Optim ledger_db get_datastore_keys

[sydhds]

get_datastore_keys() should return None if addr does not exist but for now it will return an empty BTreeSet.

So the get_keys in speculative_ledger.rs should do first an entry_exists() and then a get_datastore_keys() and it looks suboptimal because the balance (tested for entry_exists) and the datastore keys are consecutive in the ledger, but we fully independently look them up without hints (logarithmic search).

[Ben-PH]

What I'm thinking at this point, is to provide an interface that allows you to do a batch-get of the data-base entries using this method.

I'm thinking something like this (except not so jankey. Perhaps with builder pattern, and perhaps encapsulate the types somehow)

    pub fn batch_get_sub_entries(&self, address: &Address, balance: bool, bytecode: bool, datastore: Option<Vec[u8]>) -> Option<(Option<Vec[u8]>, Option<Vec[u8]>, Option<Vec[u8]>)> 

This could allow for a refactor of entry_exists() to

• when need to know if anything exists, just get something, and follow up with a is_some() further up the stack
• when used in the context of what this issue is trying to address, can get what's needed, while leaning into the rocks-db optimisations available.

unless there is a way to make assumptions about storage, thus allowing us to get all three in a single get?

[Ben-PH]

Continuing this discussion in #3437