Writing Lints

Preliminary notes on writing lints.

Lint Criteria

Borrowing heavily from the criteria for adding new checks to errorprone, lints should have the following properties.

Dart lints:

  • should be easy to understand. The problem should be obvious once the linter points it out.
  • should have a correspondingly easy fix. For example, “Remove this type annotation”, or “Delete these braces”, not “Introduce a new subclass and override methods A, B, and C.”
  • should have few false positives.

Lint Properties

Every lint has a:

Name. A short name using Dart package naming conventions. Naming is hard but strive to be concise and consistent. Prefer to use the problem as the name, as in the existing lints control_flow_in_finally and empty_catches. Do not start a lint's name with “always”, “avoid”, or “prefer”. Where possible, use existing rules for inspiration and observe the rules of parallel construction.

Description. A short description of the lint, suitable for printing in console output. For example:

[lint] DO name types using UpperCamelCase.

Kind. The first word in the description should identify the kind of lint where kinds are derived from the style guide. In summary:

  • DO guidelines describe practices that should always be followed. There will almost never be a valid reason to stray from them.

  • DON'T guidelines are the converse: things that are almost never a good idea. You'll note there are few of these here. Guidelines like these in other languages help to avoid the pitfalls that appear over time. Dart is new enough that we can just fix those pitfalls directly instead of putting up ropes around them.

  • PREFER guidelines are practices that you should follow. However, there may be circumstances where it makes sense to do otherwise. Just make sure you understand the full implications of ignoring the guideline when you do.

  • AVOID guidelines are the dual to “prefer”: stuff you shouldn't do but where there may be good reasons to on rare occasions.

  • CONSIDER guidelines are practices that you might or might not want to follow, depending on circumstances, precedents, and your own preference.

Detailed Description. In addition to the short description, a lint rule should have more detailed rationale with code examples, ideally good and bad. The style guide is a great source for inspiration. Many style recommendations have been directly translated to lints as enumerated here.

Group. A grouping. For example, Style Guide aggregates style guide derived lints.

Maturity. Rules can be further distinguished by maturity. Unqualified rules are considered stable, while others may be marked EXPERIMENTAL or PROPOSED to indicate that they are under review.

Mechanics

Lints live in the lib/src/rules directory. Corresponding tests live in test_data/rules.

Rule stubs can be generated with the rule.dart helper script and documentation gets generated with doc.dart. Helper scripts can be invoked via dart or grinder (dart run grinder docs --dir=doc_location and dart run grinder rule --name=my_new_rule respectively). Using grinder, for example

$ dart run grinder rule --name=my_new_lint

generates lint and test stubs in lib/src/rules and test_data/rules.

Analyzer APIs

The linter has a close relationship with the analyzer package and at times reaches into non-public APIs. For the most part, we have isolated these references in an analyzer.dart utility library. Whereever possible please use this library to access analyzer internals.

  • If analyzer.dart is missing something please consider opening an issue where we can discuss how best to add it.
  • If you find yourself tempted to make references to analyzer implementation classes also consider opening an issue so that we can see how best to manage the new dependency.

Thanks!

Dart Language Specification

When writing lints, it can be useful to have the Dart Language Specification handy. If you‘re working to support bleeding edge language features, you’ll want the latest draft.

Writing Tests that Depend on Dart SDK Details

Important: when writing tests that use standard dart: libraries, it‘s important to keep in mind that linter tests use a mocked SDK that has only a small subset of the real one. We do this for performance reasons as it’s FAR faster to load a mock SDK into memory than read the real one from disk. If you are writing tests that depend on something in the Dart SDK (for example, an interface such as Iterable), you may need to update SDK mock content located in the package:analyzer test utilities mock_sdk.dart.

Running Tests

The test suite run during the linter's CI, can be run locally like so:

$ dart test/all.dart

alternatively, tests can be run using pub:

$ dart test

Running a single test can be done using the rule_debug helper:

$ dart test/util/rule_debug.dart always_declare_return_types

would only test always_declare_return_types. (This can be very handy if you want to run tests in the VM debugger).

If you simply want to verify a test, you can run it solo in pub:

$ dart test -N always_declare_return_types

Utilities

You'll notice when authoring a new rule that failures cause the AST of the test case to be displayed to stdout. If you simply want to dump the AST of a given compilation unit, you can use the spelunk helper directly. For example:

$ dart tool/spelunk.dart lib/src/rules.dart

would dump the AST of rules.dart.

Performance

For performance reasons rules should prefer implementing NodeLintRule and registering interest in specific AST node types using registry.addXYZ(this, visitor). Avoid overriding visitCompilationUnit() and performing your own full CompilationUnit visits.

Feedback is Welcome!

Details are under active development. Feedback is most welcome!