2009/1/24 John Hartnup
But is that it's only purpose? To placate the LISP haters by removing parentheses?
On Jan 25, 2009, at 12:25 AM, David Morse wrote:
They definitely dropped the ball on introducing '$'. I think you just hit the seam between two of the authors, and it fell through the cracks.
Haskell isn't Cobol, but in comparison with the lambda calculus, it flows a bit better like a natural language. '$' helps with this. For example, in the line
mempty = en $ en []
...one clearly doesn't need to use $, parenthesizing (en []) has the same meaning. However, reading left-to-right, the $ coaxes one into dropping top-down into a new view of the line, partially abstracting out or forgetting what came before, without having to maintain a mental stack and brace oneself for the closing ). I've read books on numeracy and vision, that present the same conclusion: As our brain evolved new systems for seeing or counting, the old systems didn't get thrown out, they got buried by the new layers, but they're still wired in. Part of our reaction to a line of lisp that ends in seven )))))) is no different than a pigeon or a horse seeing the same number of objects; we're not sure. Then for us a new linguistic layer kicks in, "I can see two groups of three, that's even" and we rise above the animals, but our single conscious core got hijacked to sort this out. People take the ergonomics of cockpit design very seriously, to reduce the cognitive load on pilots in emergencies. On Apollo 13, they nearly blew it by flipping a switch the wrong way. I fly a lot, so I'm very grateful that the "purists" that gravitate to math and computer science aren't instead in ergonomic design. I actually love lisp, or more precisely Scheme. I came back to Haskell for strong typing and class abstractions; Haskell more directly supports making a theory about what one is coding. In order to use Scheme, I had to defang it of its parentheses. I didn't believe that any of the web proposals for this saw actual use, so I rolled my own preprocessor, using indentation to indicate outline structure, and using '|' roughly as Haskell uses '$'. There are other issues one inevitably faces, e.g. missing starts to outline levels, but I came to love '$' as Haskell uses it. Some people develop a point-free style in Haskell, e.g.
blah foo . blah . blah foo foo . blah
and different people write code that piles up uses of '$' e.g.
blah foo $ blah $ blah foo foo $ blah
With decent syntax coloring the second line can be given as light a touch as the first. Ultimately, '$' helps mitigate a left / right tension in Haskell that exists in all languages, and in algebra itself. In algebra, "f g" can mean either "f then g" or "g then f". It means "g then f" in any situation crippled by the ancestry of the function composition "f(g(x))", and it means "f then g" any time one has the free choice of easier ergonomics in a culture that reads left to right. Quick, ten times, work out out the permutation cycle product
(1 2 3) (2 3 4 5) (3 4 5 6) (4 5 6 2) (5 6 1 2) (1 2 3) (2 3 4 5) (3 4 5 6) (4 5 6 2) (5 6 1 2)
under each assumption for the order of operations. One wraps around through parentheses either way, but using the "f then g" convention, all other scanning is consistently left to right, not flitting back and forth like a deranged water bug. Now try composing a complex set of type signatures for Haskell functions. Do you feel like a deranged water bug? The arrows go the wrong way, don't they! You'd be playing a nice game of dominoes, canceling off neighbors and so forth, if the arrows went the other way. So Haskell doesn't have its left / right tension completely worked out, but I'd say '$' isn't part of the problem, it's part of the solution.