ulam cube support

[belisarius222]

          nice. What's the thinking on not having "cube"s?

Originally posted by @tinnus-napbus in #27 (review)

This issue is for discussing this question.

[belisarius222]

The argument against including cubes is that from ulam's perspective, there's no difference between them and non-singleton atoms. There are a couple arguments in favor of including them, though.

One argument is just that they're valid Hoon literal syntax, and a common syntax at that, so it would be more convenient for programmers to include them -- they won't have to remember that there's a special case of atom literal that isn't supported in ulam.

For strings, %foo saves a character over 'foo'. That's nice if you're typing them all the time.

I think the strongest argument in favor of including cubes in ulam would be if there's a way to use them to encode $% values in a way that preserves the intent, i.e. some way to make it clear from reading the ulam syntax that the value came from a $% where the cube is the tag.

There is some similarity here between ulam's %page type and a $%. The $page could be thought of as a special case of a $%, but where the tagged value is just a raw noun. It's notable that ulam can encode raw nouns already, as $coin literals, so it's not clear we need special functionality for %pages.

Instead, we could have ulam syntax for $% tagged by cube, where the tagged value is itself an ulam. This could reuse the existing Hoon irregular syntax foo+val, which in Hoon is a shortcut for [%foo val].

{
  ~zod: 0x0,
  ~bus: foo+0x1
}
::
^-  ulam
:-  %map
%-  gas:((on ulam ulam) lte-ulam)
:~  [[%coin %$ %p ~zod] [%coin $% %ux 0x0]]
    [[%coin %$ %p ~bus] [%bccn [%tas %foo] %coin %$ %ux 0x1]]
==

The syntax could be extended to something like "tall form" in ulam by allowing whitespace after the +.

foo+
[0x0 0x1]
::
^-  ulam
[%bccn [%tas %foo] [%list [%coin %$ %ux 0x0] [%coin %$ %ux 0x1] ~]]]

As for inferring that the cube could be a mark tag, I could see two options: one is to have a different syntax for something we expect to be a page, such as foo/ instead of foo+. The other option, which I think I prefer, is to treat any %bccn as a potential mark tag. Data inside the tag could be mark-converted from ulam to that mark, and if the tagged ulam is a raw noun %coin, then the mark converter would convert from the %noun mark to the specified mark.