wishlist

<B>Digital的Unix操作系统VAX 4.2源码</B>

Date: 25 Apr 1980 1546-PST (Friday)
From: mike at UCLA Computer Science VAX (Mike Urban)
Subject: ATTN: DDL Implementors
To: cc (UCLA Computer Club)

   I would like the following changes made in DDL:

1) Friendlier Parsing.  You should be able to say 
	give troll the knife
   and get away with it (with "knife" as Dobj and "troll" as Iobj).  You
   should get messages like "I don't know how to tke" in the event
   of a bad verb, and "I think 'north' is a verb" in the event
   of a weird parse like "push north wall".  You should be able
   to say things like "e.s.s.take thing. eat thing. look" and have
   the parser treat the "." as equivalent to and end-of-line.  
		/* ACCOMPLISHED 6/13/80.  Syntax errors are now
		of the form "I don't know how to xxx" if an unknown
		word is encountered (the whole word, not just 5 letters
		is printed); "I don't know the word 'xxxxxx'" for other
		unknown words; and "I don't understand, because I 
		think 'xxxxxx' is a [noun/verb/adjective/etc]" for
		other syntax errors.  The separator was already
		built into the lexer, but it's a comma, not a period.
		And VERB IOBJ DOBJ is a legal syntax now, but 
		VERB PREP IOBJ DOBJ isn't (tho it's easy to add) */
	MAYBE
   (if it isn't hard), you should be able to say "take this, that, the other"
   and have a more sophisticated control loop which causes it to act
   like "take this.  take that.  take the other" with the EXCEPTION that
   no daemon/fuse processing is performed (no turn increment) and that
   the whole thing can be short-circuited by an appropriate Exit code.

2) Generalized routine and object handling.  The routines associated with
   objects should be properties of that object just like the others.  The
   constants that tell WHICH property is, say, the Action property should
   be predefined.  So ($setp room1 LDESC Rubbl) would set the long 
   description of room1 to the routine Rubbl (which describes the rubble
   left by the explosion you just triggered...).   

	/*** ACCOMPLISHED 6/6/80 ***/

3) Improved global handling.  The programmer shouldn't have to know the
   "address" of his globals.  Instead, a construct like 
 	VAR score
   would be the equivalent of saying "score = xxx" where xxx is one less
   than the last VAR done.  

	/*** ACCOMPLISHED 5/29/80 ***/


4) Improved grammeme access.  There should be predefined VARs which
   contain the values of ($verb) ($dobj) and ($iobj).  You should
   be able to re-assign these.  SO, you should be able to say
       turn off lamp   
   and the PreVerb routine for "turn" might contain something like
	(($eq @Iobj off):
		($setg Verb douse)
		($setg Iobj 0)
	)
		/*** Accomplished 6/12/80; Only change in
			ddlrun is #defines for verbtodo, etc */

   Because verbs don't have properties, we may wish for functions
   to access their preverb and action routines as well.

5) String handling:  It'd sure be nice to have someone type
	incant "abra ca dabra" 
   and be able to talk about ($subs Dobj @var 1) or ($indx @Dobj "abra")
   or whatever.  I don't know exactly how to implement it yet.
		/*** Accomplished 7/28/80;  STRINGs are negative-valued
		objects.  If the lexer gets a string, it puts it into
		a temporary string area (that is flushed every turn) and
		returns an index (not pointer) into that are as a negative
		number.  The parser knows about these and can handle
		verb string  and  verb string prep iobj   sentences.
		(the latter being for "Incant "rise" over seed".)

		If the dobj or iobj of an input is a string, the dobj
		and iobj routines called are the ones associated with
		the object STRING (if such is defined by the DDL
		programmer).  At present, the functions $substr, $length
		and $eqst  are defined.  $substr returns a negative
		pointer into the temp string area; this can be stored
		in a global, for example.  A permanent string area
		is available, but I haven't written the function 
		($sets glbl string) which would move the string to the
		permanent area and put its index therein into glbl.
 				Mike		***/
		/*** On 8/20/80, I added a new function (which, mirable dictu,
		will allow coding of the "echo" room!), ($read), which
		pauses, gets a line of input, and returns the line as
		a string!  Gee, if we code up ($pars str verb dobj iobj)...

6) Ctrl-D should terminate the program.  Ctrl-C should interrupt whatever
   is happening and begin a new turn.
   Note that this may leave things in a bad state.

7) New language construct: (WHILE condexp : formlist) is essentially
   equivalent to (condexp : formlist ($____ formlist+condexp)) where
   "($____ formlist+condexp)" is a new operator, JUMPB (jump back),
   which is just like the current $.... jump forward instruction, but
   (obviously) negative, and unconditional.  It goes back exactly to
   the beginning of the test.
		/*** ACCOMPLISHED 8/18/80 --Mike ***/

8) New routine ($rand n) produces a somewhat random number in the range
   1..n .	/*** ACCOMPLISHED 8/18/80 ***/

9) DDLRUN sources should be reorganized so as to enable recompilation
   of less than the entire program if some less global change (such
   as the addition of a new routine, or a bug fix in the parser) is
   made.
		/*** Somewhat accomplished 10/80.  Still needs 
		     work.  In particular, changes in extvars.c must
		     be reflected in extvars.h ***/

10) The symbol table, and the hash routine, are simple-minded.  The
   5-character limitation on names is a real pain for some people.
   For example, you can't even fake the Zork puzzle room by allowing
   someone to say "push northern wall".  
   These changes don't have to go together.

		/*** ACCOMPLISHED(!!) 11/80.  Identifiers are pretty
			much as long as you like.  The symbol table,
			now alphabetized, contains indices (not pointers)
			into a pool of identifiers that starts at
			sbrk(0).  A binary search algorithm is used.
			This will enable a future version to respond
			correctly to unambiguous abbreviations for
			things at runtime (like news for the aard
			newspaper, for a cogent example).  This
			hasn't been tested for portability to a
			PDP-11 under V7, so beware. As of 11/18,
			the symbol table is still unencrypted ***/
		/*** The runtime parser recognizes unambiguous 
			abbreviations.  The symbol table is still
			unencrypted.  1/81 ***/

11) It should be available on a micro.
			/* Not yet.  But the thinking has begun! */
			/* This will necessitate some rearchitecture
			   especially wrt buffering of routines/strings */

12) The ACTION routine for verbs should execute AFTER the action 
    routine for your room; this enables east(ACTION) to correctly
    default to ($say "Can't go that way.") and allow one to generally
    live without $hit and $miss, crocky routines that they are.
    See ../sample (espec. transit.ddl) for a specimen of how this works.

			/* 11/80 */

13) Routines that get the action and preaction routines for verbs
    should be made available.  

14) Fuses and daemons aren't QUITE right (fuses go off after your daemon
	has typed a prompt, for a real-world example).  Maybe they
	shd be on a homogenized priority queue?

15) It should be frozen real soon.
			/* As of 1/81, we're STILL thinking in terms
			   of "just this ONE more feature..."  Damn. */



	Mike