[erlang-questions] dialyzer: user-defined types just synonyms?

[Dmitry Belyaev]

I saw those options already but they don't do what I'd like to see.

I have very complex system. And I am _heavily_ refactoring it. And 
thought dialyzer would be of great help for me.
Now I see many functions that long ago had (and may will have in the 
future) wider result type than now dialyzer infers. I get many messages 
like those (currently I grep them from output):

.*: The pattern .* can never match since previous clauses completely 
covered the type .*
.*: The pattern .* can never match the type .*
.*: The variable .* can never match since previous clauses completely 
covered the type .*

I don't want to change this code (it may be needed in future). I'd 
better make result type of function wider, so dialyzer won't warn about 
superflous (by this moment) code.

Tobias Lindahl wrote:
> 2010/4/22 Dmitry Belyaev <rumata-estor@REDACTED>:
>   
>> It would be nice if it would intersect spec parameters and check spec result
>> contain what dialyzer infers.
>>
>> So, if I have
>> -spec function(A) -> B.
>> and dialyzer finds that real type as function(C) -> D then it would check
>> that C contains A and B contains D and its final type would be
>> function(C) -> B.
>>
>>     
>
> Dialyzer checks that there is a non-empty intersection between the
> spec type and the infered type. In other words if the spec is possible
> at all. If not, you will get a warning.
>
> You might want to have a look at  the options -Wunderspecs and
> -Woverspecs. They give you some more information about how the specs
> relate to the inferred types. They are off by default since you
> typically get a lot of warnings from them.
>
> Tobias
>
>