Missing values

Float columns use IEEE NaN for missing numeric values. The helpers is_nan, fill_nan, any_nan, and count_nan take the float tensor from get_float_col; drop_nan(frame, name) removes rows from a frame. is_nan tests each value with neq(value, value), which is true for NaN.

module Coral.BookMissing
import Coral.Frame (FloatCol, from_pairs, drop_nan, nrows)
export (main)
def main() -> i64 = {
  frame = from_pairs([("value", FloatCol(to_tensor([1.0f32, div(0.0f32, 0.0f32), 3.0f32])))])
  kept = drop_nan(frame, "value")
  nrows(kept)
}

main returns 2.

Integer columns store an i64 tensor alongside a bool mask (true means missing). Use is_nan_col(frame, name), fill_nan_col, drop_nan_col, any_nan_col, and count_nan_col for those columns. get_int_col returns only the values tensor, so use the _col helpers when the mask matters. Grouping skips masked entries for sum, mean, min, and max; counts include the rows. Joins and concatenation carry the mask into their results.

String and bool columns have no missing-value marker. The CSV and JSON writers do not encode an integer mask: they write the underlying number, including 0 in a missing position. See I/O before writing a frame with missing integers.