Consider making it easier to add new TypeMapper instances based on old types

[felher]

Hey all,

in my code bases I use a lot of new types, often just wrapping a more fundamental type like string or int. I assume that's rather common. If I want to use such a type, I think I currently have two ways to do so:

1. Create a TypeMapper from scratch, which requires at least a bit of understanding of JDBC/SQL. For example, for a String-based type you have to decide between CHAR, VARCHAR and LONGVARCHAR. Also, you have to know what a cast[T] is in SQL.
2. Piggyback on some TypeMapper already there, delegating to it and maybe casting the type somewhere.

Both of them involve a bit of boilerplate and don't really signal to the user "you're doing it right".

I suggest adding a simple method to TypeMapper or its companion object. Maybe something like:

object TypeMapper:
  def basedOn[T, U](tm: TypeMapper[T])(toBase: U => T, fromBase: T => U): TypeMapper

Apart from being less boilerplate, the user needs to know basically nothing about SQL/JDBC to know that what they have done is valid. Also, it makes it easier to not care about the dialect you are currently using, because a implicit for the base type should be in scope whenever DialectTypeMappers are in scope, so it should naturally use the correct one.

Any thoughts? Worth it, not worth it?

Felix

[lihaoyi]

Sounds reasonable. uPickle has this and calls it bimap

[lihaoyi]

Here's a PR implementing it #27

[lihaoyi]

Should be released in Scalasql 0.1.8

[felher]

Great! Thank you 🙏

[aboisvert]

Another thanks from the peanut gallery! I will most likely use it as well.