The expressions length(list) == 0 and list == [] have slightly different behaviour when the variable list isn’t actually a proper list:
length(list) will fail if the input is not a proper list and this means that the whole clause will fail - even it might seem it would succeed otherwise - e.g. length(value) >= 0 used as a guard would fail for input [1 | 2].
list == [] will just return false for anything that is not an empty list.
Because of this difference, such a conversion is not safe to be performed automatically by the compiler - it’s extremely hard to prove it would be safe. A compiler has to be 100% sure an optimisation is safe to perform it. On the other hand, in 99% of cases it actually is safe and the programmer can switch it.
Thanks for the response, I hadn’t considered improper lists.
The problem, however, as I see it is that if I do (somehow) have an improper list:
foo = [1 | 2]
and I run the following checks:
is_list(foo)
Enum.empty?(foo)
I come to the incorrect conclusion that I have a list with one or more items.
Performance implications aside, it almost seems better/preferable to do length(foo) == 0, because at least that blows up with an argument error. Neither of the Credo suggestions do.
Improper lists are still lists, this isn’t an incorrect conclusion. If you want to make sure something is a proper list, do a dedicated check for that. If you want to know if a list is non empty, checking equality with [] is the clearest thing to do.
RE: “… checking equality with [] is the clearest thing to do.” I could have just as easily said foo = [] instead of Enum.empty?(foo) in my example. As Credo recommends either, they need to be functionally equivalent.
RE: “If you want to make sure something is a proper list, do a dedicated check for that.”
How do you check for an improper list? (i.e. differentiate [1 | 2] from [1 | [2, 3]])