Skip to main content

dict_u32_utf8

Function dict_u32_utf8 

Source
fn dict_u32_utf8() -> DataType
Expand description

A dictionary wide enough that a block cannot fill it.

Every other enumerable string in the schema is u16 and seals the block on overflow (see DICT_CAP). Attribute values are the one string here that is genuinely unbounded — a trace id, a URL, a GenAI prompt — so a u16 would turn a high-cardinality tenant into a block sealed every few thousand rows. u32 costs about one percent of the compressed size over u16 and removes the failure mode: 4 billion distinct values do not fit in a 32 MiB block.