fix: validate entire array and not single elements

[fedeci]

Description

To be merged after #1279
Closes #1230
Closes #1243
Related #755 (comment)

Maybe it is not the best time for doing this since we just released v7, but I changed my mind about forcing the entire array to comply with the validator.
We provide wildcards which are a great way to validate individual array items.
This PR is born because there are a lot of cases where validating individual items if not requested by the user leads to unexpected results.
Otherwise we can: add a param to body(''), check('')... to explicitly enable/disable array, or keep the current approach.

@gustavohenke wdyt?

To-do list

• I have added tests for what I changed.

• This pull request is ready to merge.

[coveralls]

coverage: 100.0%. remained the same
when pulling d59da5b on fedeci/fix-array-validation
into ffd7994 on master.

[darkman82]

Hi,
currently, to avoid unexpected results with arrays passed when single value is expected, we are forced to:

• use isString() on each field for the query string
• use isString() or not().isArray() for the body according to expected field type

FYI: even the AI (ChatGPT) understood that a standard behavior should not include the use of that workaround on all fields,
as, for example, when expecting a single Boolean value in the body it suggests: v.body('active').isBoolean() while the current approach must be v.body('active').not().isArray().isBoolean().

[fedeci]

I know this is a problem and I want to solve it as fast as possible, on a minor version possibly, even if that requires a breaking change.

p.s. ChatGPT is trained on data of [email protected], so whatever it is saying it is for that version of this library not 7.0.0.

[darkman82]

Hi,
I think this is not a matter of training, even tough v6 and v7 share the same behavior regarding the by-pass problem,
it's the common sense that suggests that validators should expect single values unless specified otherwise.

I discovered this problem by accident in 2019, after I had mistakenly duplicated the parameters in the query string. Then I realized that the issue was even more widespread, and obviously it crashed the entire logic after the validation.

If this was a wanted behavior, it is not obvious to me, nor to the AI.

[gustavohenke]

nit 🤓

Suggested change

	"prettier": "prettier -w ."
	"lint:fix": "prettier -w ."

Then the equivalent command from eslint can be added too, if you like.

[gustavohenke]

Similar feeling to #1279 (comment).