Note about hashing: sha256() and hash()

[Vladimir-Khm]

History of the problem

I was trying to create a small catalog in a map using a hash of an item as a key and value as a price. There was a struct item:

struct Item {
    prop1: String;
    prop2: String;
    prop3: String;
    prop4: String;
}

To calculate a hash of this item I used this code: sha256(msg.item.toCell().asSlice()) which seems to be fine, but suddenly I realized that smart contract does very strange things. After precise debugging, I found out the problem was that the code above returns absolutely the same hash despite all items being different. It is because sha256 is a math function that calculates a hash only of the first ref cell in my case. The ref cells of the struct were always the same, so the sha256() results were the same too. To calculate the hash of this struct you need to write this code: item.toCell().hash()

Proposal

From my point of view, it is an important note to prevent many hard-catching bugs and make smart contracts a little bit safer, so my suggestions are:

1. Also write hash function for structs to simply write: someStruct.hash()
2. Write down that if you want to calculate a hash of the object you need hash() and NOT sha256(). Emphases the difference between these two functions
3. Write down small examples to make the difference clear

[novusnota]

1. There's a big caution in the sha256() function reference, saying that in run-time it works only with up to 128 bytes (limitations of the underlying TVM instruction)

2. On the newly made assembly functions page there's an example of run-time (on-chain) SHA-256 function that works with arbitrary Strings (or Slices, if you s/String/Slice in two places there): https://docs.tact-lang.org/book/assembly-functions/#onchainsha256

3. The point about adding a .hash() method to Structs/Messages is interesting, however, the .toCell()-backed implementation may be a bit more expensive than we'd like to promote. For a cheaper approach, see point 2.

With those points in mind, I think this issue is resolved. If you think it's not, please reopen it and tell me why. I'll try to help :)