Joins

Coral.Join covers inner_join, left_join, and outer_join with string keys. The implementation uses host-path equality matching rather than sort-merge ordering.

module Coral.BookJoin
import Coral.Frame (from_pairs, nrows, int_col_of_list)
import Coral.Join (left_join, outer_join)
export (main)
def main() -> int64 = {
  left = from_pairs([("customer", StringCol(["a", "b", "a"])), ("qty", int_col_of_list([cast(1, int64), cast(2, int64), cast(3, int64)]))])
  right = from_pairs([("customer", StringCol(["a", "c"])), ("score", FloatCol(to_tensor([cast(10.0, f32), cast(40.0, f32)])))])
  nrows(left_join(left, right, "customer"))
}

outer_join

outer_join includes all rows from both frames. Left-only rows get NaN/empty for right columns; right-only rows get 0 with an integer mask (missing) for left int columns and NaN for left float columns.

Row order: left-sequential traversal first, then right-only rows appended at the end.

Semantics notes

• Right-hand key column is suppressed in the output
• Overlapping right column names are suffixed with _right
• Bool output columns from the joined right side are deferred
• Integer columns from the source frame have their missing-value mask propagated through joins