Be able to pay to an address

[yannham]

Is your feature request related to a problem? Please describe.
Currently, we can pay to a bunch of different recipient such as a pub key hash or a script. Technically, because IsTxSkelOutAllowedOwner is implemented for those types. However, it is not possible to pay to a generic Address.

Describe the solution you'd like
I would like to be able to pay to a generic Address.

Describe alternatives you've considered
We can deconstruct the address manually and pay to it in some cases, but it's not clear what to do if the address resolves e.g. to a script credential. More generally, even if it would only be boilerplate, it's still common to want to pay to an address, so a helper in cooked would be nice.

Additional context
The difficulty described in the previous paragraph will be there for the cooked implementation as well: it's not clear yet how to proceed for script credentials.

[mmontin]

In the current state of cooked, it is not possible to pay to an address, because when paying to a script, we require the full script to exist and be given to cooked-validators. Thus, if the address resolves to a script credential, and cooked is not already aware of the script that corresponds to this credential, the payment cannot be completed.

In other words, Address does not implement IsTxOutAllowedOwner and thus cannot be converted to a User.

Additionally, addresses can have staking credentials attached, which is not part of the owner.

A work around would be to have a function:

fromAddress :: Address -> m (Maybe (User IsEither Allocation), Maybe StakingCredential)

This function would:

• Get the staking credential part of the address and return it
• If the credential is a pub key credential, return the associated user
• If the credential is a script credential, look for a script with the right hash on the blockchain and return it if any, otherwise fail

However, fundamentally, in cooked, we cannot pay to a script for which we don't have access to the code (because we would not be able to execute it later) which is a design choice, and thus it's questionable whether such a helper should exist.