It is possible that a new import_type directive (⚠️ a previous link was removed - though GitHub doesn't seem to update the target - so that this rule suggestion is not poorly interpreted as an "attack" on the functionality...) lands in OTP soon (27 or later, I imagine). While it may have uses in some specific cases, namely, as the proposer mentions, large code bases, it can easily be avoided in smaller code bases. elvis_core shall propose this rule as "enabled by default".
Reasoning
There seems to be some value in the use of the directive:
for large code bases, where the use of "externally-defined" types is done by means of an auxiliary (internal) type
for potential removal of boilerplate, also potentially at the expense of more "difficult to reason about" code: where does the type come from? (it might be Ok if you use a language server, but not all developers do, and some might actually prefer the clarify of having Module:<type>)
but it's probably best to start by avoiding it (as we propose avoiding import). elvis_core proposes, doesn't force, and one can always go about using disable to negative the effect of the new rule 😄.
Refactoring Proposal
Instead of -import_type(m, [t/0])., then t, simply use m:t().
This would potentially be a breaking change, depending on how we consider that to be (we're not changing the interface, but adding a rule, which may cause issues in flows with warnings_as_errors) but the MIGRATION.md file hasn't been updated in a while, either.
Name
no_import_typeBrief Description
It is possible that a new
import_typedirective (⚠️ a previous link was removed - though GitHub doesn't seem to update the target - so that this rule suggestion is not poorly interpreted as an "attack" on the functionality...) lands in OTP soon (27 or later, I imagine). While it may have uses in some specific cases, namely, as the proposer mentions, large code bases, it can easily be avoided in smaller code bases.elvis_coreshall propose this rule as "enabled by default".Reasoning
There seems to be some value in the use of the directive:
Module:<type>)but it's probably best to start by avoiding it (as we propose avoiding
import).elvis_coreproposes, doesn't force, and one can always go about usingdisableto negative the effect of the new rule 😄.Refactoring Proposal
Instead of
-import_type(m, [t/0])., thent, simply usem:t().This would potentially be a breaking change, depending on how we consider that to be (we're not changing the interface, but adding a rule, which may cause issues in flows with
warnings_as_errors) but the MIGRATION.md file hasn't been updated in a while, either.