Regression caused by NH-3904

[gliljas]

Just a suggestion. Please advise.

[hazzik]

Some options here:

1. fast & easy solution: rollback this heuristic to get custom type and force user to use .MappedAs() extension method (NH-2401/NH-2401 - A method to specify IType of LINQ parameter #421, also in 4.1) .
2. properly implement GuessType (see AbstractQueryImpl.GuessType to get the idea). But this solution will still have some flaws as sometimes the wrong type will be picked anyway.
3. same as above, but also analyse surrounding expression to get the idea about constant expression true type (similar to NH-3403: Wrong parameter size in query with MsSql2000Dialect,MsSql2005Dialect and MsSql2008Dialect #480)

My preferred option is 1.
/cc @oskarb?

[gliljas]

Option 1 makes sense, since then there will only be an explicit way to use user type constants, and not also a brittle, implicit way.

Option 2 was what I wanted to do, but for an urgent fix, it seemed lika bad idea.

Option 3 is the best way, IMO, but also potentially tricky.

[hazzik]

Found that MappedAs actually does not work...