Re: qsh from cl

On Tue, Mar 6, 2018 at 3:44 PM, Evan Harris <auctionitis@xxxxxxxxx> wrote:

Personally I think the idea of a whole pile of parameters with no keywords
would be even uglier than the verbose string the CL prompter produces.

IBM-supplied commands don't let you have "a whole pile" of parameters
with no keywords. Usually just two or three. But the whole point is
that those parameters are the ones that are (generally speaking)
needed the most, and are the most obvious.

Command design is an art, and there certainly is such a thing as too
many positional parameters.

Despite the perceived ugliness including the keywords removes pretty much
all the possible ambiguity.

Not just pretty much. It removes exactly all the ambiguity. That is
why it is, as a pure matter of objective fact, the safest option.

If I was that horrified I think I would write a CL "prettifier" in Python,
but I've been looking at CL so long I find the prompts comforting.

Don't think I haven't imagined doing that. But I don't think it's
worth the investment of time. The balance of "data" versus "algorithm"
is tilted heavily toward data, because the prettifier would have to
know which keywords can be left out. You would have to essentially
build a database of prettifiable commands and their parameters. Which,
if you want to prettify a lot of commands, would probably be more
tedious than fun. The indenter portion of the prettifier would be more
algorithm-oriented, though, so that could be fun.

Actually, prompted CL code feels to me kind of like "generated code"
more than it feels like code written by humans, for humans. (After
all, it *is* kind of generated, though just one line at a time.) As
such, it would totally make sense to create some other language which
is aesthetically nice but "transpiles" to CL (in a similar way that
CoffeeScript and TypeScript transpile to JavaScript). That's more
effort but also probably more fun and potentially more useful than a
prettifier. With CL's increasing capabilities, it becomes an
increasingly attractive target for transpilation.

John Y.