Contents

Customizing static analysis

Contents keyboard_arrow_down keyboard_arrow_up

Static analysis allows you to find problems before executing a single line of code. It’s a powerful tool used to prevent bugs and ensure that code conforms to style guidelines.

With the help of the analyzer, you can find simple typos. For example, perhaps an accidental semicolon made its way into an if statement:

void increment() {
  if (count < 10) ;
  count++;
}

If properly configured, the analyzer points to the semicolon and produces the following warning:

info - example.dart:9:19 - Unnecessary empty statement. Try removing the empty statement or restructuring the code. - empty_statements

The analyzer can also help you find more subtle problems. For example, perhaps you’ve forgotten to close a sink method:

var controller = StreamController<String>();
info - Unclosed instance of 'Sink'. Try invoking 'close' in the function in which the 'Sink' was created. - close_sinks

In the Dart ecosystem, the Dart Analysis Server and other tools use the analyzer package to perform static analysis.

You can customize static analysis to look for a variety of potential problems, including errors and warnings specified in the Dart language spec. You can also configure linter rules, to ensure that your code complies with the Dart Style Guide and other suggested guidelines in Effective Dart. Tools such as dart analyze, flutter analyze, and IDEs and editors use the analyzer package to evaluate your code.

This document explains how to customize the behavior of the analyzer using either an analysis options file or comments in Dart source code. If you want to add static analysis to your tool, see the analyzer package docs and the Analysis Server API Specification.

The analysis options file

Place the analysis options file, analysis_options.yaml, at the root of the package, in the same directory as the pubspec file.

Here’s a sample analysis options file:

include: package:lints/recommended.yaml

analyzer:
  exclude: [build/**]
  language:
    strict-casts: true
    strict-raw-types: true

linter:
  rules:
    - cancel_subscriptions

The sample illustrates the most common top-level entries:

If the analyzer can’t find an analysis options file at the package root, it walks up the directory tree, looking for one. If no file is available, the analyzer defaults to standard checks.

Consider the following directory structure for a large project:

project root contains analysis_options.yaml (#1) and 3 packages, one of which (my_package) contains an analysis_options.yaml file (#2).

The analyzer uses file #1 to analyze the code in my_other_package and my_other_other_package, and file #2 to analyze the code in my_package.

Enabling stricter type checks

If you want stricter static checks than the Dart type system requires, consider enabling the strict-casts, strict-inference, and strict-raw-types language modes:

analyzer:
  language:
    strict-casts: true
    strict-inference: true
    strict-raw-types: true

You can use the modes together or separately; all default to false.

strict-casts: <bool>
A value of true ensures that the type inference engine never implicitly casts from dynamic to a more specific type. The following valid Dart code includes an implicit downcast from the dynamic value returned by jsonDecode to List<String> that could fail at runtime. This mode reports the potential error, requiring you to add an explicit cast or otherwise adjust your code.
void foo(List<String> lines) {
  ...
}

void bar(String jsonText) {
  foo(jsonDecode(jsonText)); // Implicit cast
}
error - The argument type 'dynamic' can't be assigned to the parameter type 'List<String>'. - argument_type_not_assignable
strict-inference: <bool>
A value of true ensures that the type inference engine never chooses the dynamic type when it can’t determine a static type. The following valid Dart code creates a Map whose type argument cannot be inferred, resulting in an inference failure hint by this mode:
final lines = {}; // Inference failure
lines['Dart'] = 10000;
lines['C++'] = 'one thousand';
lines['Go'] = 2000;
print('Lines: ${lines.values.reduce((a, b) => a + b)}'); // Runtime error
warning - The type argument(s) of 'Map' can't be inferred - inference_failure_on_collection_literal
strict-raw-types: <bool>
A value of true ensures that the type inference engine never chooses the dynamic type when it can’t determine a static type due to omitted type arguments. The following valid Dart code has a List variable with a raw type, resulting in a raw type hint by this mode:
List numbers = [1, 2, 3]; // List with raw type
for (final n in numbers) {
  print(n.length); // Runtime error
}
warning - The generic type 'List<dynamic>' should have explicit type arguments but doesn't - strict_raw_type

Enabling and disabling linter rules

The analyzer package also provides a code linter. A wide variety of linter rules are available. Linters tend to be nondenominational—rules don’t have to agree with each other. For example, some rules are more appropriate for library packages and others are designed for Flutter apps. Note that linter rules can have false positives, unlike static analysis.

Enabling Dart team recommended linter rules

The Dart team provides two sets of recommended linter rules in the lints package:

Core rules
Help identify critical issues that are likely to lead to problems when running or consuming Dart code. All code should pass these linter rules. Packages that are uploaded to pub.dev have a package score that’s based in part on passing these rules.
Recommended rules
Help identify additional issues that may lead to problems when running or consuming Dart code, and enforce a single, idiomatic style and format. We recommend that all Dart code use these rules, which are a superset of the core rules.

To enable either set of lints, add the lints package as a dev dependency:

$ dart pub add --dev lints

Then edit your analysis_options.yaml file to include your preferred rule set:

include: package:lints/<RULE_SET>.yaml

For example, you can include the recommended rule set like this:

include: package:lints/recommended.yaml

Enabling individual rules

To enable a single linter rule, add linter: to the analysis options file as a top-level key, followed by rules: as a second-level key. On subsequent lines, specify the rules that you want to apply, prefixed with dashes (the syntax for a YAML list). For example:

linter:
  rules:
    - always_declare_return_types
    - cancel_subscriptions
    - close_sinks
    - comment_references
    - one_member_abstracts
    - only_throw_errors
    - package_api_docs
    - prefer_final_in_for_each
    - prefer_single_quotes

Disabling individual rules

If you include an analysis options file such as the one in lints, you might want to disable some of the included rules. Disabling individual rules is similar to enabling them, but requires the use of a map rather than a list as the value for the rules: entry, so each line should contain the name of a rule followed by either : false or : true.

Here’s an example of an analysis options file that uses all the recommended rules from lints except avoid_shadowing_type_parameters. It also enables the lint await_only_futures:

include: package:lints/recommended.yaml

linter:
  rules:
    avoid_shadowing_type_parameters: false
    await_only_futures: true

Enabling analyzer plugins (experimental)

The analyzer has experimental support for plugins. These plugins integrate with the analyzer to add functionality such as new diagnostics, quick fixes, and custom code completion. You can enable only one plugin per analysis_options.yaml file. Enabling an analyzer plugin increases how much memory the analyzer uses.

Don’t use analyzer plugins if your situation meets either of the following conditions:

  • You use a development machine with less than 16 GB of memory.
  • You use a mono-repo with more than 10 pubspec.yaml and analysis_options.yaml files.

You can find a few analyzer plugins on pub.dev.

To enable a plugin:

  1. Add the package containing the plugin as a dev dependency.

    $ dart pub add --dev <your_favorite_analyzer_plugin_package>

  2. Edit your analysis_options.yaml file to enable the plugin.

    analyzer:
  plugins:
    - your_favorite_analyzer_plugin_package

    To indicate specific plugin functionality to enable, such as new diagnostics, additional setup might be required.

Excluding code from analysis

Sometimes it’s OK for some code to fail analysis. For example, you might rely on code generated by a package that you don’t own—the generated code works, but produces warnings during static analysis. Or a linter rule might cause a false positive that you want to suppress.

You have a few ways to exclude code from analysis:

  • Exclude entire files from analysis.
  • Stop specific non-error rules from being applied to individual files.
  • Stop specific non-error rules from being applied to individual lines of code.

You can also disable specific rules for all files or change the severity of rules.

Excluding files

To exclude files from static analysis, use the exclude: analyzer option. You can list individual files, or use glob syntax:

analyzer:
  exclude:
    - lib/client.dart
    - lib/server/*.g.dart
    - test/_data/**

Suppressing rules for a file

To ignore a specific non-error rule for a specific file, add an ignore_for_file comment to the file:

// ignore_for_file: unused_local_variable

This acts for the whole file, before or after the comment, and is particularly useful for generated code.

To suppress more than one rule, use a comma-separated list:

// ignore_for_file: unused_local_variable, duplicate_ignore, dead_code

To suppress all linter rules, add a type=lint specifier:

// ignore_for_file: type=lint

Suppressing rules for a line of code

To suppress a specific non-error rule on a specific line of code, put an ignore comment above the line of code. Here’s an example of ignoring code that causes a runtime error, as you might do in a language test:

// ignore: invalid_assignment
int x = '';

To suppress more than one rule, supply a comma-separated list:

// ignore: invalid_assignment, const_initialized_with_non_constant_value
const x = y;

Alternatively, append the ignore rule to the line that it applies to:

int x = ''; // ignore: invalid_assignment

Customizing analysis rules

Each analyzer error code and linter rule has a default severity. You can use the analysis options file to change the severity of individual rules, or to always ignore some rules.

The analyzer supports three severity levels:

info
An informational message that doesn’t cause analysis to fail. Example: dead_code
warning
A warning that doesn’t cause analysis to fail unless the analyzer is configured to treat warnings as errors. Example: invalid_null_aware_operator
error
An error that causes analysis to fail. Example: invalid_assignment

Ignoring rules

You can ignore specific analyzer error codes and linter rules by using the errors: field. List the rule, followed by : ignore. For example, the following analysis options file instructs the analysis tools to ignore the TODO rule:

analyzer:
  errors:
    todo: ignore

Changing the severity of rules

You can globally change the severity of a particular rule. This technique works for regular analysis issues as well as for lints. For example, the following analysis options file instructs the analysis tools to treat invalid assignments as warnings and missing returns as errors, and to provide information (but not a warning or error) about dead code:

analyzer:
  errors:
    invalid_assignment: warning
    missing_return: error
    dead_code: info

Resources

Use the following resources to learn more about static analysis in Dart: