Ameba 1.4.0 has been released

This release brings 6 new rules, colored severity output, Docker images available on GHCR, and several internal refinements. Check out the release notes to see a full scope of changes.

New rules

Lint/LiteralAssignmentsInExpressions

A rule that detects literal assignments used in expressions — a common source of bugs when == was intended instead of =.

if x = 42
  puts "matched!"
end

if x == 42
  puts "matched!"
end

Lint/UnusedBlockArgument

A dedicated rule for reporting unused block arguments. Previously this was partially covered by Lint/UnusedArgument, but now it gets its own rule with clearer messages and autocorrection support.

def foo(&block)
  # `block` is never called
end

Use &_block or the anonymous & to silence it.

Lint/MissingBlockArgument

Reports yielding methods that lack an explicit block argument — making it harder for callers to pass a named block.

def foo
  yield 42
end

Adding & (or &block) makes the block explicit and introspectable.

Lint/Formatting

Integrates crystal tool format checks directly as an Ameba rule. You can now replace your separate crystal tool format --check CI step with an Ameba rule, keeping all linting unified.

Style/ParenthesesAroundCondition

Renamed from the former Style/RedundantParentheses with a more flexible behavior. This rule enforces consistent use of parentheses around conditions in if/unless/case/while/until. It replaces the old ExcludeAssignments option with AllowSafeAssignment — when set, assignments in conditions are expected to be wrapped in parentheses.

if (x > 0)
  puts x
end

if x > 0
  puts x
end

Style/QueryBoolMethods

Reports boolean properties defined with property or getter instead of their ?-suffixed variants.

class Person
  property deceased = false
end

class Person
  property? deceased = false
end

Breaking changes

Boolean properties are now query-style methods

Boolean rule properties now generate query-style methods (enabled? instead of enabled). This affects the public API for rule configuration — if you reference rule properties like rule.enabled or rule.excluded in custom code or extensions, you’ll need to update them to rule.enabled? and rule.excluded? respectively. This is a more idiomatic Crystal pattern and makes the API consistent.

Default severity for Performance and Metrics groups changed to warning

Rules in the Performance and Metrics groups now default to warning severity instead of convention. This makes them stand out more in the output, reflecting their higher importance for code quality.

Tweaked issue location reporting

Issue locations now point to the full affected node range (e.g. &block) instead of just the first character (&). This makes the highlighted source locations more accurate and useful for autocorrection.

Removed buggy autocorrect from Performance/AnyInsteadOfEmpty

The autocorrect for Performance/AnyInsteadOfEmpty was producing invalid code in several edge cases. It has been removed to avoid generating incorrect suggestions — the rule still reports issues, but corrections must now be applied manually.

Autocorrection improvements

• Style/UnlessElse — now has autocorrect support, automatically rewriting unless/else blocks to if/else.
• Lint/UnusedArgument and Lint/UnusedBlockArgument — extended with corrections that prefix unused arguments with _ or replace named block arguments with anonymous &.

Other notable changes

• Colored severity output — each severity level now has a dedicated color in the terminal output, making issues easier to scan at a glance.
• Docker images on GHCR — Ameba is now published as a Docker image to the GitHub Container Registry, instead of Docker Hub.
• Anonymous block arguments — Lint/UnusedArgument no longer reports anonymous block arguments (&), delegating that to the new Lint/UnusedBlockArgument rule.
• Infrastructure updates — CI definitions were refreshed, Dependabot was added for automated dependency updates, and OpenSSL was removed from the Docker image to reduce its size.

Check out the full changelog for a complete list of changes.