Check price of data object bloat bonds

[L3pereira]

funds_needed_for_upload(params: UploadParameters<T>) created in pallet storage
I think It's more advantageous to use UploadParameters as a parameter than to separate the fields into params.

This function calculates the storage fee and the net prize an actor has to pay if he wants to update X amount of data objects.
It only works on updates, so the net deletion prize is always positive (NetDeletionPrize::<T>::Pos(b)).

net_prize = number of objects to update * DataObjectDeletionPrizeValue.
storage_fee = DataObjectPerMegabyteFee * round_up_to_nearest_integer(size of total objects in bytes / 1_048_576)

funds_needed_for_upload = net_prize + storage_fee

• Closes Check price of data object bloat bonds #3607

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Ignored Deployment

Name	Status	Preview	Updated
pioneer-testnet	⬜️ Ignored (Inspect)		May 4, 2022 at 11:14AM (UTC)

[bedeho]

This implementation is doing too much. You should not be checking and verifying lots of checks that are not relevant for fee estimation, those will be checked later as part of actual call to upload. Checking for existance of bag is OK, since we need contained informatino to do actual evaluation, but the other stuff is not.

[bedeho]

Won't net_prize include old deletin prize? If so, wouldn't funds_needed_for_upload in fact be the new total cost, rather than cost of doing this additional upload, which is what we care about?

[L3pereira]

When we call upload_data_objects this function will call try_mutating_storage_state with
BagOperationParams::Update(object_creation_list)
then

1. we will ensure bag exists and bag has a deletion prize, that is some for dynamic bags or none for static ones.
2. Construct objects to upload where deletion prize is the old one
3. we compute net prize init_value + (deletion_prize * num_obj_to_create)
4. We add net_prize to storage_fee

So net prize include deletion prize in the calculation, but that is what try_mutating_storage_state is doing to do the ensure_funds validation
I think the goal here is to separate funds validation from try_mutate_storage, right?

[L3pereira]

Now we have a much simpler funds calculation function. This function doesn't fail.
I didn't check for bag existence, let's move that responsibility somewhere else, since the main goal of this function is calculating funds.
The signature is very simple

    fn funds_needed_for_upload(
        num_of_objs_to_upload: usize,
        objs_total_size_in_bytes: u64,
    ) -> BalanceOf<T>;

I tested for overflow saturation, but I'm thinking now, these numbers are so huge, that no quantity of memory will ever amount to that.
This function is will roundup to the next greater integer when objs_total_size_in_bytes % 1048576 (1Mb) > 0

I also move to fixtures some helper functions, that were in the middle of the unit tests.

[bedeho]

This looks very simple and correct, originally, this is what I thought we would need, something very very simple. Why were you able to simplify so much only now?

[L3pereira]

I thought your goal was to separate the try_mutate in such a way we would have validation and then the mutation safe parts for the content pallet, and the funds function would have those validations, But for now, you just want a calculator.