Mavryk_raw_protocol_001_PtAtLas.Cache_reprFrequently used data should be kept in memory and persisted along a chain of blocks. The caching mechanism allows the economic protocol to declare such data and to rely on a Least Recently Used strategy to keep the cache size under a fixed limit.
Take a look at Environment_cache and Environment_context for additional implementation details about the protocol cache.
The protocol has two main kinds of interaction with the cache:
1. It is responsible for setting up the cache with appropriate parameter values and callbacks. It must also compute cache nonces to give the shell enough information to properly synchronize the in-memory cache with the block contexts and protocol upgrades. A typical place where this happens is Apply. This aspect must be implemented using Cache.Admin.
2. It can exploit the cache to retrieve, to insert, and to update cached values from the in-memory cache. The basic idea is to avoid recomputing values from scratch at each block when they are frequently used. Script_cache is an example of such usage. This aspect must be implemented using Cache.Interface.
module Admin : sig ... endThe following module acts on the whole cache, not on a specific sub-cache, unlike Interface. It is used to administrate the protocol cache, e.g., to maintain the cache in a consistent state with respect to the chain. This module is typically used by low-level layers of the protocol and by the shell.
A client uses a unique namespace (represented as a string without '@') to avoid collision with the keys of other clients.
val create_namespace : string -> namespacecreate_namespace str creates a valid namespace from str
module type CLIENT = sig ... endTo use the cache, a client must implement the CLIENT interface.
module type INTERFACE = sig ... endAn INTERFACE to the subcache where keys live in a given namespace.
val register_exn :
(module CLIENT with type cached_value = 'a) ->
(module INTERFACE
with type cached_value = 'a)register_exn client produces an Interface specific to a given client. This function can fail if client does not respect the invariant declared in the documentation of CLIENT.
val cache_nonce_from_block_header :
Block_header_repr.shell_header ->
Block_header_repr.contents ->
cache_noncecache_nonce_from_block_header shell_header contents computes a cache_nonce from the shell_header and its contents.