magmalake iceberg

iceberg · API

iceberg.schema

Schema — an Iceberg struct type with a schema id and identifier fields.

Structs

AccessorField

struct AccessorField

A field reachable by name: its id, full dotted name and type index.

Implements Copyable, Movable

Fields
id: Int
name: String
type: Int
required: Bool
parent: Int: Position in Schema.flat of the field this one is nested in, or -1.
child_pos: Int: Position among the parent struct's fields, or -1 for a list element and a map key or value — the places a dotted path cannot walk through.

simple

def simple(id: Int, var name: String, type: Int, required: Bool) -> Self

Schema

struct Schema

A named struct type. Owns its TypeStore.

root is the index of the top-level struct node; by_id / by_name are flattened indexes over all nested fields, built once at construction. Nested names are dotted (addr.city); list elements use .element, map keys/values .key / .value, matching the reference implementations.

Implements Copyable, Movable

Fields
store: TypeStore
root: Int
schema_id: Int
identifier_field_ids: List[Int]
flat: List[AccessorField]

__init__

def __init__(out self, var store: TypeStore, root: Int, schema_id: Int)

find_index

def find_index(self, id: Int) -> Int

Position in flat of field id, or -1.

find_field

def find_field(self, id: Int) -> AccessorField

has_field

def has_field(self, id: Int) -> Bool

find_by_name

def find_by_name(self, name: String) -> AccessorField

find_name_index

def find_name_index(self, name: String) -> Int

Case-sensitive first, then case-insensitive (Iceberg allows both).

has_name

def has_name(self, name: String) -> Bool

name_of

def name_of(self, id: Int) -> String

columns

def columns(self) -> List[NestedField]

The top-level fields only.

top_ancestor_id

def top_ancestor_id(self, id: Int) -> Int

The id of the top-level column id lives in, or id itself.

struct_path

def struct_path(self, id: Int) -> List[Int]

Child positions from the top-level column down to field id.

Empty for a top-level field. Raises when the path crosses a list element or a map key/value, which a dotted name cannot address and a row predicate cannot evaluate.

in_struct_only

def in_struct_only(self, id: Int) -> Bool

True when id is reachable from the top through structs alone.

is_top_level

def is_top_level(self, id: Int) -> Bool

column_type

def column_type(self, id: Int) -> Self

A one-column schema holding just the type of field id.

select

def select(self, ids: List[Int]) -> Self

Project to the named ids, keeping ancestors of any selected field.

Selecting a nested field keeps the structs above it and prunes the ones beside it: select(["a.b"]) gives back a struct a whose only field is b. Selecting a whole nested column keeps all of it. A list or a map is never partially selected — its element type is pruned when the selection reaches inside it, but the container itself stays.

highest_field_id

def highest_field_id(self) -> Int

from_json

def from_json(doc: Json, i: Int) -> Self

parse

def parse(text: String) -> Self

to_json

def to_json(self) -> String