A case for the BASIC CASE statement

OK Rick, I concede your point, but as you know, the code itself is quite simple and there would be no other way the interpreter would do it. The simplicity leads to oddities and bugs.

Quite interesting is that a WHEN clause ends at the next WHEN, which then jumps to the next ENDCASE. This can be used to exit a CASE structure prematurely, and even provides a way to have the effect of GOTO without line numbers. ;-)

A bug is that the WHEN will search for the next ENDCASE wherever, and that means past END and into the DEF s. ;-(

It behaves the same as real BASIC.

I think it would be rather surprising if it didn’t. The only variation I can really imagine would be one that allowed multiple matches, that would print HMM and then YES! as well – definitely not in spec, but not unreasonable logic.

Quite interesting is that a WHEN clause ends at the next WHEN, which then jumps to the next ENDCASE.

I never thought about it behaving any differently.

A bug is that the WHEN will search for the next ENDCASE wherever, and that means past END and into the DEF s. ;-(

Admittedly I’ve never done anything complicated, and only for me anyway, but my programming style probably stop me from discovering that because my early shell setups tend to be of the form

DEFPROC
REM This will do something if I put code in
ENDPROC

and

CASE X OF
REM This will test stuff
ENDCASE

i.e. empty shell elements.

I can just about imagine having nested CASE structures and miscounting the ENDCASEs. I’ve done it with nested IF{}s in PHP. Quite recently.

Quite interesting is that a WHEN clause ends at the next WHEN, which then jumps to the next ENDCASE.
I never thought about it behaving any differently.

Note that I said next WHEN not next WHEN clause. A WHEN clause starts at the beginning of a line, but the second WHEN that triggers this behaviour can be anywhere, and not a WHEN clause.

Regarding Rick’s insane troll logic example program, the function with three entry points that modifies a global variable. I don’t know if anyone else in the thread has mentioned this, but I wanted to say that it doesn’t work in the Brandy BBC BASIC interpreter – it stops with an error ‘Program execution has run into a PROC or FN’.

I do most of my BASIC programming with the Brandy interpreter on Linux, and when I saw Rick’s example, I expected that it wouldn’t work in Brandy for that reason, which is why I tested it.

I can just about imagine having nested CASE structures and miscounting the ENDCASEs. I’ve done it with nested IF{}s in PHP. Quite recently.

Quickly tested by using the “fold” feature in an appropriate editor¹, properly started and ended procedures etc fold up…

¹Edit don’t hack it here.

Patrick:

but I wanted to say that it doesn’t work in the Brandy BBC BASIC interpreter

I believe ABC would choke on that as well. It’s one of those “undefined” areas of something that works, but probably shouldn’t…

Steve:

Edit don’t hack it here.

Fixed that for you. (looks best in Firefox! :-) )

properly started and ended procedures etc fold up…

I don’t think I’ve ever done such a thing. I know Notepad++ can, as I found the option by mistake (and a brief panic at everything vanishing). My way of dealing with this situation is keeping with a fixed degree of code indentation (two spaces in BASIC, three in C, no tabs EVER); and if the closing element of something is far enough down that the opening element may not be visible, to put a comment afterwards explaining what that closing is actually closing. I’m not so good at remembering this part, but I do try more these days.
Example, in Harinezumi is some code to parse the BootRun file:

   while ( !feof(scr) )
   {
      ln[0] = '\0'; // this prevents annoying ghost line at EOF
      fgets(ln, (CLSIZE - 1), scr);
 

      ...two screenfuls of stuff...


   } // end of block: while ( !feof(scr) )

Further down, there are multiple closing braces together in do_prefileexec() so they are described too:

            }
         } // end of buffer.type == OBJ_FILE block
      } // end of r.r[3] != 0 block
   } while (regs[4] != -1);

The first is closing an if…else so doesn’t need any explanation. The last one is a while clause so is self-explanatory. The other two are explained.

It takes longer, but helps to make the code more readable and ward off problems with unbalanced braces and such. It is perhaps better in C than BASIC as BASIC will fail during program execution, while C will fail hard during compilation. ;-)

Edit don’t hack it here.

Fixed that for you.

Well, yes. Not really a great use for Edit when you’ve got StrongED (which does folding).

I don’t think I’ve ever done such a thing. I know Notepad++ can, as I found the option by mistake (and a brief panic at everything vanishing).

Presumably it’s not an obvious feature in Zap if you’ve never found yourself using it. NotePad++ is a regular work item, can’t understand people at work who do text file comparison with one utility and edit with another.

My way of dealing with this situation is keeping with a fixed degree of code indentation (two spaces in BASIC, three in C, no tabs EVER);

That’s just neatness and easier to follow. You’ll be putting in loads of comments next.

You’ll be putting in loads of comments next.

<cough> <splutter> http://heyrick.co.uk/random/errorcancel.s.txt

BTW, if anybody wants a copy of the RISC OS build of Brandy v1.20 that works on the Pi2…

http://heyrick.ddns.net/files/brandy.ff8

Rename it to be simply “brandy” and set it’s file type to &FF8, then drop it into !Boot.Library so you can *brandy at any time to enter Brandy.

BTW, don’t bother trying to mess around with WimpSlot – this Brandy is hardwired to 524288 bytes unless you pass a different size in the command line. For instance, if the default half-megabyte isn’t good enough, try this:

Brandy -size 2M

for a 2MB version.

Or, if you’re a bit insane:

Brandy -size 511M

Yes, that’s M, not K! And it’s the most you can claim for an application… Works great in the TaskWindow. A mere 535,822,336 bytes free (!).

Have fun!

My way of dealing with this situation is keeping with a fixed degree of code indentation

I do that when I’m being a good boy, or when I start feeling as though I’m liable to get tangled up.

You’ll be putting in loads of comments next.

I would do that if I was being a really good boy. You can tell how often that happens by looking at my code. On second thoughts, probably better for all concerned if you give that option a miss…

Presumably it’s not an obvious feature in Zap

Indeed. What exactly IS “folding”?

On the Mac (writing PHP) I use Atom. It’s got all sorts of features I’ve no idea how to use, but it does what I want it to do. I’ve never miscounted ENDCASES (or ENDIFs) in BASIC (only said I could imagine doing it), but I have miscounted }s in PHP. It’s just so tempting to put them all on the same line, instead of properly indented on separate lines…and then when you’ve just done a restructuring of a nest by cutting and pasting chunks of code here and there, and not bothered to fix the no-longer-appropriate indenting …ooops!

BTW, if anybody wants a copy of the RISC OS build of Brandy v1.20 that works on the Pi2…

Does it work on the Pi3?

but I have miscounted }s in PHP. It’s just so tempting to put them all on the same line,

Gah!

Does it work on the Pi3?

Should. It’s just compiled C. Try it.

Gah!

Yeah. So it crashes. So you go and find out why it crashes, fix it. (They’re still all on the same line. Just there’s one fewer of them now, so it doesn’t eat the one that isn’t supposed to have finished yet…)

Try it.

Will do. And report in due course.

Edit: can’t say it’s thoroughly tested yet, but it is installed and it runs, and runs at least a few small programs. Will report again when I’ve run some of my own apps in it.

So it crashes.

Remember – this is running on Servage’s server in Hamburg, not on my own machine. Happily the server is clever enough not to let my crash cause a system wide crash…didn’t even crash my whole website, just served a few users (possibly – possibly just me) a load of garbage a couple of times.

Indeed. What exactly IS “folding”?

In BASIC imagine clicking on the left edge of the DEF PROC and seeing the full procedure chunk become just the DEF PROC line with a marker at the side of it. If you fold all then all the procedures and functions etc fold up to a start line – great for an overview of the top to bottom code and you can unfold individual subroutines.

As I mentioned earlier it’s also handy for subroutine start/finish matching and more.

A competent programmer would probably do way more with it than I have ever dreamed of doing.
From the lack of other comment from Zap users I suspect a Zap user would wonder what folding was.

… In the normal world, a script error should never bring down the entire machine…

I suspect a Zap user would wonder what folding was.

It would be more difficult to do in !Zap because it would have to remember all the line numbers of the lines not being shown so that folding and unfolding did not alter any line number. One of the main reasons to edit a BASIC programme is after an ‘error at line 34560’ and StrongEd’s habit of mungling line numbers means it is useless for editing BASIC.

… In the normal world, a script error should never bring down the entire machine…

Indeed. And it didn’t.

If you fold all then all the procedures and functions etc fold up to a start line

Ah. Atom does that. I didn’t know it was called folding, that’s all. There’s probably a name on a menu for it somewhere, but I discovered by accident that clicking on little arrows at the left does it.

There are indeed lots of folding options in menus in Atom. Will investigate what they all do!

Zap does have a handy “list functions and procedures” option, which lists them (with line numbers if you have them switched on).

It would be more difficult to do in !Zap because it would have to remember all the line numbers of the lines not being shown so that folding and unfolding did not alter any line number.

Folding is ‘just’ a visual effect, it does not change how things are stored in memory and so line numbers do not play a part in this.

StrongEd’s habit of mungling line numbers means it is useless for editing BASIC.

We’ve been over this before. Conclusion: StrongED is perfectly fine for editing BASIC it just doesn’t fit in with what you’re trying to do.

Zap does have a handy “list functions and procedures” option, which lists them (with line numbers if you have them switched on).

Yeah, StrongED has a similar feature using the List of Found window, which I keep open all the time next to the main code window. If I could save the extent of the LoS window, as well as its default opening position, I’d be super-happy ;-)

One of the main reasons to edit a BASIC programme is after an ‘error at line 34560’ and StrongEd’s habit of mungling line numbers means it is useless for editing BASIC.

Not sure I understand this – I use StrongED to edit BASIC files all the time, and it works perfectly well at identifying line numbers after errors: the message reports the line number with STR$ERL, and then I press F5 to go to the corresponding line in the file. What is happening when you try it?

Not sure I understand this

StrongED does not use the embedded BASIC line numbers, it just counts the physical lines. That’s all good with BASIC saved from StrongED with the default (re)numbering of 1,1 as set in Global Choices.

It goes wrong with BASIC numbered 10,10, by a factor of 10, divide ERL by 10 to find the right line in StrongED. Save the file and StrongED will renumber to 1,1 matching the physical line numbers and the issue goes away.

Zap works directly on tokenised BASIC and uses the BASIC line numbering so there is never an issue.

There are some wierd numbering schemes out there in the wild depending on how the BASIC has been edited. Once in a while Zap will be more useful than StrongED.

StrongED is fine. If one has not noticed a problem perhaps there is not much of a problem there, mostly.

If one wants to be aware or manipulate the BASIC line numbers then use Zap. The point being made is that StrongED “mungles” line numbers which is true but does not matter if one not concerned by the line numbering and is content to let StrongED do what it does

HTH.

HTH

Ah, I see. Thanks for the explanation.