False positive in `array_init`

[liam-i]

New Issue Checklist

• Updated SwiftLint to the latest version
• I searched for existing GitHub issues

Describe the bug

A clear and concise description of what the bug is.

Complete output when running SwiftLint, including the stack trace and command used

$ swiftlint lint
/Users/liam/Documents/Example/AppDelegate.swift:26:45: Array Init Violation: Prefer using `Array(seq)` over `seq.map { $0 }` to convert a sequence into an Array. (array_init)

Environment

• SwiftLint version (run swiftlint version to be sure)? 0.45.0
• Installation method used (Homebrew, CocoaPods, building from source, etc)? CocoaPods
• Paste your configuration file:

only_rules:
  - array_init

• Are you using nested configurations?
  If so, paste their relative paths and respective contents.
  No

• Which Xcode version are you using (check xcodebuild -version)?
  Xcode 13.1

• Do you have a sample that shows the issue? Run echo "[string here]" | swiftlint lint --no-cache --use-stdin --enable-all-rules
  to quickly test if your example is really demonstrating the issue. If your example is more
  complex, you can use swiftlint lint --path [file here] --no-cache --enable-all-rules.

enum BizError: Swift.Error {
}

let reqResult: Result<String, BizError> = .success("")

// False positive in `array_init`
let result: Result<Any, BizError> = reqResult.map { $0 }

print(result)

[aranasaurus]

I spent some time looking at this today, and I could be wrong since this is my first time jumping into the codebase, but it looks like the way to fix this issue would require changing the ArrayInitRule's KindType to SwiftDeclarationKind, but that would make this rule stop being able to detect occurrences of .map instead of Array() as arguments to function calls.

I couldn't see a way to make one ASTRule able to operate on more than one KindType. Maybe there's another way that I'm just missing?

[SimplyDanny]

My take on this is that it cannot be fixed without knowing the type of the object at which .map is called. By default, SwiftLint deals with an AST that is not type checked. I only recently learned that there are a few so called "Analyzer Rules" which have a type checked AST available to work on. They are currently an experimental feature, though.

[jpsim]

They are currently an experimental feature, though.

Analyzer rules have been pretty stable and we could remove the "experimental" label on them.

[jpsim]

#3884

[SimplyDanny]

So, is it okay to convert the array_init rule to an analysis rule to fix the bug, @jpsim? If yes, I could have a look into this.

[jpsim]

You could make a new analyzer rule that does what array_init tries to do, but more reliably.

However, I wouldn’t remove the current array_init rule due because of what I explained in #3884 (comment).

[SimplyDanny]

That's what I also thought after reading your comment. Let me give it a try.

[SimplyDanny]

Can this one be closed now that there is TypesafeArrayInitRule available, @jpsim? I mean, the bug is still valid but certainly won't be fixed.

[jpsim]

Yes let's close this, if someone wants a type-checked version of the rule they can use TypesafeArrayInitRule, albeit requiring the compilation database.