This commit creates a new autopilot heuristic which simply returns
normalized betweenness centrality values for the current graph. This
new heuristic will make it possible to prefer nodes with large
centrality when we're trying to open channels. The heuristic is also
somewhat dumb as it doesn't try to figure out the best nodes, as that'd
require adding ghost edges to the graph recalculating the centrality as
many times as many nodes there are (minus the one we already have
channels with).
This commit removes an extra filter on address availability which is not
needed as the scored nodes are a already prefiltered subset of the whole
graph where address availability has already been checked.
This commit adds betweenness centrality to the available node metrics.
Betweenness centrality is a per node centrality measure which for an
arbitrary node v equals to the sum of shortest paths going trough v
divided by the number of all shortest paths for for each vertex pair
k, s where k != s != v.
This commit adds the NodeMetric interface which will be used for all
graph metrics not directly part of the autopilot but are useful in
composite heuristics to drive autopilot decisions and improve node
scores.
This PR fixes an issue that happens when adding a new channel edge
between two nodes in a memChannelGraph. Originally a channel edge held a
node value which made the graph different when iterating from the two
endpoints of an edge. This is simply fixed by holding pointers instead.
This commit adds a new signal to the autopilot agent, meant to signal
when any of the available heuristics has gotten an update.
We currently use this to trigger a new channel opening after the
external scores have been updated.
When appending to a slice, there is no guarantee the slice won't be
modified. So instead of appending to the global slice
availableHeuristics, we create a temporary local one.
Previously we waited only for the number of channels to become what we
expected, but this wasn't enough. Sometimes the agent had't yet updated
its internal balance, causing the test to fail with an unexpected
balance.
To make the autopilot able to account for fees, we let it use the
subtractFees option when opening channels.
This makes sure that each channel we attempt to open will eat at most
Amt out of our budget. Previously fees would eat into our funds in
addition, causing us to deplete our funds more than expected on each
channel opening.
This commit moves the logic querying the available heuristics out of the
autopilot agent and into the autopilot manager. This lets us query the
heuristic without the autopilot agent being active.
If called without the agent being active, the current set of channels
will be considered by the heuristics. If the agent is active also the
pending channels will be considered.
This commit fixes a regression in how we allocate funds to attempted
channels. We would earlier stay within the channel size limits, but we
wouldn't account for funds consumed by other channels being opened in
parallel.
We fix this by introducing a loop which greadily tries to distribute the
funds among the channels to open, and reduces the number of channels to
open in case not enough funds are available to satisfy the channel size
limits.
ScoreSettable is an interface that let caller set external scores for
the heuristic. The ExternalScoreAttachment and WeightedCombAttachment
heuristics implement this interface.
This commit adds a method SetNodesScores to the WeightedCombAttachment
heuristic.
Since the heuristic keeps a list of sub-heuristics, it will attempt to
recursively apply the scores to the sub heuristics.
This commit adds a new autopilot heuristic that is scoring based. It is
a simple heuristic that will keep a list of pubkeys and scores, and will
try opening channels with the nodes with the largest score first.
