Feerate should be returned from Transaction construction

[LLFourn]

When you get TransactionDetails from tx_builder.finish() it has the fee for the overall transaction but does not include the feerate of the resulting transaction. There is no simple way for the caller to replicate this information since the transaction does not have witness data at this point. This is annoying because after constructing the psbt the user should know the feerate before signing it.

This can either be done by returning another value from builder.finish() or by adding a (perhaps optional) field to TransactionDetails.

[csralvall]

Would make sense to save the overall weight or size of the transaction and the fee rate instead of the fee amount? I think that the weight would be useful in the TransactionDetails, and we can implement a method to compute the fee amount. WDYT? I forget to mention that I was talking about a modification to CoinSelectionResult that would allow to pass the fee_rate to Transaction Details inside wallet.create_tx. However, now I think that it is an overkiller, and we can add the field just inside TransactionDetails without touching the CoinSelectionResult struct.

[thunderbiscuit]

Just want to comment to say that this is also a feature I'd use if implemented.

[i5hi]

I am giving this issue a try and have a few thoughts:

Transaction size can be useful since you can calculate the fee_rate with it.

When using create_tx does it make sense to have Some(Transaction) always returned so its easy to calculate size?

Or we can just have size as a field in TransactionDetail?

fee_rate helps too but feels repetitive since there is already a fee field.

Edit: (weight + max_satisfaction_weight)* instead of size*

[LLFourn]

When using create_tx does it make sense to have Some(Transaction) always returned so its easy to calculate size?

Even if you did that tx doesn't have the correct weight since it is missing signatures.

[i5hi]

When using create_tx does it make sense to have Some(Transaction) always returned so its easy to calculate size?

Even if you did that tx doesn't have the correct weight since it is missing signatures.

The max satisfaction weight from the descriptor would give you that, right?

[LLFourn]

The max satisfaction weight from the descriptor would give you that, right?

If all the inputs are yours and you know which input had which descriptor. But in general no.

[thunderbiscuit]

I think we're bumping into a related issue when we attempted to RBF bump a transaction on bdk-ffi/Kotlin.

BIP 125 says that nodes will relay transactions for which the fees are bumped by at least the amount of the minimum transaction relay rate, i.e. if the minimum is 1 sat/vbyte your "currently paying" 3.0 sat/vbyte transaction will need to have a replacement transaction with at least 4.0 sat/vbyte to be related by nodes.

But since we don't know the feerate, calculating how much fees should be added to the RBF replacement transaction is not straightforward (I assume it's doable by looking at nominal fee paid and raw transaction size but having it returned explicitly as part of the transaction would make it much easier).