ChainGraph API for electrum

[LLFourn]

ChainGraph API for electrum

[rajarshimaitra]

I am working on this next..

[LLFourn]

Side note this means removing insert_txid from chain graph since this is incompatible with this design.

[evanlinjin]

We want to have the chaingraph enforce that for every txid in sparse_chain there is a full tx in the graph. To do this we'll have to make those fields private and we'll have to have an error case when the changeset is applied.

For the rules of applying the changeset, I think as long as the update chain's txid is included in either graphs (whether it be in original or the update) we can allow the update.

This allows the following logic:

1. Grab txids of given scriptPubKeys.
2. Formulate initial update (nothing in txgraph).
3. Find missing transactions (by looking at original graph).
4. Fetch missing transactions and add to update's graph.
5. Apply update chaingraph (which does not contain full txs of every txid mentioned in update.chain, but collectively with the original chaingraph, all txs will exist).

[LLFourn]

@evanlinjin I think we should provide this guarantee as the API can result in unintentional unwraps when you think a full tx should be there but it isn't. For electrum I don't think the update process should start with a ChainGraph at all:

1. grab txids of given script pubkeys.
2. Figure out which of these we need full transactions for i.e. return something like a HashSet<Txid> or maybe a Vec<(Txid, I)> where I is TxHeight in this case.
3. Fetch those full transactions and transform them into ChainGraph<TxHeight> with only the transactions in missing set.
4. Apply the ChainGraph.

[evanlinjin]

@LLFourn what about transactions being confirmed? Because they might already exist in the mempool of the original chain

[LLFourn]

@evanlinjin the update ChainGraph<TxHeight> will have the updated height of the transactions. I don't see a problem. The TxHeight of the transaction in the original chain is ignored as part of this. We only care whether it is present in the original ChainGraph or not.

[LLFourn]

The best way to go about this might be to do electrum with the current API first. It's not too hard to figure out which transactions are missing and you need to fetch with the current API. Once we can see what that looks like we can try and optimize both the electurm API and the chaingraph API to make it simple to document and explain,

[rajarshimaitra]

For ref, I tried the electrum sync with existing API in #77 ..