panproto-mig

Validates, compiles, and applies schema migrations.

What it does

A migration describes how to map one schema version to another: vertex A in the old schema corresponds to vertex B in the new schema, edge X maps to edge Y, this field was renamed, that field was removed. This crate takes that description, checks that it is consistent with the rules of both schemas (existence checking), then compiles it into a form that can be applied to actual data records.

Applying a migration has three modes, named after mathematical lifting operations. restrict drops everything in the data that the migration does not cover; it is the right operation when you want to project old records into a new, smaller schema. lift_wtype (also called the delta lift) remaps what the migration maps and preserves everything else. lift_wtype_sigma (the sigma lift, or left Kan extension) fills in new fields with computed defaults. The three modes give you control over exactly how much data is preserved or synthesized during a schema transition.

For cases where you do not already know the migration, hom_search can discover candidate morphisms automatically using a backtracking constraint solver, and discover_overlap finds the largest sub-schema that two schemas share.

Quick example

use panproto_mig::{Migration, check_existence, compile, lift_wtype};

let report = check_existence(&protocol, &src_schema, &tgt_schema, &migration, &theories);
assert!(report.valid, "migration has errors: {:?}", report.errors);

let compiled = compile(&src_schema, &tgt_schema, &migration)?;
let new_instance = lift_wtype(&compiled, &src_schema, &tgt_schema, &old_instance)?;

API overview

License

MIT