Conditional Hyperblocks?

[glenbull]

@toontalk posted a discussion of conditional uses of hyperblocks here:

How to use hyperblocks with conditional functions

In some cases instead of using map of a conditional function, e.g. [image] One can accomplish the same thing with hyperblocks [image] If the list is one million numbers then the hyperblock version is 22 times faster. But note that the hyperblock version computes the false branch of the conditional so there will be many situations where it is slower. Here is an illustrative project:

I would like to add "2" to each number in a list that is greater than "3". For example, the list [1 2 3 4 5 6] would become [1 2 3 6 7 8]. Can this be accomplished with conditional hyperblocks?

[ego-lay_atman-bay]

Like this?

[glenbull]

Yes, exactly like that. (Thanks!) However, it is not obvious to me why multiplication by two adds two.

Is there a reason that the operations work in this manner?

[scratchmodification]

Because it's a boolean. True is considered as 1 and false 0.

[glenbull]

Aha! So if "False" = 0 and "True" = 1, then "1 x 2" = 2. (Much appreciated!)

[bh]

I go back and forth about whether I like False/True = 0/1. It allows powerful manipulations like this, but it's a sort of unadvertised shift in domain. Sometimes I think we should more fully embrace it, as APL does, with no distinguishing notation for Boolean values, just using 0 and 1. And sometimes I think we should report an error when asked to do arithmetic on Booleans. But this middle ground makes me uncomfortable.

[jens]

I agree about feeling uncomfortable with silently type-casting Booleans to integers. I even wanted to disable that altogether only to find out that all too many projects already depend on it. I do have a bunch of prototypes for variadic inputs to reporter-IF, so stay tuned for more news about this as we keep thinking about solutions.