Module column 

Columnar data model used by the engine and every consumer of query results.

A column is a typed, possibly-nullable, possibly-dictionary-encoded buffer of values plus statistics, encoding metadata, and a none-bitmap.

Invariant: a column’s RowMask and NoneBitmap are always exactly aligned to its logical row count. Operations that grow or shrink the buffer must update the mask and bitmap atomically with respect to consumers; an off-by-one between length and mask silently produces wrong results in filter and aggregation paths.

Modules

buffer

ColumnBuffer and its operation surface: the typed-per-primitive buffer that stores actual values.

columns

compressed

data

ColumnData is the polymorphic accessor (type, length, encoding, nullability, statistics, value extraction) that each concrete data implements. Canonical is the dense default data used for primitive-typed columns. The CompareOp and SearchResult enums express search-and-compare operations the engine performs over arrays during selection (binary search and range probing).

encoding

frame

headers

mask

nones

push

Typed Push<T> trait and its blanket implementations for every primitive ColumnBuffer accepts.

row

row_ref

stats

transform

Pure column transforms: append, filter, and take.

view

Read-only, structurally-typed projections over columns.

Structs

ColumnWithName