Add back in helper function semantic::Policy::_translate_pk

[tcharding]

Recently while removing recursion I mistakenly removed the _translate_pk function forgetting that this was a trick to prevent compilation explosion caused by the generics.

[tcharding]

Needs #613

[apoelstra]

I don't understand this PR.

[tcharding]

There was a post sometime recently by @sanket1729 about how we were getting an explosion of code once all the generics were expanded [0]. And something about the helper function not needing to be re-compiled each time ...

Can you help me out Sanket, I clearly did not fully grok it when I read it.

[0] Is "expanded" the correct term?

[apoelstra]

There are a couple things that I think you might mean:

1. When we take generic parameters that can take a T or &mut T and have functions that call each other with a call tree a few layers deep, we wind up compiling functions for T, &mut T, &mut mut T, etc., which is wasteful. This happens with the rust-bitcoin encoding stuff but never here.
2. When we call functions from themselves we get the same thing but infinitely many times and compilation fails. No longer an issue here.
3. When a function takes a generic AsRef (for example) and all we do with the generic argument is call .as_ref(), then everything after that line uses concrete types, and it's wasteful to compile the whole function multiple times with different generic parameters. Better to have a helper function with concrete types, then the real function calls as_ref() before calling that.

I think you are trying to address issue (3). But the current code here is a no-op. You have two functions with identical signatures, with identical generics, and one is calling the other.

[tcharding]

Thanks for the explanation, should have trusted myself more originally when I removed the helper because I couldn't see it was achieving anything. Closing.

[tcharding]

I dug through the git index and the _translate_pk function was introduced when translate_pk had a different signature, when we changed it we didn't notice that the helper was no longer needed. The "mysterious" reason I vaguely thought existed must be some unrelated thing floating around in my brain.

For interest it was introduced in commit ef1d69c81e3acc34ed2ce71b6484d719cc631ab6