[ADP-3406] Scaffold signTxBody for deposit wallet

[Anviking]

• Scaffold signTxBody function
• Add rootXSignKey field to WalletState

Issue Number

ADP-3406

[Anviking]

Wasn't sure where to put a type alias, so went with XPrv for now

[Anviking]

Or well — at least Maybe Write.Tx as rootSignKey might be nothing.

But is returning IO (Maybe result) in the .IO module our desired approach then like

cardano-wallet/lib/customer-deposit-wallet/src/Cardano/Wallet/Deposit/IO.hs

Lines 205 to 208 in 7d766ff

	createPayment
	:: [(Address, Read.Value)] -> WalletInstance -> IO (Maybe Write.TxBody)
	createPayment a w =
	Wallet.createPayment a <$> readWalletState w

? ... which contrasts with how we've in cardano-wallet moved towards IO result with a potential WalletException thrown.

[HeinrichApfelmus]

🤔 For the pure version, I would definitely add explicit failure.

I would also leave out the [BIP32Path] argument, because that can be computed from the combination of TxBody and WalletState. (My feeling is that this argument would be on topic for a function whose abstraction is at the level of XPrv, e.g. signTxBody :: XPrv → [BIP32Path] → …, and does not know about the WalletState.)

How about this:

-- | Add signatures for *all* inputs if possible, `Nothing` otherwise.
signTxBody :: Write.TxBody -> WalletState -> Maybe Write.Tx

For the version without Maybe, I would probably give it a different name, which makes the semantics more clear:

-- | Add signatures for as many inputs as possible.
cosignTx :: Write.Tx -> WalletState -> Write.Tx

We can have either one or both, but for the beginning, I think that signTxBody is a good start.

[HeinrichApfelmus]

Thanks! 😊

🤔 I think I prefer the version with explicit Maybe, definitely in the pure part.

[HeinrichApfelmus]

🤔 For the pure version, I would definitely add explicit failure.

I would also leave out the [BIP32Path] argument, because that can be computed from the combination of TxBody and WalletState. (My feeling is that this argument would be on topic for a function whose abstraction is at the level of XPrv, e.g. signTxBody :: XPrv → [BIP32Path] → …, and does not know about the WalletState.)

How about this:

-- | Add signatures for *all* inputs if possible, `Nothing` otherwise.
signTxBody :: Write.TxBody -> WalletState -> Maybe Write.Tx

For the version without Maybe, I would probably give it a different name, which makes the semantics more clear:

-- | Add signatures for as many inputs as possible.
cosignTx :: Write.Tx -> WalletState -> Write.Tx

We can have either one or both, but for the beginning, I think that signTxBody is a good start.

[HeinrichApfelmus]

Suggested change

	signTxBody :: Write.TxBody -> [BIP32Path] -> WalletState -> Write.Tx
	signTxBody :: Write.TxBody -> WalletState -> Maybe Write.Tx

[HeinrichApfelmus]

Suggested change

	signTxBody :: Write.TxBody -> [BIP32Path] -> WalletInstance -> IO Write.Tx
	signTxBody :: Write.TxBody -> WalletInstance -> IO (Maybe Write.Tx)

But is returning IO (Maybe result) in the .IO module our desired approach then like

I think so 🤔 — The ubiquitous ErrNoSuchWallet is gone, and by making the failure explicitly in the result, we can also present it an explicit HTTP status code 4xx response. I would only sweep the status code 500 "internal server error" into the IO exceptions.