B/rbt: Rulebooks Template.
 
@Purpose: To work through the rules in a rulebook until a decision is made.
 
@-------------------------------------------------------------------------------

@p Latest Rule Result.
This used to be a large data structure which kept track of the effect of
procedural rules, but in January 2011 procedurals were abolished. It retains
only one purpose: as a place to record the result of the most recently
completed rule. This used to sit on the top of the stack, and is now the
only thing which ever sits on it. So the "stack" has just one 3-word
record now. The meanings of these are as follows. The first word is one of
the following:

(1) |RS_SUCCEEDS| indicates that the most recent rule or rulebook processed
ended in success. Word 2 is |false| if there's no value, or the kind if there
is, in which case word 3 contains the value itself.
(2) |RS_FAILS| is similar, but for a failure. Note that failures can also
return values.
(3) |RS_NEITHER| is similar except that it cannot return any value, so that
words 2 and 3 are meaningless.

@c
Constant RS_NEITHER		= 0;
Constant RS_SUCCEEDS	= 1;
Constant RS_FAILS		= 2;

Array latest_rule_result --> 3;

[ RecordRuleOutcome usage rule1 rule2;
	if ((latest_rule_result-->0 == RS_SUCCEEDS or RS_FAILS) &&
		(KOVIsBlockValue(latest_rule_result-->1)))
		BlkValueFree(latest_rule_result-->2);
	if ((usage == RS_SUCCEEDS or RS_FAILS) && (KOVIsBlockValue(rule1)))
		rule2 = BlkValueCopy(BlkValueCreate(rule1), rule2);
	latest_rule_result-->0 = usage;
	latest_rule_result-->1 = rule1;
	latest_rule_result-->2 = rule2;
];

@p Following.
Until January 2011, there were two ways to invoke a rulebook: to "follow" it
or simply "process" it. With the demise of procedural rules, these became
equivalent.

In the early days of Inform 7, stack usage became a serious issue since
some forms of the Frotz Z-machine interpreter provided only 4K of stack
by default. ("Only" 4K. In the mid-1980s, one of the obstacles facing
IF authors at Infocom was the need to get the stack usage down to fewer
than 600 bytes in order that the story file could be run on the smaller
home computers of the day.) |FollowRulebook| was the major consumer of
stack space, on average, because of its frequent recursion. Now that the
process is simpler, this has become less problematic, since the routine
now has fewer local variables.

|FollowRulebook| takes three arguments, of which only the first is
compulsory:

(a) The |rulebook| is an I7 value of kind "rule", which means it can be
either the ID number of a rulebook -- from 0 up to $N-1$, where $N$ is the
number of rulebooks compiled by Inform, typically about 600 -- or else the
address of a routine representing an individual rule.
(b) The |parameter| supplied to the rulebook. Much as arguments can be supplied
to a function in a conventional language's function call, so a parameter can be
supplied whenever a rulebook is invoked.
(c) |no_paragraph_skips| is a flag: if explicitly set |true|, then the rulebook
is run with paragraph breaking suppressed. This is the process by which
paragraph division points are placed between rules, so that if two rules both
print text then a paragraph break appears between. While that is appropriate
for rulebooks attached to actions or for "every turn" rules, it is disastrous
for rulebooks attached to activities such as "printing the name of
something".

|FollowRulebook| returns |R| if rule |R| in the rulebook (or rule) chose to
"succeed" or "fail", and |false| if it made no choice. (To repeat: if
the rule explicitly fails, then |FollowRulebook| returns |true|. It's easy
to write plausible-looking code which goes wrong because it assumes that the
return value is success vs. failure.) The outcome of |FollowRulebook| is
stored as described above: thus the most recent rule or rulebook succeeded
or failed if --

	|(latest_rule_result-->0 == RS_SUCCEEDS)|
	|(latest_rule_result-->0 == RS_FAILS)|

and otherwise there was no decision.

@c
Global process_rulebook_count; ! Depth of processing recursion
Global debugging_rules = false; ! Are we tracing rule invocations?

[ FollowRulebook rulebook parameter no_paragraph_skips
	rv ss spv;
	ss = self;
	if ((Protect_I7_Arrays-->0 ~= 16339) || (Protect_I7_Arrays-->1 ~= 12345)) {
		print "^^*** Fatal programming error: I7 arrays corrupted ***^^";
		@quit;
	}
	if (parameter) { self = parameter; parameter_object = parameter; }
	spv = parameter_value; parameter_value = parameter;
	! we won't need parameter again, so can reuse it
	parameter = debugging_rules;
	#ifndef MEMORY_ECONOMY;
	if (debugging_rules) {
		DebugRulebooks(rulebook, parameter);
		process_rulebook_count = process_rulebook_count + debugging_rules;
	}
	#endif;
	if ((rulebook >= 0) && (rulebook < NUMBER_RULEBOOKS_CREATED)) {
		rv = rulebooks_array-->rulebook;
		if (rv ~= EMPTY_RULEBOOK) {
			if (rulebook ~= ACTION_PROCESSING_RB) MStack_CreateRBVars(rulebook);
			if (say__p) RulebookParBreak(no_paragraph_skips);
			rv = rv(no_paragraph_skips);
			if (rulebook ~= ACTION_PROCESSING_RB) MStack_DestroyRBVars(rulebook);
		} else {
			rv = 0;
		}
	} else {
		if (say__p) RulebookParBreak(no_paragraph_skips);
		rv = indirect(rulebook);
		if (rv == 2) rv = reason_the_action_failed;
		else if (rv) rv = rulebook;
	}
	if (rv) {
		#ifndef MEMORY_ECONOMY;
		if (debugging_rules) {
			process_rulebook_count = process_rulebook_count - debugging_rules;
			if (process_rulebook_count < 0) process_rulebook_count = 0;
			spaces(2*process_rulebook_count);
	  	    if (latest_rule_result-->0 == RS_SUCCEEDS) print "[stopped: success]^";
	  	    if (latest_rule_result-->0 == RS_FAILS) print "[stopped: fail]^";
		}
		#endif;
	} else {
		if (debugging_rules)
			process_rulebook_count = process_rulebook_count - debugging_rules;
		latest_rule_result-->0 = RS_NEITHER;
	}
	debugging_rules = parameter;
	self = ss; parameter_value = spv;
	return rv;
];

[ RulebookParBreak no_paragraph_skips;
	if ((no_paragraph_skips == false) && (say__pc & PARA_NORULEBOOKBREAKS == 0))
		DivideParagraphPoint();
];

@p Specifying Outcomes.
The following provide ways for rules to succeed, fail or decline to do
either.

|SetRulebookOutcome| is a little different: it changes the outcome state
of the most recent rule completed, not the current one. (It's used only
when saving and restoring this in the actions machinery: rules should not
call it.)

@c
[ ActRulebookSucceeds rule_id;
	if (rule_id) reason_the_action_failed = rule_id;
	RulebookSucceeds();
];

[ ActRulebookFails rule_id;
	if (rule_id) reason_the_action_failed = rule_id;
	RulebookFails();
];

[ RulebookSucceeds weak_kind value;
	RecordRuleOutcome(RS_SUCCEEDS, weak_kind, value);
];

[ RulebookFails weak_kind value;
	RecordRuleOutcome(RS_FAILS, weak_kind, value);
];

[ RuleHasNoOutcome;
	RecordRuleOutcome(RS_NEITHER, 0, 0);
];

[ SetRulebookOutcome a;
	latest_rule_result-->0 = a;
];

@p Discovering Outcomes.
And here is how to tell what the results were.

@c
[ RulebookOutcome a;
	a = latest_rule_result-->0;
	if ((a == RS_FAILS) || (a == RS_SUCCEEDS)) return a;
	return RS_NEITHER;
];

[ RulebookFailed;
	if (latest_rule_result-->0 == RS_FAILS) rtrue; rfalse;
];

[ RulebookSucceeded;
	if (latest_rule_result-->0 == RS_SUCCEEDS) rtrue; rfalse;
];

[ ResultOfRule RB V F K a;
	if (RB) FollowRulebook(RB, V, F);
	a = latest_rule_result-->0;
	if ((a == RS_FAILS) || (a == RS_SUCCEEDS)) {
		a = latest_rule_result-->1;
		if (a) return latest_rule_result-->2;
	}
	if (K) return DefaultValueOfKOV(K);
	return 0;
];

@p Printing Rule Names.
This is the I6 printing rule used for a value of kind "rule", which as
noted above can either be rulebook ID numbers in the range 0 to $N-1$ or
are addresses of individual rules.

Names of rules and rulebooks take up a fair amount of space, and one of the
main memory economies enforced by the "Use memory economy" option is to
omit the necessary arrays. (It's not the text which is the problem so
much as the table of addresses pointing to that text, which has to live in
precious readable memory on the Z-machine.)

@c
#IFNDEF MEMORY_ECONOMY;
{-array:Phrases::Manager::RulebookNames}
#ENDIF; ! MEMORY_ECONOMY

[ RulePrintingRule R p1;
#ifndef MEMORY_ECONOMY;
	if ((R>=0) && (R<NUMBER_RULEBOOKS_CREATED)) {
		print (string) (RulebookNames-->R);
	} else {
{-call:Phrases::Manager::compile_rule_printing_switch}
		print "(nameless rule at address ", R, ")";
	}
#ifnot;
	if ((R>=0) && (R<NUMBER_RULEBOOKS_CREATED)) {
		print "(rulebook ", R, ")";
	} else {
		print "(rule at address ", R, ")";
	}
#endif;
];

@p Casting.
Nothing needs to be done to a rulebook value to make it a rule value.

@c
[ RULEBOOK_TY_to_RULE_TY r;
	return r;
];

@p Debugging.
Two modest routines to print out the names of rules and rulebooks when they
occur, in so far as memory economy allows this.

@c
[ DebugRulebooks subs parameter i;
	spaces(2*process_rulebook_count);
	print "[", (RulePrintingRule) subs;
	if (parameter) print " / on O", parameter;
	print "]^";
];

[ DB_Rule R N blocked;
	if (R==0) return;
	print "[Rule ~", (RulePrintingRule) R, "~ ";
	#ifdef NUMBERED_RULES; print "(", N, ") "; #endif;
	if (blocked == false) "applies.]";
	print "does not apply (wrong ", (address) blocked, ").]^";
];