[FEA] type casting between fixed-point decimal and other numeric types

[sperlingxx]

What is your question?
According to the checklist of issue 3556, we are able to cast to/from floats, ints and decimal with other precision/accuracy in cuDF. How can we achieve that?
cudf::logical_cast alone doesn't seem to be the right solution. For integer types, we can perform cudf::logicial_cast after rescaling fix_point data to zero. But how to deal with float_point types?
@codereport can you help to give some advice? Thanks!

[harrism]

@codereport I unchecked the box next to

cast to/from floats, ints and decimal with other precision/accuracy

because I looked at cast_ops.cu and I see multiple CUDF_FAIL("Fixed point unary ops currently not supported");.

Casting between types is a unary operation. Apparently this is not implemented yet. Converting this issue into a feature request.

[harrism]

Assigning to @codereport for 0.17.

[sperlingxx]

@codereport @harrism
May I ask about the progress on this issue? The release_0.3 of spark-rapids project heavily depends on this feature, so we really want to know when this feature will be ready.
Sorry for any disturbing :)

[codereport]

I am wrapping up cudf::round and this will be next. Should be done before 0.17 is released.

[shwina]

Oops - apologies. Can someone please move this back to P0? @harrism @kkraus14