Canonical Value format doesn't work on script evaluation

[golddydev]

When mint using helios api, the minted value is sorted in canonical format. (length first then, alphabetical)

But when script evaluation it causes an error, because on smart contract when we handle value, that is sorted alphabetically.

[papag00se]

The nuance here is that for hardware wallets, the spec is slightly different. When creating the Tx for hardware wallets, "multiasset" types have to be in CBOR canonical order (shortest first) and this is what the tx hash should be based on. It seems cardano-node doesn't care either way, but for hardware wallet purposes, they should be built in canonical order per CIP-21

HOWEVER, when calculating the execution budget/fees, "multiassets" are always presented in lexicographical order (not based on length) in the ScriptContext.

It's a frustrating difference. Helios has to build the tx one way for CIP-21, but evaluate it a different way for execution budget.

cc: @christianschmitz

[papag00se]

I now wonder if this can account for at least some portion of the execution budget miscalculations

[christianschmitz]

This is an interesting issue.

@papag00se is indeed right that the current ordering is due to hardware wallet requirements.

Thanks for figuring out a possible solution @papag00se! I should be able to implement a fix soon.

Let's hope budget miscalculation is due to this

[christianschmitz]

It looks like the fix is trivial: use compareBytes with shortestFirst=false when sorting assets inside a new ScriptContext

https://helios-lang.io/docs/sdk/codec-utils/compareBytes

[christianschmitz]

I've applied the fix:

• ledger -> v0.8.2
• tx-utils -> v0.7.2

@papag00se If you want to patch your fork of ledger/tx-utils, here is the change to Assets.toUplcData() (in ledger):

    /**
     * Used when generating script contexts for running programs
     * @param {boolean} [isInScriptContext] if true: tokens are in strict lexicographical order
     * @returns {MapData}
     */
    toUplcData(isInScriptContext = false) {
        return makeMapData(
            this.assets.map(([mph, tokens]) => [
                mph.toUplcData(),
                makeMapData(
                    (isInScriptContext
                        ? tokens.sort((a, b) => compareBytes(a[0], b[0], false))
                        : tokens
                    ).map(([tokenName, qty]) => [
                        makeByteArrayData(tokenName),
                        makeIntData(qty)
                    ])
                )
            ])
        )
    }

And change to Value.toUplcData() (in ledger):

    /**
     * Used when building script context
     * @param {boolean} isInScriptContext
     * @returns {MapData}
     */
    toUplcData(isInScriptContext = false) {
        const map = this.assets.toUplcData(isInScriptContext)

        if (this.lovelace != 0n || isInScriptContext) {
            map.items.unshift([
                makeByteArrayData([]),
                makeMapData([
                    [makeByteArrayData([]), makeIntData(this.lovelace)]
                ])
            ])
        }

        return map
    }

[papag00se]

Thanks @christianschmitz. I'm glad it was an easy fix :)

[christianschmitz]

For future reference: I've added a comment to the SDK documentation: https://helios-lang.io/docs/sdk/ledger/Assets#touplcdata

[golddydev]

For future reference: I've added a comment to the SDK documentation: https://helios-lang.io/docs/sdk/ledger/Assets#touplcdata

What I use actually is txBuilder.mintPolicyTokensUnsafe function, which accepts [ByteArrayLike, IntLike][] type as tokens.

And that was ordered canonical (If I remember correctly), and now you fixed that is ordered in lexicographically?