change xi macro to #% tag macro

[gilch]

As discussed in #1439. This replaces xi with a more Clojure-like #% tag macro.

[ekaschalk]

Do you have thoughts on whether to highlight the %i? I don't know how clojure does it.

[gilch]

You mean, "Does clojure-mode?" It does, I checked. Should Hy? I'm less certain. This is not builtin syntax. This is not even a core macro.

But, if we conventionally used %-prefixed anaphors in other macros (as Clojure does), then highlighting such symbols would make sense. It does improve readability. But some might argue the old xi symbols were more readable, because that's what won out in #879, even though I presented % symbols as an option.

We could use #xi with the old xi symbols. Maybe it's more readable, but it's also more likely to interfere with user symbols. Names like x1, x2 could totally come up. Tradeoffs.

[ekaschalk]

I meant "does and what face", I can check into it.

A middle-ground is to implement is as an option, your call whether to enable it by default or not.

[gilch]

I added a %** anaphor for the &kwargs parameter. This makes #% usable for partial application even if the function takes keyword arguments. I also changed the &rest parameter from Clojure's %& to %* for consistency.

For example, #%(foo #* %* #** %**) would do the same thing as function foo, regardless of its signature.

[Kodiologist]

This is cool. I like it. Is it ready for review? Still, I would finish up your outstanding PRs (and review those of others) before starting new ones like this.

[gilch]

Hy can't be my top priority all the time. I'm busy with other things right now, but this seemed like a really easy fix, especially since I wrote the code in question in the first place.

It's a tag macro, so I've generalized #% to work on arbitrary expressions.
For example, you can just use #%[%1 %2] instead of (xi identity [x1 x2]) like before.

Clojure would have to use #(identity [%1 %2]) or #(vector %1 %2) for that, but it saves the extra % in the more common case of a normal s-expression. We'd have to build function literals into the compiler to do it Clojure's way.

It should be ready for review now.

[Kodiologist]

A lot of the diff is whitespace changes to old code. If you want to include them in this PR, could you put them in a separate commit?

Instead of "presence of a % parameter", I would write "presence of symbols named %*, %**, or %i for some number i".

The tests seem to pass without the (import hy) in anaphoric.hy, despite the use of HySymbol. So I think you can remove it.

(if (coll? expr) (flatten expr) (, expr)) can be written (flatten [expr]).

You have a commented-out require at the bottom of anaphoric.hy. I'm guessing it's old debugging code.

Otherwise, all looks good.

[gilch]

whitespace

Done.

Instead of "presence

I reworded it, but not quite that way.

(import hy)

Yep.

(flatten [expr])

More elegant, I like it.

old debugging code

Removed.

@Kodiologist, how about now?

[Kodiologist]

Stray comma.