feat(eslint-plugin): [strict-bool-expr] add fixes and suggestions

[phaux]

Added automatic fixes and suggestions to make this rule more easy to use.

I added following fixes and suggestions for particular types in boolean context (also added to docs):

• boolean - Always allowed - no fix needed.
• string - (when allowString is false) - Provides following suggestions:
  • Change condition to check string's length (str → str.length > 0)
  • Change condition to check for empty string (str → str !== "")
  • Explicitly cast value to a boolean (str → Boolean(str))
• number - (when allowNumber is false) - Provides following suggestions:
  • Change condition to check for 0 (num → num !== 0)
  • Change condition to check for NaN (num → !Number.isNaN(num))
  • Explicitly cast value to a boolean (num → Boolean(num))
• object | null | undefined - (when allowNullableObject is false) - Provides autofix:
  • Change condition to check for null/undefined (maybeObj → maybeObj != null)
• boolean | null | undefined - Provides autofix:
  • Explicitly treat nullish value the same as false (maybeBool → maybeBool ?? false)
• string | null | undefined - Provides following suggestions:
  • Change condition to check for null/undefined (maybeStr → maybeStr != null)
  • Explicitly treat nullish value the same as an empty string (maybeStr → maybeStr ?? "")
  • Explicitly cast value to a boolean (maybeStr → Boolean(maybeStr))
• number | null | undefined - Provides following suggestions:
  • Change condition to check for null/undefined (maybeNum → maybeNum != null)
  • Explicitly treat nullish value the same as 0 (maybeNum → maybeNum ?? 0)
  • Explicitly cast value to a boolean (maybeNum → Boolean(maybeNum))
• any and unknown - Provides following suggestions:
  • Explicitly cast value to a boolean (value → Boolean(value))

Edit: updated order and messages to match the actual ones

---

I used the word "explicitly" where it doesn't change the runtime behavior, even when types are wrong.

The autofix for nullable object theoretically changes the behavior (e.g. if you somehow get empty string or 0 instead of null or object), but it's not enabled by default and there's only one obvious way of handling it if enabled. In practice it shouldn't make a difference when you compare object to null vs implicitly cast to boolean.

The second autofix for nullable boolean doesn't change the behavior and it's usually what the programmer wants. That's why I decided to make it automatic instead of suggestion.

I left some TODO comments about suggestions which are wrong for bigint (it's treated the same as number in the rule). They shouldn't cause a huge breakage when applied (in worst case TypeScript should report error after applying the suggestions). I think they can be fixed separately in a followup PR.

[typescript-eslint]

Thanks for the PR, @phaux!

typescript-eslint is a 100% community driven project, and we are incredibly grateful that you are contributing to that community.

The core maintainers work on this in their personal time, so please understand that it may not be possible for them to review your work immediately.

Thanks again!

---

🙏 Please, if you or your company is finding typescript-eslint valuable, help us sustain the project by sponsoring it transparently on https://opencollective.com/typescript-eslint. As a thank you, your profile/company logo will be added to our main README which receives thousands of unique visitors per day.

[codecov]

Codecov Report

Merging #2847 (03fe065) into master (c32f803) will increase coverage by 0.04%.
The diff coverage is 98.11%.

@@            Coverage Diff             @@
##           master    #2847      +/-   ##
==========================================
+ Coverage   92.89%   92.94%   +0.04%     
==========================================
  Files         315      316       +1     
  Lines       10704    10802      +98     
  Branches     3036     3065      +29     
==========================================
+ Hits         9944    10040      +96     
  Misses        344      344              
- Partials      416      418       +2     

Flag	Coverage Δ
unittest	92.94% <98.11%> (+0.04%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...ackages/eslint-plugin/src/util/getWrappingFixer.ts	93.93% <93.93%> (ø)
...int-plugin/src/rules/strict-boolean-expressions.ts	98.70% <100.00%> (+0.81%)	⬆️
.../eslint-plugin/src/rules/member-delimiter-style.ts	94.91% <0.00%> (+0.47%)	⬆️

[bradzacher]

Logic itself LGTM.
One thing that needs to be addressed

---

I agree with all of the fixes except for the autofix for nullable boolean.

Explicitly treat nullish value the same as false (maybeBool → maybeBool ?? false)

two reasons:

1. I think that it's a stylistic choice here and a lot of people won't like the autofix to ??.
2. (the big one which blocks this) Not everyone is on TS 3.7, which would make this a breaking change because we'd be autofixing to unsupported syntax and break their code.

Suggested solution:
make it a suggestion fixer with the following choices:

• maybeBool ?? false
• maybeBool === true

---

Side note for a (potential) future enhancement (not saying lets do it now - god no! - just saying someone might like it in the future).

Additional fixer for the verbose and explicit:
x !== null && x !== undefined
and the super verbose but super explicit and always safe:
x !== null && typeof x !== 'undefined'

[phaux]

Done!

I changed the suggestions for nullable boolean as you suggested.

I also added a special case for number in boolean context when the number is array.length in which case it provides autofix instead of suggestion.

In addition, I moved getWrappingFixer to utils because it might be useful in other rules.

Then I also fixed a bug with getWrappingFixer so that it adds a semicolon if it detects that it inserted parens at the beginning of an ExpressionStatement.

[bradzacher]

thanks so much for your work on this!
This is a really valuable set of changes.