Replace _txInputs of Tx with nonempty-containers Set in Shelley about cardano-ledger-byron HOT 5 CLOSED

We should leave this for now I think. The non-empty library needs a bit of work before I would want to consider using it. The main issues are:

• No Foldable instances for Data.NonEmpty.Set
• Unclear type naming: all the types are named T for some reason. I believe this is so we could make our own aliases for anything isomorphic, but still slightly odd.
• Missing Aeson instances

We could contribute the missing instances and discuss naming with the library author, but that's a longer-term solution.

from cardano-ledger-byron.

Can't we just define something like data NESet e = NESet e (Set e) and add instances for it?

from cardano-ledger-byron.

Yeah, we could do something like that, but I was hoping to have a library-based solution. We should still think about it though - maybe I'll reopen this.

from cardano-ledger-byron.

After some investigation it turns out that we cannot use a set in the OBFT era and will have to wait for Shelley to make that change. The reason is to do with transaction witnesses and matching them to specific transaction inputs. In the old implementation a transaction contained a list of transactions inputs. Another datatype TxAux holds the transaction and a list of witnesses, one for each input. The only way to match these is the position in the list. If we turn the list of inputs into a set, we don't guarantee the same ordering, so finding the right witness becomes harder. We would have to try to verify against several witnesses until we found a matching one, which is wasteful.

Instead, we can wait until the Shelley implementaion to use a non-empty set.

from cardano-ledger-byron.

In Shelley we'll be working from the executable spec, so we'll deal with this when translating from that

from cardano-ledger-byron.