Type Alias unc_primitives::shard_layout::ShardVersion

pub type ShardVersion = u32;

This file implements two data structure ShardLayout and ShardUId

ShardLayout A versioned struct that contains all information needed to assign accounts to shards. Because of re-sharding, the chain may use different shard layout to split shards at different times. Currently, ShardLayout is stored as part of EpochConfig, which is generated each epoch given the epoch protocol version. In mainnet/testnet, we use two shard layouts since re-sharding has only happened once. It is stored as part of genesis config, see default_simple_nightshade_shard_layout() Below is an overview for some important functionalities of ShardLayout interface.

version ShardLayout has a version number. The version number should increment as when sharding changes. This guarantees the version number is unique across different shard layouts, which in turn guarantees ShardUId is different across shards from different shard layouts, as ShardUId includes version and shard_id

get_parent_shard_id and get_split_shard_ids ShardLayout also includes information needed for resharding. In particular, it encodes which shards from the previous shard layout split to which shards in the following shard layout. If shard A in shard layout 0 splits to shard B and C in shard layout 1, we call shard A the parent shard of shard B and C. Note that a shard can only have one parent shard. For example, the following case will be prohibited, a shard C in shard layout 1 contains accounts in both shard A and B in shard layout 0. Parent/split shard information can be accessed through these two functions.

account_id_to_shard_id Maps an account to the shard that it belongs to given a shard_layout

ShardUId ShardUId is a unique representation for shards from different shard layouts. Comparing to ShardId, which is just an ordinal number ranging from 0 to NUM_SHARDS-1, ShardUId provides a way to unique identify shards when shard layouts may change across epochs. This is important because we store states indexed by shards in our database, so we need a way to unique identify shard even when shards change across epochs. Another difference between ShardUId and ShardId is that ShardUId should only exist in a node’s internal state while ShardId can be exposed to outside APIs and used in protocol level information (for example, ShardChunkHeader contains ShardId instead of ShardUId)