lang: remove the state and interface attributes

[Aursen]

TODO:

• Fix the client
• Fix all examples
• Check unnecessary code snippets
• Remove interface attribute
• Clean the doc

[vercel]

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

A member of the Team first needs to authorize it.

[Aursen]

@Henry-E for the basic 4 example, do I use init-if-needed with a pda?

[Henry-E]

personally i would prefer to not include init_if_needed in any of the basic examples

[Aursen]

• Examples:
  • deleted the interface
  • deleted misc2
  • disabled lockup
  • updated misc to work stateless
  • updated basic4 to work stateless
  • updated version of tic-tac-toe anchor
• Client:
  • removed state and RequestNamespace
  • updated instructions function to remove unused account metas
• Lang:
  • removed cpi_state and state accounts
  • removed state count generation (i.e.: codegen/program/accounts.rs)
  • removed state and ctor sighash and ctor args generation (ie: codegen/program/common.rs)
  • suppression of the generation of state_cpi_methods (ie : codegen/program/cpi.rs)
  • removed the generation of ctor, state, interface arms (ie : codegen/program/dispatch.rs)
  • removed ctor, state and interface handlers (ie : codegen/program/handlers.rs)
  • removed ctor and state variants in instructions (ie : codegen/program/instruction.rs)
  • removed the state in the idl
  • removed state parser
• TS:
  • delete state.ts
  • update borsh (remove encodeState, state method map and update parseIxLayout)

[Henry-E]

minor english maybe use it calculate ...

The key with making a trait that automatically calculates size is that it doesn't have a high runtime cost. It really ought to be closer to a constant, although need to account for variable sized elements.

[Henry-E]

Looks really nice, thanks for such a great effort!

Some loose thoughts,

• is it possible to keep the interface trait without state? Maybe we can just remove the interface test example which uses state but leave the interface attribute in the code just for people that really want it and mark it as deprecated.
• references to the account types ProgramState and CpiState and also the constraint state should probably also be removed from the syn/.
  • If you want, I can do this part but I would need you to first rebase your PR to the current master so that I can safely pull and update it (the rebasing part is actually the most time consuming/annoying, probably more annoying than finding and deleting the account + constraint references)

Another question as well to check before we merge for my own sanity

• In the non-example parts of the anchor code, e.g. lang/ was state relatively self contained or where there any places where you had to modify the workings of any traits/functions/impls in order to get it to compile?

[Aursen]

Concerning the state and the interface the two are extremely linked because the interface does not call the global function in the dispatcher. And the interface in the idl was in the state category, so I concluded that if the program had no state, the interface was not usable. It is better for this kind of practice to make a client from the idl as anchor-gen or mannually.

Program state is used by the idl internally, from what I remember.

For the account size, I may delete it because I'm going to do a derive macro for it.

Then for all that is constraint it is true that I did not touch it, because of memories there is no constraint related to the state.

No everything is relatively contained concerning the state, I just had to modify a TS function for the instructions. And the same in the client.

And for the rest I will do it

[Henry-E]

Cool, I will play around with this locally tomorrow morning and try to merge.

[Henry-E]

Thanks for all your hard work on this!