Support multi-digit macro arguments in angle brackets

[Rangi42]

This allows access to arguments past \9 without using shift.

[Rangi42]

There's no open issue for this feature but I remember it being suggested before. (Full expressions inside parentheses wouldn't be practical because that's parsing, not lexing.)

Edit: it was issue #460. At the time it was closed because of "the difficulty in implementing the original request, on top of a suitable workaround existing". However, readParentheticMacroArgNum is not a complex implementation, and the suggested workaround was to "expand \#, split on commas, and pick the Nth element", which is more cumbersome than the original approach using shift.

@aaaaaa123456789 's original reversedb example would be simplified to:

reversedb: MACRO
  for ___current, _NARG, 0, -1
    db \<___current>
  endr
ENDM

And pokecrystal's object_event would be:

object_event: MACRO
	db \3, \2 + 4, \1 + 4, \4
	dn \6, \5
	db \7, \8
	dn \9, \<10>
	db \<11>
	dw \<12>, \<13>
_NUM_OBJECT_EVENTS = _NUM_OBJECT_EVENTS + 1
ENDM

[Rangi42]

This also removes the out-of-date documentation lines about macros having max 256 arguments.

[ISSOtm]

While I certainly don't have any problems with the code change (except maybe how many forward declarations are needed? :S), I'm less sure about the syntax change.

This syntax couldn't be \{42}, for example, because that's already an escape used to remove {'s special meaning. But then, giving \( special meaning means that we won't be able to give ( any special meaning. There's nothing wrong with the change per se, but I want us to be sure that we won't regret this decision later.

[Rangi42]

Agreed about the forward declarations; I'll see if there's a rearrangement that reduces their overall count.

You're right about parentheses: #704 proposes keeping track of parentheses nesting in macro arguments and allowing commas inside them, so e.g. mac fn(1, 2) would pass fn(1, 2) as \1. To disable that the parentheses would have to be escapable, like mac fn\(1, 2).

Alternate syntaxes: brackets \[10], angle brackets \<10>, or various paired punctuation (\$10$, \^10^, \:10:, etc). (Not single quotes \'10', that would conflict with potentially escaping \' within '' for character literals or some kind of special string.) Brackets might be getting used for inline fragments and/or structs, and while neither of those uses currently demands a \[ escape, it's plausible that we'll end up with one. I don't think \< would ever be necessary though; it's just for the < and << operators. How about it?

[ISSOtm]

\$ sounds like hex, so not that one; \^ looks weird imo. That seems to leave \: and \<, so I guess \<. It should be possible for us to be bitten regardless, but it seems unlikely, so the risk is imo acceptable.

[Rangi42]

(Angle brackets are kind of appropriate for being related to macro args. Some older assemblers used them to surround argument values that needed actual commas, e.g. mac <1,2>,3 would have \1 be 1,2 and \2 be 3.)

(Also, this assembler allowed 36 arguments, \0-\9 and then \A-\Z. Being case-sensitive, \N wouldn't even conflict with \n. :P Of course that's irrelevant now with \<36> and above.)