Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

allow breaking inside for_blocks closures using ControlFlow #925

Open
wants to merge 1 commit into
base: master
Choose a base branch
from

Conversation

antonilol
Copy link
Contributor

@antonilol antonilol commented Aug 19, 2023

bitcoind will keep sending blocks even if ControlFlow::Break(_) if returned, so they have to be received to not interfere with future calls. blocks technically do not have to be parsed then, which happens now at the sender side of blocks_recv, but this makes more sense to optimize that together with using bitcoin_slices (#913 moves parsing of the block to the receiver side).

is it possible to send bitcoind some message to stop sending blocks? otherwise it would be possible to request a few at a time and stop when ControlFlow::Break(_) is returned to prevent having to receive (and bitcoind having to read from disk and send) many blocks.

i also rewrote lookup_transaction to use this method of breaking, which will save at least parsing unneeded blocks later (#913). the only change to this function right now is for_blocks will keep the state with the ControlFlow enum instead of lookup_transaction with Option

i also happened to need this when making a poc implementation for is_unused (#920)

@romanz romanz self-requested a review August 19, 2023 14:13
@romanz
Copy link
Owner

romanz commented Aug 19, 2023

Thanks for the contribution!
Will review after #913 is merged.

@antonilol antonilol force-pushed the for_blocks-control-flow branch from ba04a6e to 4775424 Compare August 22, 2023 20:02
@romanz
Copy link
Owner

romanz commented Aug 27, 2023

IIUC, this optimization is used for optimizing txid index scan (by early-exiting the iteration).
Since currently txid index shouldn't have collisions (except BIP-30 transactions), I'd prefer to keep the existing code for simplicity.
I have opened #928 to make sure there are no more collisions.

@antonilol
Copy link
Contributor Author

IIUC, this optimization is used for optimizing txid index scan (by early-exiting the iteration)

yes, exiting early means that after the right one if found no more block parsing and hashing has to be done. actually, i just found out that #913 changed this function to return the last matching transaction, whereas before it returned the first one. i think it is best to get the last one as according to bip 30 it is allowed to have duplicate txids, just not duplicate utxos, and i think that it is more likely that a transaction is found in more recent blocks, so reversing blockhashes before it is passed to for_blocks will fix this and optimize this even more, and make the algorithm consistent with what it is now (current master, after #913)

Since currently txid index shouldn't have collisions (except BIP-30 transactions)

the may be collisions in txids as long as no utxos from the utxo set would be overwritten by it.

I'd prefer to keep the existing code for simplicity.

i dont see a complexity increase in this code, and find this well worth the optimization

@antonilol antonilol force-pushed the for_blocks-control-flow branch from 4775424 to f46cffc Compare August 28, 2023 12:37
@antonilol
Copy link
Contributor Author

reversing blockhashes is somewhat harder that sticking a .rev() adapter in between, but with rocksdb::IteratorMode a backwards iterator can be made, out of scope for this pr, but surely an interesting optimization

(btw that force push was just a rebase)

@romanz
Copy link
Owner

romanz commented Aug 29, 2023

the may be collisions in txids as long as no utxos from the utxo set would be overwritten by it.

Not sure - do you mean that two different transactions having the same txid?
I think this is extremely unlikely (assuming that SHA-256 is collision resistant).

@antonilol
Copy link
Contributor Author

the may be collisions in txids as long as no utxos from the utxo set would be overwritten by it.

Not sure - do you mean that two different transactions having the same txid? I think this is extremely unlikely (assuming that SHA-256 is collision resistant).

i meant that two transaction having the same txid is allowed in bitcoin, as stated in bip 30. if this would happen i believe it would be more likely because of a bug than a collision of sha 256

if a user requests a transaction by txid, which transaction should be returned? before #913 this was the first one, now every transaction it finds is assigned to a variable, overwriting a previous result, resulting in the last one being returned

@romanz
Copy link
Owner

romanz commented Aug 31, 2023

I think that #933 should resolve the ordering (returning the first matching tx), as well as not parsing the next blocks.

@antonilol
Copy link
Contributor Author

I think that #933 should resolve the ordering (returning the first matching tx), as well as not parsing the next blocks.

i think as well (it reuses the old logic). actually, such a thing should probably be clarified in the electrum server protocol spec.

that fixes one point of this pr (i actually found out about it later than making this). i would like to have the possible optimization i mentioned in the first message, to save bitcoin core from having to send all blocks after the tx has already been found, using this method of 'telling' for_blocks the at first requested blocks are no longer needed. i will try to do some testing on how much time/resources this saves and implement it in the coming week

@antonilol antonilol force-pushed the for_blocks-control-flow branch from f46cffc to dcae47f Compare September 2, 2023 20:57
@antonilol antonilol force-pushed the for_blocks-control-flow branch from dcae47f to cbb0ed1 Compare May 25, 2024 11:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants