Add token owners to RPC and archive db #10989
Conversation
psteckler
added
the
ci-build-me
| @@ -51,6 +51,18 @@ module Public_key = struct | |||
| public_key | |||
| end | |||
|
|
|||
| module Token_owners = struct | |||
| (* hash table of token owners, updated for each block *) | |||
| let owner_tbl : Account_id.t Token_id.Table.t = Token_id.Table.create () | |||
There was a problem hiding this comment.
How would this work when the archive process restarts? Should we initialize the table with the tokens from the database?
There was a problem hiding this comment.
You're only adding tokens to the db when there's a transaction in a block that mentions them. In that case, those tokens will be in the tokens_used list, and hence, in this table.
There was a problem hiding this comment.
Actually, tokens section in zkapps integration test should be sufficient, right? since it exercises the replayer flow now
We have to add that, right?
| let open Deferred.Result.Let_syntax in | ||
| let value = Token_id.(to_string token_id) in | ||
| match owner with | ||
| match Token_owners.find_owner token_id with |
There was a problem hiding this comment.
Not finding an owner doesn't mean this is default token, right? (See the comment in Token_owners module definition)
There was a problem hiding this comment.
There should not be any tokens other than the default that have no owner, unless there's a bug, hence the asserts.
There was a problem hiding this comment.
It might be safer to have the table contain options, so that a lookup should never fail.
src/app/archive/archive_lib/diff.ml
Outdated
| let token_id = acct.token_id in | ||
| let owner = Mina_ledger.Ledger.token_owner ledger token_id in | ||
| (token_id, owner) ) | ||
| |> List.dedup_and_sort |
There was a problem hiding this comment.
Nit: it will usually be more efficient to dedup_and_sort first, then look up the token_owner
There was a problem hiding this comment.
The accounts_accessed are all already distinct, but the tokens they use are not.
Just realized what you were trying to say here. I will push a commit with a fix.
There was a problem hiding this comment.
Indeed, I meant
- get the tokens from the account IDs
- sort the tokens
- hit the ledger to find out the token owners.
Yes, I can do that, but I think it's for a separate PR. |
a follow up PR? (so that this code change is tested before we make the next qa-net release) |
Actually, tokens section in zkapps integration test should be sufficient, right? since it exercises the replayer flow now |
Add a field
tokens_usedto theBreadcrumb_addedRPC message, a list of token id, owner pairs. Add the same field to precomputed blocks and extensional blocks.The data in that list is used to populate a table of token owners in the archive processor, which is consulted when adding a token id to the archive db.
This change required additional code in
extract_blocksto find the tokens used in each block. Tested by extracting the blocks from an archive db created with a local testnet, and then runningarchive_blockson the resulting blocks.(Note: added an empty token list to the precomputed block sample, in the expectation that these will be regenerated.)
Closes #10772
Closes #10793