Reject use of export_all

[kbaird]

Keeping export_all in place after initial development work is typically considered a bad practice.

A rule that complains when that is still present in a file might be useful.

[igaray]

👍

[elbrujohalcon]

👍 and can you add a similar issue in the guidelines, please?

[kbaird]

inaka/erlang_guidelines#60, although it appears to have some errors from gadget/compiler and gadget/xref.

[bullno1]

I guess, using a -ifdef(TEST) to conditionally export_all should be acceptable?

[elbrujohalcon]

@bullno1 you were correct somewhere else where you pointed out that the compiler itself can give you this same warning. Therefore, this rule should not be default and the implementation of this rule should be considered a very very low priority.
In any case, I wouldn't add any exception like that for ifdef, if you want to do that, just don't include this rule in your elvis.config or just ignore the module where you do that altogether.

[paulo-ferraz-oliveira]

I guess, using a -ifdef(TEST) to conditionally export_all should be acceptable?

I've used it and find no issue with it, unless the consumer is being smart and deploying production code with TEST defined. :D

To build on what @elbrujohalcon commented, I find this rule won't bring relevant added value. Even more so since for a few Erlang/OTP versions already, warn_export_all is on by default, so you explicitly have to turn the warning off.

[elbrujohalcon]

Closing this ticket.