[C++][Compute] Add floor, ceiling, and truncate kernels

[asfimport]

Kernels to round each value in an array of floating point numbers to:

• the nearest integer less than or equal to it (floor)

• the nearest integer greater than or equal to it (ceiling)

• the integral part without fraction digits

  Should return an array of the same type as the input (not an integer type)

Reporter: Ian Cook / @ianmcook
Assignee: Eduardo Ponce / @edponce

Related issues:

• [C++] Arithmetic kernels for numeric arrays (is a child of)
• [C++] Implement mod compute function (relates to)
• [C++][Compute] Add rounding kernel (relates to)

PRs and other links:

• GitHub Pull Request #10727

_{Note: This issue was originally created as ARROW-12745. Please see the migration documentation for further details.}

[asfimport]

Eduardo Ponce / @edponce:
These are special cases of the round kernels (ARROW-12744) or viceversa.

[asfimport]

Eduardo Ponce / @edponce:
@ianmcook why would ceiling/floor spec be to not return integral values? By definition their result is an integral value, and R and SQL both return an integer type. On the other hand, numpy returns float64.

[asfimport]

Ian Cook / @ianmcook:
@edponce I believe that is incorrect; R's ceiling and floor functions return numeric vectors (R's double type). In most SQL engines I've looked at, the ceil and floor functions return a column with the same type as the input.

[asfimport]

Eduardo Ponce / @edponce:
You are correct, I ran examples and R returns "double" type, and SQL engines match output to numeric type of input. I have no objection against this, and now recognize that my doubt arose from mixing the mathematical definition of an integral value with the type system of computing tools. Thanks!

[asfimport]

David Li / @lidavidm:
Issue resolved by pull request 10727
#10727