home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
AMIGA PD 1
/
AMIGA-PD-1.iso
/
Programme_zum_Heft
/
Programmieren
/
Kurztests
/
ACE
/
archive
/
ACE23.LHA
/
MAIN.lha
/
utils
/
a-a
/
a-a.doc
< prev
Wrap
Text File
|
1994-10-22
|
12KB
|
256 lines
From XRACTON@FULLERTON.EDU Thu Jun 24 00:39:05 1993
Return-Path: <XRACTON@FULLERTON.EDU>
Received: from csu.Fullerton.EDU by leven.appcomp.utas.edu.au (4.1/SMI-4.1)
id AB05310; Thu, 24 Jun 93 00:38:13 EST
Received: from FULLERTON.EDU by FULLERTON.EDU (PMDF #2446 ) id
<01GZPKDRJYSS002QFT@FULLERTON.EDU>; Wed, 23 Jun 1993 07:36:24 PST
Date: 23 Jun 1993 07:36:23 -0800 (PST)
From: ROLAND ACTON <XRACTON@FULLERTON.EDU>
Subject: A-A documentation
To: dbenn@leven.appcomp.utas.edu.au
Message-Id: <01GZPKDRJYSU002QFT@FULLERTON.EDU>
X-Vms-To: IN%"dbenn@leven.appcomp.utas.edu.au"
Mime-Version: 1.0
Content-Transfer-Encoding: 7BIT
Status: OR
AMOS to ACE
Version 1.0
June 22, 1993
INTRODUCTION
------------
This program is intended for people like me: people who have a
big, bulky application program written in AMOS BASIC. AMOS is nice
for development, because it's interpreted, and runs fast. But
programs written in it have a lot of problems - such as opening up
their own screen (which is not Intuition-compatible), not
multitasking with each other, and not working properly with the
68040. This makes AMOS programs murder to run for any serious
computer user - and most of them won't.
ACE has its own set of problems, but most of them are compiler
limitations. They're not things that would affect the end user of
the application program. When you're writing programs for fame,
instead of fortune, that's an important consideration.
AMOS and ACE are not completely compatible. Most of the commands
found in ACE are in AMOS, but some of them have the parameters in a
different order. There are a few cases where both programs have the
same function, but require a different command to execute it. This
can make converting from one to the other rather tedious - and once
you do, you have to either drop the original version, or (try) to
develop them both in parallel.
AMOS to ACE is designed to help. With it, you can keep one version
of your program (in AMOS format) and convert it to ACE when you're
ready to compile it. It has quite a few limitations, and can't
convert anywhere near all of AMOS's command set (most of them have
no analogues in ACE), but should be very useful to serious
developers (who don't want to admit that they're writing programs in
AMOS). The program is user-extensible, and enough information is
provided in this documentation to allow interested users to
customize the program.
SUPPORTED CONVERSIONS
---------------------
X, Y, A$, B$, ect. represent the positions of variables in commands.
AMOS VERSION ACE VERSION
Inc X ++X
Dec X --X
Add X,Y-Z X=X+Y-Z
Free FRE(-1)
Instr(A$,B$,X) INSTR(X,A$,B$)
Upper$(A$) UCASE$(A$)
End If END IF
If(X=Y) IF (X=Y) THEN
Do REPEAT
Loop UNTIL 1=0
Rnd(X) RND
Fix(A) FIX A+1
Say "hello",1 SAY "hello"
Open Out 1,"out_file" OPEN "O",#1,"out_file"
Open In 1,"junk_file" OPEN "I",#1,"junk_file"
Append 1,"my_foo_bar" OPEN "A",#1,"my_foo_bar"
Hex$(X,Y) HEX$(X)
Bin$(X,Y) BIN$(X)
Deek(A) PEEKW(A)
Doke A,Y POKEW A,Y
Leek(A) PEEKL(A)
Loke A,Y POKEL A,Y
Procedure HELLO[X,Y,Z] SUB HELLO(X,Y,Z)
Procedure HELLO SUB HELLO
End Proc END SUB
Proc HELLO[X,Y,Z] CALL HELLO(X,Y,Z)
Proc HELLO CALL HELLO
Set Buffer 100 <blank line>
String$(A$,X) STRING$(X,A$)
Rename "hello" To "goodbye" NAME "hello" AS "goodbye"
Lower$(A$) LCASE$(A$)
Note that the LCASE$ function is not part of ACE's command set. It
is found in <strings.h> and must be included.
Although a version of the MID$ command, with different syntax, is
also in <strings.h>, A-A does not convert to it from AMOS. AMOS's
version of the MID$ command has a serious bug, and should never be
used. (Note that AMOS's MID$ FUNCTION works properly)
I've been told that in a recent version of AMOS Pro the MID$
command bug has finally been found and corrected, but in all prior
versions (and in all versions of the original AMOS) the bug still
exists.
HOW IT WORKS
------------
A-A works by comparing each program line against several dozen
internal templates. If a match is found, the "corrected" version of
the match is output. If not, the first character of the line is
deleted from the internal buffer and output to the destination file,
and the process begins again. A typical template looks like this:
CONVANYFROM$(11)="Rnd(\1)"+Chr$(10)
The "R", "n", "d" and left parenthesis characters are "constants" -
if they don't match with the input line, the comparison immediately
fails. The backslash is special - it is the "variable" character.
A-A has nine internal "variables" which it uses to rearrange parts
of the input line. The backslash must be followed immediately by the
variable number, and then by the "success-character" and the
"failure-character". When A-A sees the backslash, it scans through
the input line, looking for both the success and failure characters.
If the failure-character appears before the success-character, or
the success-character does not appear at all, the match fails.
Otherwise, the input line, up to the point that the
success-character was found, is placed in the specified "variable",
and A-A continues comparing the template against the input line,
starting from the position right after where the success-character
was found. Notice that the success-character in the example is a
right parenthesis, and the failure-character is a linefeed (the
Amiga's end-of-line character).
If any match fails, all variables are cleared before the input
line is compared against the next template. Also, if the same
variable appears multiple times in a template, the appropriate
section of the input line is added onto the existing contents of the
variable.
If a template matches completely, a definition such as this is
used to determine what to output:
CONVANYTO$(11)="RND"
All characters specified, except for the backslash character, are
output verbatim. Again, the backslash character is special - it is
followed immediately by a variable number, and the contents of that
variable are inserted into that position in the output. In the
example, the contents of the variable have been discarded - ACE's
version of the Random function does not take a parameter.
There are two kinds of templates in A-A. The ones in the arrays
beginning with CONVBEGIN are checked against the input line only
once, starting with the first character. This is because certain
commands (such as IF) can only legally appear at the beginning of
the line, and so it is not necessary to check for them across the
entire line. Commands in the arrays beginning with CONVANY (such as
the Instr function) can appear at any place in the input line, and
so are checked for at every position.
The template conversion method has some problems. The worst of
them appears in connection with strings. If A-A sees the input line
PRINT "Let Freedom Ring"
it will erroneously convert the Free in Freedom to FRE(-1).
Another problem occurs because of AMOS's IF statements. There are
actually two versions of the IF statement in AMOS - one that has a
"then" and one that does not. The scope of the "then" version is
limited to the current line; the other version must be terminated
with an END IF. A-A will always add a THEN onto the end of an IF;
this will cause ACE to generate an error if there was already a THEN
there. Fortunately, though, AMOS has an odd restriction about the
"then" version - it can't be used if you're already inside an IF (of
either kind). Thus I never use the "then" version, because I would
have to rewrite it if I ever put another IF around that block of
code. Hopefully, most other AMOS users think the same way.
A third problem occurs because the conversion routines are not
recursive. If A-A converts an IF statement, it will not convert the
statement's parameters.
A-A does have one "special case" patch - if the success-character
in a variable is a right parenthesis, A-A will keep track of the
number of left and right parentheses it encounters in the input
line, and not match with any nested right parentheses.
These problems, and others, can be solved (with some annoyance) by
using the conditional conversion feature.
CONDITIONAL CONVERSION
----------------------
Sections of code can be designated as AMOS-only or ACE-only. This
is done in the following way:
Rem Begin AMOS
...
<AMOS-only code>
...
Rem End AMOS
Rem ACE ...
Rem ACE <ACE-only code>
Rem ACE ...
Note that these Rem statements must be typed EXACTLY as shown.
When A-A sees the text "Rem Begin AMOS", it will discard all lines
after that until it sees the text "Rem End AMOS". It will then place
"Rem AMOS" into the output stream to mark the place (in case you
want to look at the output file) and continue as normal.
When A-A sees the text "Rem ACE", it will strip it out and place
the remainder of the line into the output stream, without performing
any conversions. It will then continue as normal.
Although this feature was mainly intended to allow you to replace
AMOS commands (particularly graphic-oriented ones) with ones that
will work under ACE, it can also be used to get around the template
converter's limitations. Make a copy of the "problem" line, manually
do whatever conversions are necessary, and designate the old and new
lines as AMOS- and ACE-only, respectively.
Making use of this feature will affect the size of your source
code and (marginally) the speed of your interpreted program. It will
not affect the final compile, because only the ACE sections will get
to the compiler.
VARIABLE TYPE DIFFERENCES
-------------------------
AMOS has two kinds of numeric variables: integers (4 bytes) which
do not have a qualifier, and floating-point variables (4 bytes),
which have a "#" as a qualifier. ACE has four kinds: short integer
(2 bytes, "%" qualifier), long integer (4 bytes, "&"),
single-precision (4 bytes, "!"), and double-precision (8 bytes,
"#").
In AMOS, if no qualifier is appended to a variable name, it is
treated as an integer. In ACE it's treated as single-precision.
This isn't a problem because A-A automatically puts a DEFLNG a-z
directive at the beginning of the output file, which makes all
integers the same length as AMOS.
The difference is with the floating-point variables. In AMOS, a
4-byte float has a trailing hash character; in ACE it has a trailing
exclamation point. The easiest course is simply to leave the
variable alone; ACE will just see it as a variable of greater
precision than it originally was. This should cause no problems (I
hope).
At the moment, double-precision variables aren't implemented in
ACE - they're treated by the compiler as single-precision. So, for
now, there is no potential incompatibility problem.
OTHER STUFF
-----------
There are two extra subroutines in A-A called BEFORECHECK and
AFTERCHECK. These are called on each character "step" through the
input line, before and after (respectively) the line is compared
against the templates. At the moment, BEFORECHECK is being used only
to implement the conditional conversion feature, and AFTERCHECK is
blank. You can place extra code in the subroutines to make A-A
handle your specific personal requirements.
Interested parties can contact me at:
Roland Acton
8001 Bluebird Lane
La Palma, CA, 90623
U.S.A.
Internet: xracton@ccvax.fullerton.edu