Add Wallet::apply_changeset

[evanlinjin]

Description

My proposed alternative for #229.

To Do:

Expand on the "Warning" for the apply_changeset doc comments. We really don't want most users doing this until LocalChain changesets become monotone.

Changelog notice

• Add Wallet::apply_chaneset method.

Checklists

All Submissions:

• I've signed all my commits
• I followed the contribution guidelines
• I ran cargo fmt and cargo clippy before committing

New Features:

• I've added tests for the new feature
• I've added docs for the new feature

[coveralls]

Pull Request Test Coverage Report for Build 15037397098

Details

• 0 of 11 (0.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.1%) to 85.958%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
wallet/src/wallet/mod.rs	0	11	0.0%

Totals
Change from base Build 14921254296:	-0.1%
Covered Lines:	7266
Relevant Lines:	8453

---

💛 - Coveralls

[icota]

Does the OoO caveat apply to apply_update above too? If it does not I wonder if the increased serialized size (I'm getting a few kb) is a valid tradeoff

[evanlinjin]

That is a good point. I'm still a bit concerned about serializing Updates because of how CheckPoints can be used (as it is part of an Update).

CheckPoints are used to communicate between the wallet and the chain source.

• The wallet uses CheckPoints to tell the chain source it's initial blockchain state (before any updates).
• Then the chain source sends CheckPoints back to the wallet which is extended from what was provided from the wallet (as updates).

These CheckPoints can be sparse, or may include the whole chain. Additionally, we've been looking at making CheckPoints generic so that alternative data types such as BlockHeader (instead of just BlockHash) can be stored (which would allow us to calculate MTP properly).

Chain sources will be improved to incorporate more data into CheckPoints. For example, we've come to realize that the only correct way to use Electrum is basically download every block (except for some of those below Electrum checkpoints - different to BDK CheckPoints). Eventually, we'll have a streaming version of bdk_electrum that will include block headers in CheckPoints and have a lot more CheckPoints than what we have now.

To summarize, what I am saying is that I expect a lot more data being shoved in CheckPoints in the future (hope you are okay with this?). If you are okay with this point, I will be happy to support this feature.

However, because CheckPoint is a linked list (so it's a deeply nested data structure), we may risk blowing the thread-stack if we purely use derive-generated, purely recursive impls of Serialize and Deserialize. To move forward, I propose that we do some custom Serialize/Deserialize impls into a list. ChatGPT gave some suggestions (not sure if it compiles):

use serde::ser::{Serialize, Serializer, SerializeSeq};

impl Serialize for CheckPoint {
    fn serialize<S: Serializer>(&self, serializer: S) -> Result<S::Ok, S::Error> {
        // First, count how many nodes (or stash them somehow).
        let mut len = 0;
        let mut node = Some(self);
        while let Some(n) = node {
            len += 1;
            node = n.next.as_deref();
        }

        let mut seq = serializer.serialize_seq(Some(len))?;
        let mut node = Some(self);
        while let Some(n) = node {
            seq.serialize_element(&n.data)?;       // whatever you need to emit
            node = n.next.as_deref();
        }
        seq.end()
    }
}

[evanlinjin]

Another way to do this, assuming that the wallet on the device is a "dumb wallet", is to just send over the canonical txs/UTXOs where the device is basically a slave of the main wallet. Not sure if this is appropriate with what you guys are going for.

[evanlinjin]

Additionally, the caveat with ChangeSets will disappear once we modify/replace LocalChain with a GraphChain (which @ValuedMammal is working on).