "error",
  "errors": [
    {
      "type": "UndefinedReference",  // Error variant name
      "name": "foo"                   // Optional: specific error details
    }
  ]
}

Test Categories#

Namespace Imports#

Tests various import syntaxes:

namespace_alias - use com.example; creates alias example → com.example
namespace_alias_nested - Multi-level namespace aliasing
implicit_main_import - use place.stream.profile; imports profile type
undefined_namespace - Error when importing non-existent namespace

Namespace Exports#

Tests how types are exported from modules:

implicit_main - Type matching namespace suffix becomes main export

Constraints#

Tests constraint validation and refinement:

refinement_valid - Valid constraint narrowing (maxLength: 100 → 50)
refinement_invalid - Invalid constraint widening (maxLength: 50 → 100)
type_mismatch - Wrong constraint types (string constraints on integers)
enum_refinement - Enum subset validation

Running Tests#

# Run all integration tests
cargo test -p mlf-lang --test integration_test

# Run with output
cargo test -p mlf-lang --test integration_test -- --nocapture

# Run specific test (not directly supported, but you can filter by removing tests)

Test Output#

The test runner outputs:

✓ for passing tests
✗ for failing tests with error details
Summary: X passed, Y failed

Example:

✓ namespace_imports/namespace_alias
✓ namespace_imports/namespace_alias_nested
✓ constraints/refinement_valid
✗ constraints/bad_test: Test expected success but got errors: ...

Results: 8 passed, 1 failed

Scope: mlf-lang Only#

These tests use only the mlf-lang crate (parsing, validation, workspace resolution). They do not test:

Code generation (mlf-codegen, mlf-codegen-*)
CLI commands (mlf-cli)
Error formatting (mlf-diagnostics)
Multi-crate workflows

For those, see workspace-level tests in /tests/ at the project root.

Future Lang-Specific Tests#

More tests to add here (still mlf-lang only):

raw_identifiers - Keyword escaping and edge cases
lenient_parsing - Error recovery and multiple errors
esoteric - Edge cases like circular references, empty unions
annotations - @main, @key, @encoding parsing
unions - Open vs closed unions
circular_references - A refs B refs A

Adding New Tests#

Create a new directory under the appropriate category:
```
mkdir -p tests/lang/your_category/your_test
```

Add MLF files:

echo 'record foo { bar!: string; }' > tests/lang/your_category/your_test/test.mlf

Add expected result:

echo '{"status": "success"}' > tests/lang/your_category/your_test/expected.json

Update derive_namespace_from_path() in integration_test.rs if needed

Run tests:

cargo test -p mlf-lang --test integration_test

Namespace Derivation#

The test runner automatically derives namespaces from file paths:

test.mlf → Uses test directory name
defs.mlf in namespace_alias/ → com.example.defs
actor_defs.mlf in namespace_alias_nested/ → app.bsky.actor.defs

Update derive_namespace_from_path() for custom namespace logic.

Notes#

Tests run against Workspace::with_std() so prelude types are available
Multiple .mlf files in a test directory are loaded in sorted order (test.mlf last)
Error tests check for error type presence, not exact messages
Warnings don't cause test failures

Configure Feed