c9f7427967
* Organize docs: front and back end; custom predicates. * Whoops forgot to hit save before git commit last time -- delete stuff moved out of values.md * Update book/src/values.md --------- Co-authored-by: Ahmad Afuni <root@ahmadafuni.com>
1.3 KiB
How to hash a custom predicate
Every predicate, native or custom, is identified on the backend by a predicate ID.
The native predicates are numbered with small integers, sequentially. The ID of a custom predicate is a hash of its definition; this guarantees that two different predicates cannot have the same ID (aside from the miniscule probability of a hash collision).
This document explains in some detail how the definition of a custom predicate is serialized and hashed.
Custom predicates are defined in groups (also known as batches); see an example. The definition of a custom predicate in a group involves other predicates, which may include:
- native predicates
- previously-defined custom predicates
- other predicates in the same group.
Predicate hashing is recursive: in order to hash a group of custom predicates, we need to know IDs for all the previously-defined custom predicates it depends on.
The definition of the whole group of custom predicates is serialized (as explained below), and that serialization is hashed (using a zk-friendly hash -- in the case of the plonky2 backend, Poseidon) to give a group ID. Each predicate in the group is then referenced by
predicate_ID = (group_ID, idx)
(here idx is simply the index of the predicate in the group).