Lnd: SubscribeInvoices, can't get settle_index 1

[ottosuess]

Background

Documentation for the SubscribeInvoices call says about the settle_index:

If specified (non-zero), then we’ll first start by sending out notifications for all settled indexes with an settle_index greater than this value.

That means if I set the settle_index to 0 i only get notified about new Invoices.
If I set the settle_index to 1 the first Invoice I get notified about is the one with an settle_index of 2.

So I'm always missing the first first invoice.

Would be cool if I could just rely on SubscribeInvoices to get notified about all my invoices. But right now I have to use the ListInvoices call to page through all my invoices until the first invoice is settled to make this work.

same for the add_index.

Possible Solutions

• make settle_index start at 2
• send out notifications for all indexes with an index greater or equal than the provided value
• let users of the api set settle_index to 0 and make the default value something else (null or -1)

Your environment

• lnd version 0.5.1-beta commit=v0.5.1-beta-154-g71444e74acc9896d31205bd16b70fd0f85fc691b
• Linux btc 4.15.0-42-generic #45-Ubuntu SMP Thu Nov 15 19:32:57 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux
• Bitcoin Core Daemon version v0.17.0.0-ge1ed37edaedc85b8c3468bd9a726046344036243

[cfromknecht]

@ottosuess have you been able to reproduce this on your node? I think it may just be a documentation issue, as from my understanding, we send out all indexes greater than or equal to the specified index

[ottosuess]

@cfromknecht yes, when I set settleIndex = 1, the first invoice I get is the one with settleIndex == 2.

[halseth]

Could we add new parameters to the API that uses greater or equal, then deprecate the old ones?

[riperk]

What about treating 0 differently from undefined? i.e. If you only want new settlements, you don't set a settle_index. If you set a settle_index to value (including 0) then you get all above that value

[halseth]

@riperk I think the default value is 0, so you cannot differentiate that from not set.

[instagibbs]

Any update on this wrinkle? New hold invoice system avoids this issue but curious to see what the cleanest fix is.

[Roasbeef]

We're planning a new sub-server RPC call to fix all the issues with the existing base service invoice subscription endpoint. Until then, the hacky way would be to use LIstInvoices to just grab that first invoice (if you really need it), then you subscription for the rest. I'd imagine an application is storing the two indexes (settle and add) on their end so they can resume consuming the stream after a restart.

[instagibbs]

Yes the hack ended up not being so bad.

On Wed, Jun 3, 2020, 8:45 PM Olaoluwa Osuntokun [e-mail]
wrote:

We're planning a new sub-server RPC call to fix all the issues with the
existing base service invoice subscription endpoint. Until then, the hacky
way would be to use LIstInvoices to just grab that first invoice (if you
really need it), then you subscription for the rest. I'd imagine an
application is storing the two indexes (settle and add) on their end so
they can resume consuming the stream after a restart.

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
https://github.com/lightningnetwork/lnd/issues/2469#issuecomment-638533746,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABMAFU4V2WQBBBT6FG3B3S3RU3VBZANCNFSM4GPZ4UAQ
.

[tim2CF]

@riperk I think the default value is 0, so you cannot differentiate that from not set.

Unfortunately protobuf standard don't support generics, so we can't just define Maybe a generic type in native terms of protobuf. This will not compile:

message Nothing {

}

message Maybe a {
  oneof unMaybe {
    a just = 1;
    Nothing nothing = 2;
  }
}

So absence representation is not possible in general, but it's still might be done this way:

https://github.com/protocolbuffers/protobuf/blob/d61aede89cf188367766b971f59cf57d7835d8e8/src/google/protobuf/wrappers.proto#L31-L39

The same practice can be used to represent absence of value of any custom enum type as well. These small messages-wrappers are a bit similar to newtype expressions in Haskell or Rust.