Hacker News Reader

Comment by Starwatcher2001 | Hacker News Reader

Starwatcher2001

There was a bug in the TRS-80 basic (which was indeed Microsoft). It had an "on x goto" command that allowed the program to move to different lines depending on the value of a variable:

ON X GOTO 200,300,400,500 depending if x=1, x=2,x=3...

They'd obviously used the same parser for the standard "If a=1 then goto 200" construct, which allowed an undocumented shortcut of "If a=1, 200" which a few people noticed and started using. However, a later model fixed this, which broke a number of programs.

View on HN

This was legal in BASIC from the very first versions of the language:

   IF a=1 THEN 200

and when they added ON..GOTO, they basically made these two mirrored syntactically. So you could do:

   IF a=1 GOTO 200

   ON a THEN 100,200,300

(I suspect it was simply treated as identical token?)

ooh I never worked on TRS80's - so they had a version of the Arithmetic IF :-)

I could have Told BillG that was bad idea back when I was 19

  x = x + (x < 5)

No If needed.

Atari Basic would allow crazy things like:

  10 Goto ((x < 5)*100) + 50

Goes to either line 50 or 150, depending on the comparison result being 1, or 0)

Or...

  Goto (100*x)

Faster On goto with no checking, lol.

Bill Gates was not involved with the original Atari Basic. That could be purchased on a cartridge for running MS Basic programs.

Did Atari BASIC use 1 for TRUE? Most BASICs used -1 (i.e. all bits 1), because this allowed the same operators to work for both bitwise and Boolean logic, in the absence of a dedicated Boolean data type.

https://atariwiki.org/wiki/Wiki.jsp?page=Boolean%20Logic%20i...

Yes, it used 0 and 1. And it did not contain bitwise ops. Those were quite the surprise!

My first efforts were in Applesoft. Computers at school. Got an Atari machine and programmed the crap out of it. Wrote my own assembler, a few games, etc...

Then I got an Apple and finding them in Applesoft was a nice surprise! Prior to that, I thought they were an assembly only thing.

The trade off in the Atari basic was doing things mathematically was pretty easy and it generally, but not always, was faster.

A few of us wrote fairly complex programs with basically no if statements. Kind of a fun way to think about things back then.

I much prefer -1 and 0 and the bit ops, frankly. Internally, I remembered this kind of thing as "really fucking true" as in every bit better signal, or it's false!

The problem with using bit ops as logical is that they don't short circuit. Although it was unfortunately common even in languages that had proper Booleans at the time - e.g. standard Pascal didn't have it, and Turbo Pascal required a compiler directive/switch to enable it, which few even knew of. So it feels worse in retrospect than it actually was back then.

On the other hand, one thing that BASIC still has to this day (in VBA) that no other language I can think of does, is bitwise equivalence (EQV) and implication (IMP) operators. Furthermore, since VBA has a Boolean data type, these operators are also available for Booleans; implication comes in handy sometimes.

Not sure what you mean by short circuit.

I had no idea about those VBA features. Will read about them.

Having the bit ops themselves is a very strong feature.

My simple assembler would definitely been faster and simpler with them. Graphics, some math, all benefit. People would make little language extensions or callable machine language routines to make the bit ops available.

In basics with bit ops, and the -1 as true:

X = 100 & (y = 5)

Atari

X = 100 * (y = 5)

Not doing the multiply seems like a net gain in a limited CPU too. That basic did not really have types, like MS basic did. Pretty much floating point only.

Interestingly, basics that have the bit ops do not always allow the expressions in a goto. Or, could just be my own memory, or lack of trying. Not really needed.

In fact, outside of trying to optimize, or just explore the language, the only case that made real sense was a sort of jump table making good use of line numbers like one would do with memory addresses in assembly.

Say, branching directly on a checked or masked character value.

And the spiffy thing was making pseudo labels. Put line numbers into variables and use them like labels.

Foo = 1000

Goto Foo

Overall though, it was a weird little basic. It nicely supported the machine, and the device independent I/O system in the OS ROM, but was slow otherwise.

This guy has actually made an N: driver meaning Atari Basic programs written long ago can operate across the Internet today.

Lol, cool old stuff.

https://m.youtube.com/user/tschak909

Short circuiting means not evaluating the right side if the left side is sufficient - so e.g. in (Foo() OR Bar()), if Foo() returns true, it won't even evaluate Bar().

Most modern languages are like that, but BASIC is an exception to this day - even VB.NET still doesn't do short circuiting by default. You need to use special operators, AndAlso/OrElse, to get it.

Computing the line number in a GOTO is definitely an extension - the original Dartmouth BASIC never had anything like that. For jump tables and such, you were supposed to use ON..GOTO.

Makes perfect sense. Maybe nobody cares enough about Basic, or figures there are a zillion better performing options? Seems like an optimization like that would have been done at some point

Yeah, the computed goto is odd.

Back in the day, I used it and some conditional math to greatly shorten and improve performance in some programs. My first machine was only 16k. Using the graphics system in any meaningful way cost a lot in terms of program space. So, every dirty trick, including poking new addresses into string variable address pointers to overlap the graphics screen got done.

Absolutely terrible code though. I am sure if I saw some of that code today, I would have a bugger of a time sorting it out.

There were no bitwise operators in the Atari BASIC, and numbers in the tokenized program were always stored as floating-point numbers in BCD (Binary Coded Decimal) format.

Reply on news.ycombinator.com