Support group keys for RFQ negotiation flows #1382
Conversation
GeorgeTsagk
force-pushed
the
rfq-negotiation-groupkey
branch
from
ded24ba to
a76775a
Compare
|
Rebased on |
There was a problem hiding this comment.
Great set of commits! Minimal, yet impactful diff.
I think I've spotted a bug where we'll start to compare sats balance instead of assets balance when trying to select the "best" channel. We should follow up there with a unit test to confirm the error, then correct it (if it does indeed exist).
In order to properly test this e2e, I think we may need to expose a group key arg during asset invoice creation. I don't think this requires the full on multi asset group channels (we just want to verify that we can make invoices based no the group key).
Reviewed 1 of 1 files at r1, 2 of 2 files at r2, 2 of 2 files at r3, 1 of 1 files at r4, 1 of 1 files at r5, 1 of 1 files at r6, 2 of 2 files at r7, all commit messages.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @GeorgeTsagk)
rfq/manager.go line 963 at r3 (raw file):
// Store the result for future calls. m.assetIDToGroup.Store(id, groupKeyBytes)
Not something we need to worry about today, but eventually we'll want have some sort of eviction policy here.
rpcserver.go line 7865 at r6 (raw file):
// the best local balance. fn.ForEach(balances, func(b channelWithSpecifier) { if b.channelInfo.LocalBalance >
Is .LocalBalance here actually the asset, or sats balance? I ask as it's lnclient.ChannelInfo, so I assume it would be a sats balance.
We should ensure that for the best balance here, we're looking at the asset balance
rpcserver.go line 7909 at r6 (raw file):
for assetIdx := range assets { assetOutput := assets[assetIdx]
Looks like the inner loop here could be an actual Filter, you'd need to contend with the extra return error value, but that's where the Result type can be nice when working with higher order functions/types.
itest/rfq_test.go line 429 at r7 (raw file):
// testRfqNegotiationGroupKey checks that two nodes can negotiate and register // quotes based on a specifier that only uses a group key. func testRfqNegotiationGroupKey(t *harnessTest) {
We should complement this test with one on litd that makes RFQ quotes with a group key.
Think aloud, in order to properly test that, I think we may need to support creating asset invoices with a group key at the very least.
rpcserver.go line 7886 at r6 (raw file):
type channelWithAsset struct { // assetInfo is the information about one of the assets in a channel. assetInfo rfqmsg.JsonAssetChanInfo
Interesting how this seems to duplicate the information in the attribute below it?
asset/asset.go
Outdated
| @@ -245,6 +245,11 @@ func (i ID) String() string { | |||
| return hex.EncodeToString(i[:]) | |||
| } | |||
|
|
|||
| // IsEqual returns true if the ID matches the provided ID. | |||
| func (i ID) IsEqual(a ID) bool { | |||
There was a problem hiding this comment.
just a drive-by comment: asset.ID is an array, so it supports the == operator.
There was a problem hiding this comment.
Ah, I see. I'd argue that we still should keep this method, it's a top-level way of comparing without having to dive into the type and see what it's made of
There was a problem hiding this comment.
Reviewable status: 7 of 8 files reviewed, 5 unresolved discussions (waiting on @guggero and @Roasbeef)
rpcserver.go line 7865 at r6 (raw file):
Previously, Roasbeef (Olaoluwa Osuntokun) wrote…
Is
.LocalBalancehere actually the asset, or sats balance? I ask as it'slnclient.ChannelInfo, so I assume it would be a sats balance.We should ensure that for the best balance here, we're looking at the asset balance
good catch, it's actually the sats balance
rpcserver.go line 7886 at r6 (raw file):
Previously, Roasbeef (Olaoluwa Osuntokun) wrote…
Interesting how this seems to duplicate the information in the attribute below it?
this should not be removed, it includes the decoded data of the CustomChannelData field of lndclient.ChannelInfo
rpcserver.go line 7909 at r6 (raw file):
Previously, Roasbeef (Olaoluwa Osuntokun) wrote…
Looks like the inner loop here could be an actual
Filter, you'd need to contend with the extra return error value, but that's where theResulttype can be nice when working with higher order functions/types.
you mean an fn.Filter? Seems like it would have to obfuscate some errors we want to send upstream, will see if I can pull it through
itest/rfq_test.go line 429 at r7 (raw file):
Previously, Roasbeef (Olaoluwa Osuntokun) wrote…
We should complement this test with one on
litdthat makes RFQ quotes with a group key.Think aloud, in order to properly test that, I think we may need to support creating asset invoices with a group key at the very least.
Trying to keep the scope of this PR small, focusing on the establishment of quotes on group keys, before involving send/receive flows anyhow.
Could implement your recommendation in a follow up PR, taking advantage of the imported LitD oracle harness too
rfq/manager.go line 963 at r3 (raw file):
Previously, Roasbeef (Olaoluwa Osuntokun) wrote…
Not something we need to worry about today, but eventually we'll want have some sort of eviction policy here.
True, will throw a todo, or just do something simple like flushing on intervals
asset/asset.go
Outdated
| @@ -245,6 +245,11 @@ func (i ID) String() string { | |||
| return hex.EncodeToString(i[:]) | |||
| } | |||
|
|
|||
| // IsEqual returns true if the ID matches the provided ID. | |||
| func (i ID) IsEqual(a ID) bool { | |||
There was a problem hiding this comment.
Ah, I see. I'd argue that we still should keep this method, it's a top-level way of comparing without having to dive into the type and see what it's made of
GeorgeTsagk
force-pushed
the
rfq-negotiation-groupkey
branch
from
a76775a to
269ad4b
Compare
Pull Request Test Coverage Report for Build 13768108410Details
💛 - Coveralls |
rpcserver.go
Outdated
| assetGen := assetOutput.AssetInfo.AssetGenesis | ||
| assetIDBytes, err := hex.DecodeString( | ||
| assetGen.AssetID, | ||
| ) | ||
| if err != nil { | ||
| return false, fmt.Errorf("error "+ | ||
| "decoding asset ID: %w", err) | ||
| } | ||
|
|
||
| var assetID asset.ID | ||
| copy(assetID[:], assetIDBytes) | ||
|
|
||
| match, err := r.cfg.RfqManager.AssetMatchesSpecifier( | ||
| ctx, specifier, assetID, | ||
| ) | ||
| if err != nil { | ||
| return false, err | ||
| } |
There was a problem hiding this comment.
This is the exact same code as in the previous commit. Perhaps we should extract that into a re-usable method in the Manager:
func (m *Manager) ChannelCompatible(ctx context.Context,
jsonAsset rfqmsg.JsonAssetChanInfo, specifier asset.Specifier) (bool,
error) {
gen := jsonAsset.AssetInfo.AssetGenesis
assetIDBytes, err := hex.DecodeString(
gen.AssetID,
)
if err != nil {
return false, fmt.Errorf("error decoding asset ID: %w", err)
}
var assetID asset.ID
copy(assetID[:], assetIDBytes)
match, err := m.AssetMatchesSpecifier(ctx, specifier, assetID)
if err != nil {
return false, err
}
// TODO(george): Instead of returning the first result,
// try to pick the best channel for what we're trying to
// do (receive/send). Binding a baseSCID means we're
// also binding the asset liquidity on that channel.
return match, nil
}
There was a problem hiding this comment.
Great recommendation, will add it. Although, I think the correct function signature should be
func (m *Manager) ChannelCompatible(ctx context.Context,
jsonAssets []rfqmsg.JsonAssetChanInfo, specifier asset.Specifier) (bool,
error)as we need to check against all assets, to see if the channel satisfies the specifier
We add a helper to the asset specifier which checks whether it is empty, returning an error if empty.
Previously our sell & buy request message validators would consider the message invalid if the included specifier didn't set an ID. Now we want to support group keys, so having either a group key or ID defined is fine.
GeorgeTsagk
force-pushed
the
rfq-negotiation-groupkey
branch
from
269ad4b to
7c660bd
Compare
There was a problem hiding this comment.
LGTM ⛵
Reviewed 1 of 1 files at r6, 8 of 8 files at r8, 2 of 2 files at r9, 2 of 2 files at r10, 1 of 1 files at r11, 1 of 1 files at r12, 1 of 1 files at r13, 2 of 2 files at r14, all commit messages.
Reviewable status: all files reviewed, 15 unresolved discussions (waiting on @GeorgeTsagk and @guggero)
rpcserver.go line 8017 at r13 (raw file):
// the best local balance. fn.ForEach(balances, func(b channelWithSpecifier) { if b.assetInfo.LocalBalance >
👍
This commit adds a new interface to the rfq manager which helps us look up the group an asset ID may belong to. The reason we can't reuse the same interface from tapsend is a circular import issue. We also introduce an in memory look-up map to skip a roundtrip to the db, since this information is static.
We add some asset specifier related helpers, for checking whether an asset matches a specifier, and also another helper that checks if a channels' assets satisfy a specifier. These methods will be used in a following commit.
When trying to add a local scid alias we would look up all channels and filter them based on whether the asset ID matches the one of the specifier. We now call the previously introduced helper method to easily check the asset against the specifier.
When querying for channels we now provide the specifier, allowing the helper methods to check against the specifier with all the helpers that we introduced in the previous commits. Now a channel with an asset ID that belongs to a group, may qualify as valid "balance" of that group.
This commit adds a simple itest which checks that Alice and Bob can negotiate an rfq quote, using a group key as the specifier.
GeorgeTsagk
force-pushed
the
rfq-negotiation-groupkey
branch
from
7c660bd to
6cf7190
Compare
|
Added TODOs, will work on them when related PRs are merged & after this is merged |
github-merge-queue
bot
removed this pull request from the merge queue due to failed status checks
github-merge-queue
bot
removed this pull request from the merge queue due to failed status checks
github-project-automation
bot
moved this from 👀 In review
to ✅ Done
in Taproot-Assets Project Board
Description
This PR introduces the minimum functionality for two nodes to negotiate and register a buy or sell quote, which identifies the underlying asset via a group key instead of an ID.
It includes a basic itest which lets Alice & Bob negotiate a quote based on a specifier which only includes a group key.
Related to #1028, but does not close it.
This change is