Crate nescio 

nescioDB

nescio (lat.): “I do not know”. A working implementation of Aporia — a database whose primary object is ignorance, not values.

A slot without evidence is not NULL; it is a region of maximal entropy. Evidence narrows regions, time widens them again (erosion as storage physics), couplings let knowledge flow between slots. A classical relational database is the limit case in which every evidence is axiomatic and every region is a point.

The verbs

• BOUND — engine::Query::bound: credible region + entropy in bits + MAP estimate. How ignorant is the DB, really?
• SAMPLE — engine::Query::sample: one consistent world, deterministic under a seed, couplings respected.
• RESOLVE — engine::Query::resolve: which minimal-cost evidence would push a slot’s entropy under a target? The DB plans its own data procurement, across slot boundaries, Monte-Carlo-validated.
• FIND — engine::Query::find: region queries across entities (“all objects whose price certainly lies below 600k”).
• JOIN — engine::Query::join: entity pairs matching a relation, each with a probability and a three-valued certainty — joining two regions is itself uncertain.
• certainly — engine::Query::certainly: three-valued predicates as region containment: true / possible / false.

Quick start

use nescio::prelude::*;
use std::collections::BTreeMap;

let mut slots = BTreeMap::new();
slots.insert("price".into(), Domain::Continuous { lo: 0.0, hi: 1e6, n_bins: 200 });
let schema = Schema { slots, couplings: vec![] };
let broker = Source { name: "broker".into(), reliability: 0.85,
                      half_life_days: Some(90.0), axiomatic: false };
let mut db = Db::in_memory(schema, vec![broker]).unwrap();

db.ingest(EvidenceRecord {
    entity: "house_1".into(),
    claim: Claim::Interval { slot: "price".into(), lo: 400_000.0, hi: 500_000.0 },
    source: "broker".into(),
    observed_at: 0,
}).unwrap();

let q = Query::new(&db, 86_400); // one day later
let b = q.bound("house_1", "price", 0.95).unwrap();
assert!(b.entropy_bits < b.max_entropy_bits);

Modules

binlog

The binary evidence log — the production on-disk format.

calibrate

Half-life calibration: learn a source’s decay physics from the log.

engine

The Aporia engine: BOUND, SAMPLE, RESOLVE, FIND over the evidence log.

error

model

The data model: what nescioDB stores and reasons about.

prelude

rng

Deterministic RNG: splitmix64 seeded via FNV-1a over string parts.

server

nescio serve: the database as an HTTP/JSON service.

store

Persistence: a database is a directory.

time

Time: unix seconds internally, ISO dates at the edges.

watch

Watches: standing questions with knowledge horizons.