Expand description
Column-oriented query results that keep the query’s column order.
The connectors used to hand back HashMap<String, Vec<String>>, which has no
order at all: SELECT a, b FROM t could profile as ["b", "a"], and hash
iteration is not even stable between processes, so two runs of the same query
could disagree. Every other input path reports columns in source order — CSV
header order, Parquet schema order, JSON first-seen field order — so the
database path was the one place where a format conversion reshuffled a
report.
QueryColumns is that map with the order kept: a vector of named columns,
built in the order the driver reports them.
Structs§
- Query
Columns - A query result as columns, in the order the query selected them.
Functions§
- merge_
column_ batches - Merge batches of a streamed query into one result.