Struct SchemaBuilder 

pub struct SchemaBuilder { /* private fields */ }

Declaration order never matters: build sorts types by name and fields by ID before encoding, so equivalent declarations produce identical schema ids.

Implementations

impl SchemaBuilder

pub fn new() -> SchemaBuilder

pub fn set_default( self, struct_name: &str, field_id: u16, default: Value, ) -> SchemaBuilder

Give a scalar field a custom default: the value a reader synthesizes via StructReader::get_or_default when the field is absent. default must be a scalar Value matching the field’s type (Value::Enum for an enum field). Applied at build.

pub fn add_struct( self, name: &str, fields: Vec<(u16, &str, Dt)>, ) -> SchemaBuilder

pub fn add_dense_struct( self, name: &str, fields: Vec<(u16, &str, Dt)>, ) -> SchemaBuilder

A dense struct stores no presence bitmap: every field is always present (encoding with one missing is an error). Smaller and slightly faster to read; use for records whose fields are all mandatory, e.g. geometric points, samples, matrix rows.

pub fn add_packed_struct( self, name: &str, fields: Vec<(u16, &str, Dt)>, ) -> SchemaBuilder

A packed struct stores a presence bitmap followed by slots for the present fields only. Field offsets are recovered in O(1) with popcount rank queries over the bitmap — sparse-data wire sizes without giving up zero-copy random access or paying for per-object vtables. Limited to 64 fields. Best for wide records that are usually sparse (events with many optional attributes, config with many defaults).

pub fn add_enum(self, name: &str, variants: Vec<(u32, &str)>) -> SchemaBuilder

pub fn build(self, root: &str) -> Result<Schema>