AutoMapper in the DAL: When to use Mapper.Reset()?

I'm using AutoMapper in a g开发者_运维知识库enerated Data Access Layer. That works fine. It was a little confusing when using AutoMapper in another layer and realizing the mappings created in the DAL with Mapper.CreateMap<T1, T2>() were still present. I see Mapper.Reset() which will remove these however I'd rather not have to have the other layers worry about the DAL. Would the best practice be to put a Mapper.Reset() before and after my mapping operations in the DAL? Or is there a way to give these DAL mappings a non-default key to let them persist but not interfere with the use of AutoMapper in other layers?

Note: The use of AutoMapper in the DAL has some specific options such as a number of .ForMember(...) calls that my other layers should not use (without a Mapper.Reset() they would reuse these options).

AutoMapper works as a singleton/single instance. Does it really matter though?

EDIT : This may help you Using Profiles in Automapper to map the same types with different logic

If your other layers aren't worried so much about the DAL classes chances are they aren't going to be calling Map on an instance of the DAL class anyway.

If you call Reset() then your DAL classes will need to restate them when they next need to do some mapping which will add extra very unnecessary overhead.

EDIT : If you call Reset at the start of every DAL call then you can only have a single threaded Data Access strategy. If you call Reset in the middle of a mapping for another DAL project then you are going to obviously break this - so you will have to lock on every DAL method.

This is not the way to use Automapper so I would be leaning towards either looking into those profiles, or not using it all together.

ALSO : Can you post a sample code for what is wrong with having lots and lots of multiple mappings going on? Are there different mapping strategies between two types depending on where in the DAL they are being called from?