Inferring a schema

Draft a schema from example documents instead of writing one by hand.

infer(samples) drafts a schema mechanically from a list of example Documents:

from omnist import infer, doc

s = infer([doc({"id": 1, "tags": ["a"]}), doc({"id": 2, "tags": ["b", "c"]})])
print(s.to_osd())

record Root {
    "id": integer,
    "tags" [0,]: string,
}
root Root

infer returns the raw mechanical inference — it does not normalize. Nested-record names stay 1:1 with the sample's own labels, so structurally-identical shapes under different labels come out as duplicate records. Call .normalize() on the result where a canonical minimal schema is wanted.