to_account_metas()macro with Vec::with_capacity(#length)

[Eliascm17]

Problem

• Up until this point, the to_account_metas() macro would use .extend() to append account metas to an already allocated vec which would make for unnecessary memory allocations which would also waste compute units.

Solution

• Introduce ToAccountMeta trait that allowed us to:
  • introduce to_account_meta() method that returns a single account meta for the same number of ToAccountMetas implentations
  • update to_account_metas() method implementations that return a vec of to_account_meta()
• update to_account_metas() macro to:
  • take into account the number of accounts in Context<T> and use that as an arg to pass into Vec::with_capacity(#length)
  • use .push() instead of .extend() when adding account metas to the vec which rids of the unnecessary memory allocations

Previously:

Now:

cc: shoutout @cavemanloverboy & @nickgarfield for doing all of the heavy lifting. Just trying to get this PR submission across the finish line.

[vercel]

@Eliascm17 is attempting to deploy a commit to the coral-xyz Team on Vercel.

A member of the Team first needs to authorize it.

[cavemanloverboy]

The changes here are rather impactful. The snippet shown above does upwards of 16 memory allocations to construct a single Vec<AccountMeta>.

[Aursen]

Please can you remove cpi_account, cpi_state, loader and state from the accounts module?

[Henry-E]

Going to run through this now but once we merge #2386 , the macros will have to be added to the new account types as well.

@Eliascm17 For some reason the InitSpace macro isn't being recognised and this is causing cargo build to fail. Are you sure you rebased to the latest master?

Overall I like the idea the of switching to using a macro because extending vectors like this is super wasteful but I will have to read through this in more detail to understand whether it correctly covers the case of composite accounts.

[Henry-E]

yeh, this basically breaks how composite accounts work. Since you're calling to_account_metas on the another account context type which has itself used the derive(Accounts) macro.

[cavemanloverboy]

There may be a way to handle this by checking if the type is a composite account.

Alternatively, would be nice if the derive(Accounts) macro also implemented Iterator and iterated through the accounts in order. This could allow you to call to_account_meta on each element of a composite account.