panproto-inst

Represents concrete data records and transforms them as schemas change.

What it does

Schemas describe structure; instances are the actual data that conforms to a schema. This crate provides three shapes of instance to match the three shapes of data you encounter in practice: tree-shaped records (JSON objects, XML documents), table-shaped records (SQL rows, CSV files), and graph-shaped records (property graphs, RDF). All three live under a single Instance enum so generic migration code does not need to special-case each shape.

The main operation on instances is the restrict/extend pipeline. When a schema changes and a compiled migration is available, wtype_restrict transforms a tree record from the old schema to the new one by following the migration map. wtype_extend (the left Kan extension) goes further and fills in new fields using default expressions from the migration. For cases where you need the inverse direction, restrict_with_complement saves the dropped data into a Complement value so the full original record can be reconstructed later. These operations are what the schema data migrate CLI command calls under the hood.

JSON serialization is schema-guided: parse_json uses the schema to resolve ambiguities (which union variant applies, whether a null means absent vs. explicitly null), and to_json uses the same schema to produce correctly structured output. Validation checks the seven structural axioms that every well-formed tree instance must satisfy.

Quick example

use panproto_inst::{parse_json, to_json, validate_wtype, wtype_restrict};

let instance = parse_json(&schema, "user", &json_value)?;
let errors = validate_wtype(&schema, &instance);
assert!(errors.is_empty());

let new_instance = wtype_restrict(&compiled_migration, &src_schema, &tgt_schema, &instance)?;
let output = to_json(&tgt_schema, &new_instance);

API overview

License

MIT