Some small column improvements. One can now hand other int / float types (int8, float32 etc.) as arguments to toColumn or straight df["key"] = ....
Note, for arraymancer releases prior to https://github.com/mratsim/Arraymancer/pull/437 will see slight performance regressions on assignments of float and int tensors, since asType will map the whole tensor regardless of the input / output types.
Also []=(df: DataFrame, key) now returns a var Column instead of Column. This allows to directly change the retrieved column. Since both the data frame, the column and arraymancer tensor are reference objects anyways, this only makes sense.
Some small column improvements. One can now hand other int / float types (int8, float32 etc.) as arguments to
toColumn
or straightdf["key"] = ...
.Note, for arraymancer releases prior to https://github.com/mratsim/Arraymancer/pull/437 will see slight performance regressions on assignments of
float
andint
tensors, sinceasType
will map the whole tensor regardless of the input / output types.Also
[]=(df: DataFrame, key)
now returns avar Column
instead ofColumn
. This allows to directly change the retrieved column. Since both the data frame, the column and arraymancer tensor are reference objects anyways, this only makes sense.So
df["x"][idx] = val
is now valid.