This document is produced by TC X3J14 as its clarification of questions raised about ANSI X3.215-1994, American National Standard for Information Systems - Programming Languages - Forth.

The questions covered herein were raised by query

Q0007, regarding the distinction between "immediacy" and "special compilation semantics".

There are four parts in this document:

  1. The original question as received.
  2. The TC's reply.
  3. The Letter Ballot issued by TC Chair.
  4. TC Chair's statement of ballot results.

Q0007 as received

Subject: Q0007 Request Recognized & Draft 0
To: X3J14@minerva.com
Cc: wykoh@pado.krict.re.kr
Date: Sat, 23 Dec 95 02:11:38 PST  
The following query has been assigned number Q0007. - Greg Bailey, by direction 951223 0700Z
Wonyong Koh wrote on Fri, 22 Dec 95 16:27:03 PST :
Dear Mr. Bailey,

I would like you to ask to interpret the ANS Forth Stanard. You might already know about it. In recent discussion in comp. lang.forth I and Mr. Jonah Thamas have different opinion on CONSTANT. I interpret that making CONSTANT create IMMEDIATE words does not violate ANS Standard while Mr. Thomas thinks it does.

I think that it is a obvious choice to make CONSTANT words STATE-smart in subroutine-threaded Forth (i.e. native Forth compiler). IMMEDIATE CONSTANT words pushes a number on top of the stack in interpretation state while compiles in-line machine code in compilation state. I beleive that it is not TC's intention to leave significant number of native-code Forth compilers out of ANS Standard System. Please clarify this.

Wonyong Koh
wykoh@pado.krict.re.kr

TC Reply to Q0007

From: Elizabeth Rather
Subject: Q0007R, "immediacy" and "special compilation semantics", Official Response
To: X3J14 Technical Committee
Cc: lbarra@itic.nw.dc.us
Date: Mon, 19 Feb 96 14:10


Doc#:  X3J14/Q0007R
Reference Doc#:  X3.215.1994 ANS Forth
Date:  February 19, 1996
Title: Response to Request for Interpretation Q0007, Clarifying the
distinction between "immediacy" and "special compilation semantics"

Q0007: Clarifying the distinction between "immediacy" and "special compilation semantics".

Wonyong Koh wrote on Fri, 22 Dec 95 16:27:03 PST:
Dear Mr. Bailey,

I would like you to ask to interpret the ANS Forth Stanard. You might already know about it. In recent discussion in comp. lang.forth I and Mr. Jonah Thamas have different opinion on CONSTANT. I interpret that making CONSTANT create IMMEDIATE words does not violate ANS Standard while Mr. Thomas thinks it does.

I think that it is a obvious choice to make CONSTANT words STATE-smart in subroutine-threaded Forth (i.e. native Forth compiler). IMMEDIATE CONSTANT words pushes a number on top of the stack in interpretation state while compiles in-line machine code in compilation state. I beleive that it is not TC's intention to leave significant number of native-code Forth compilers out of ANS Standard System. Please clarify this.

Wonyong Koh
wykoh@pado.krict.re.kr
Wonyong Koh is literally correct, in the sense that optimizing compilers which add special compilation semantics to classes of definitions such as CONSTANTs are specifically supported by ANS Forth. Applying the concept, in these terms, to ANS Forth has very clear meaning and leads to no inconsistencies.

However, using the term "immediate" to speak of such compilation semantics is not recommended since this term carries with it all of the implications of the implementation model prescribed by earlier standards such as FORTH-83. Attempting to discuss the advanced capabilities available to ANS Forth compilers in terms of "immediacy" ineviteably leads to confusion, misunderstanding, and the eventual "discovery" of apparent inconsistencies that do not, in fact, exist in the Standard (although they may indeed exist in a particular implementation.) This in turn can lead to arguments over moot points.

In particular, a CONSTANT with special compilation semantics is *not* the same thing as is a CONSTANT declared as follows:

     123 CONSTANT JOE  IMMEDIATE
... but in ANS Forth terms, this usage is the *only* sense in which the word "immediate" or the concept of "immediacy" is defined or meaningful. Always discuss ANS Forth compilation semantics as such, not as "immediacy", for clarity and to avoid pitfalls, even when discussing an implementation that uses classical immediacy as its way for expressing compilation semantics.

Analysis of Specific Case

However, upon review of the usenet discussion you reference, it would appear that you may have added special compilation semantics to CONSTANTs without adding all of the communication with the compiler that is necessary to deliver the proper semantics in all cases when such CONSTANTs are used. The test case that failed seems likely to have encountered one of these probable inconsistencies, as opposed to "caring" whether the CONSTANT definition in question was "immediate" in your implementation as such.

At the same time, the ability to declare an IMMEDIATE constant, as in 123 CONSTANT JOE IMMEDIATE had been precluded.

Careful review of the semantic requirements for definitions when used in compilation, interpretation, and execution, as well as the results of obtaining execution tokens and using them with EXECUTE or COMPILE, , and when referenced following [COMPILE] (if COMPILE, or [COMPILE] are to be supported) should make it clear that significant flow of information is required between the compiler and the definition when extending normally "default" compilation and/or interpretation semantics of compliant definitions.

*The standard does not define the way in which this information flow takes place.* While a standard program may extend "macros" such as control flow words as illustrated in the examples, it is not possible to define words of the sort that normally have "default" compilation and interpretation semantics such that they have special compilation or interpretation semantics without depending on implementation specific knowledge. If this is done without knowing how to communicate with the compiler, the resulting definitions may deliver the correct semantics in *some*, but cannot in *all*, cases.

For example, consider a Standard Program that adds special compilation semantics to DUP as in the following example:

    VARIABLE #DUPS
    : DUP   1 #DUPS +!  POSTPONE DUP ;  IMMEDIATE
If all this did was to count compilations of DUP, it would be compliant since the additional work done affects nothing else and is therefore OK (See section 3.)

However, this definition does not comply because it does not yield the correct semantics in all cases.

And, coincidentally, execution of its tick is abysmally slow (which would take more information flow to resolve, but which is not required since it only affects performance as opposed to correct execution.)

If the word was made state smart to distinguish between interpret and compile states, so that it counted #DUPS and did the POSTPONE when in compilation state and simply executed DUP when interpreting, then the situation changes but is still not correct:

And execution of the tick, even when correct, is still slow.

There is, at minimum, a bidirectional flow of information between the compiler / interpreter and the definition of DUP that must occur around ticking and [COMPILE] ; in its absence, the compiler does not have any way of asking the definition for an *execution* token, and the definition does not have any way to provide one; and, in the case of [COMPILE], the compiler has no way of knowing that this definition is supposed to be treated as though it had default compilation semantics.

The TC did not attempt specifying this compiler interface because it affects only new capabilities which were almost universally impossible in earlier standards, and active experimentation is still underway with these new capabilities, including as it appears your own. At some time in the future after more has been learned it may be appropriate to standardize such a message passing mechanism, however at present any *application* which adds compilation semantics like this will need to incorporate implementation dependent code... or it will yield noncompliant definitions, as illustrated above.

In the problem you have been discussing, you are the implementor so you certainly have the power to create and then employ the necessary interface. However, it appears that you have overlooked this necessity, and that is probably why you have been having difficulty.

Discussion of Immediacy, Transparency, and Compliance

In FORTH-83 it was necessary to document the "I" (Immediate) attribute of every standard word (and for that matter every word in each application) because the application needed to be aware of this attribute in order to make correct selections between use of COMPILE and [COMPILE] , and in order to predict the consequences of using what are now called "execution tokens". Since knowledge of immediate versus nonimmediate attribute for each existing word appeared in application programs in both explicit and implicit ways, it was not practical to change the "immediacy" of any word in the system, nor indeed in most applications. The attribute of immediacy was the only tool for making or discussing the distinc- tion between compilation and execution behavior, and the need for program awareness of this attribute as well as its inelastic nature led to a number of well known inconsistencies. In addition, it made compliance with FORTH-83 impractically cumbersome for many advanced compilers, including the incremental optimizing sort.

ANS Forth addresses these problems by defining all words in terms of explicit compilation, interpretation, and execution semantics; by defining such words as IMMEDIATE and [COMPILE] and the effects of EXECUTE with execution tokens as specific transformations and selections on these semantics; by introducing POSTPONE to solve a well known dilemma in selection of the appropriate semantics; and finally by deleting the "I" designations since it is no longer necessary for the application programmer to be aware of them as such. Finally, having done these things, it is possible for a compiler to add any desired compilation semantics, including those useful for optimization purposes, to definitions without affecting the use of those definitions by applications. IMMEDIATE remains available for application use, but its effects are specified in semantic terms that can coexist with arbitrary default compilation semantics.

In ANS Forth, the existence and nature of compilation semantics that are in effect messages to the compiler is concealed well from the programmer, and the application code is the same whether or not, for example, DUP is "immediate" (in the old terms) or "has special compilation semantics" (in ANS Forth terms). The only significant caveat is that when a system supports [COMPILE] it must be very careful to treat all definitions that the Standard says have "default" compilation semantics, but to which the system has given "special" compilation semantics for its own purposes, as though they had "default" compilation semantics when referenced by [COMPILE].

An implementation complies with the standard if all Forth words and classes of words defined by the Standard behave, in all cases, consistently with the semantics prescribed in the Standard. The test for correctness is specifically limited to semantics, as opposed to nonsemantic considerations such as, for example, the size, speed, or binary representation of compiled programs.

Please refer to D.4, D.5, D.6.7, and E.4.1 for further discussion of optimizing compilers and ANS Forth handling of "immediacy".

Letter Ballot

X3 Subgroup Letter Ballot
Authorized by X3 Procedures - Distributed by X3 Subgroup X3J14
Project: X3J14, ANS Forth
Doc#:  X3J14/LB018
Reference Doc#s:  X3J14/Q0007R, X3.215.1994 ANS Forth
Date:  February 19, 1996
Title:  Response to Request for Interpretation Q0007, Clarifying the
distinction between "immediacy" and "special compilation semantics"
Ballot Period:  30 Days
Ballot Closes NOON DATE:  March 21, 1996
Respond to:  greg@minerva.com
        or:  Elizabeth D. Rather, Chair
             FORTH, Inc.
             111 N. Sepulveda Blvd.  Suite 300
             Manhattan Beach, CA  90266
             (310) 372-8493    FAX (310) 318-7130
             erather@forth.com

Statement:
    Document X3J14/Q0007R contains a proposed Response to Request for
    Interpretation Q0007.

Question:
    Do you agree that this response represents the intended interpretation of
    X3.215.1994 ANS Forth?


/------------------------  begin response area----------------------\
|
|  YES____ NO____ ABSTAIN____
|
|  Signature:  [not required for email ballots]
|  Name:
|  Organization:
|
|  Explanation (REQUIRED for NO or ABSTAIN votes):
|    
\------------------------  end response area  ----------------------/

INSTRUCTIONS:
Please return the entire letter ballot with your response _either_ by email
to greg@minerva.com _or_ by regular mail or fax or email to me at the above
address, before the closing date & time.

   If replying electronically PLEASE edit only within the response area
   indicated above, inserting any explanatory text in place of .
   Any changes made outside that area will likely be overlooked.

All TC members must vote.  Failure to vote in two consecutive ballots may
cause you to lose your voting rights in X3J14.

Thank you for your participation.

Elizabeth D. Rather, Chair, X3J14

Results of Letter Ballot

Letter ballot 18 closed at noon March 21 with the following results:

        Y  N  A NV
LB18:  12, 0, 1, 1

Abstention from John Hayes.