PSPossibleIncorrectComparisonWithNull rule is questionable

[KirkMunro]

I disagree with the use of the PSPossibleIncorrectComparisonWithNull rule. Script authors need to understand how comparisons against $null work when it comes to collections, however there are scenarios where you may want to use a comparison operator with a collection against $null. For example, if you have some null values in a collection, you can remove them like this:

$d = @('a',$null,'b',$null,'c') -ne $null
$d.Count # returns 3

In that scenario, you don't want to $null on the left hand side of the -ne operator, because that is a totally different function.

Further, unlike other languages, it is not possible to unintentionally cause assignment when using the PowerShell comparison operators. For example, in C++, I compare values using ==. But if I mistakenly type a single '=', then I'm performing an assignment. In PowerShell, you can't do anything to -eq/-ne to make them do assignment.

The only potential risk with $null on the right hand side is when the item on the left hand side is an array. If it is not, there is no need to make the change. Why then would PSPossibleIncorrectComparisonWithNull raise what is essentially a compiler warning for values that are not arrays?

If this rule truly must stick around, then I would like it changed such that it only generates a warning when it either does not know the type of the object on the left hand side of the equality comparison operator or when the type of the object on the left hand side is a collection.

[imfrancisd]

This rule can also be limited to Boolean expressions so that this rule has a higher chance of being correct.

For example, with this line:

    $a = $someArray -ne $null

we can't know for sure if "$someArray -ne $null" is expected to be an array or a Boolean.

However, with this line:

    $a = ($someArray -ne $null) -and $b

or this line:

    if ($someArray -ne $null)

we know for sure that "$someArray -ne $null" is supposed to be a Boolean (because PowerShell will turn it into a Boolean if it isn't), so the rule has a higher chance of being correct.

[KirkMunro]

I like that idea. That sounds like a better alternative than the current implementation.

[yutingc]

Thanks @KirkMunro and @imfrancisd , we will modify the rule accordingly.

[quoctruong]

Fixed with #250. The rule is now raised only if the object on the LHS has an unspecified type or is an array.