Converting MINI-BASIC in MASM over to C++?

**matsp** · 04-20-2009

Right, so I've just reorganized most fo the code that Sebastiani has written into separate .cpp files, renaming the header-files into .h instead of .hpp.

It compiles and I think it works just as well as before (pitman works reasonably well, but some console stuff gets a bit messy).

I think we should do more refactoring and split up "minibasic.cpp" into a couple of files, but I haven't looked into it very much.

--
Mats

**Sebastiani** · 04-20-2009

OK, I'll be posting some updates as well, soon, including some major changes to the FOR loop functionality (whereas before you could only use named variables, you can now use @ variables as well), the addition of the TIMER function and other minor extensions, and formatted the code to the original stype. By the way - where is console.h and token_scanner.h? I don't see them anywhere. Oh, and although I really don't mind changing the '.hpp' files to '.h', it is a pretty standard extension for C++ header files. No big deal, though, just thought I'd point that out. And I think the experimental code is stable enough to put it in the main directory.

Thoughts?

>> I think we should do more refactoring and split up "minibasic.cpp" into a couple of files, but I haven't looked into it very much.

Ugh. My only objection there is that it is extremely time consuming. =/

**matsp** · 04-21-2009

Drat! I forgot to add those files. They aren't hard to recreate, but I guess it's better to wait until tonight (my time) when I can submit them. Expect around 7PM BST.

--
Mats

**Sebastiani** · 04-21-2009

Oh, don't worry about it - done.

**matsp** · 04-21-2009

I don't know if I've broken something - it seems to not work at all right now, exits either immediately or after the first letter. Did you test your code before pushing?

--
Mats

**matsp** · 04-21-2009

Found it. It was a case of "assign instead of equality test"

Code:

diff -r 75584b6b32c9 -r d7a2a568bcb3 minibasic/experimental/console.cpp
--- a/minibasic/experimental/console.cpp        Tue Apr 21 21:50:42 2009 +0100
+++ b/minibasic/experimental/console.cpp        Wed Apr 22 00:18:29 2009 +0100
@@ -217,7 +217,7 @@
                        }
                }
        }
-       if( ch = '\r' )
+       if( ch == '\r' )
                put(ch);
        return ch;
 }

I'll submit that and a few minor cosmetic changes.

--
Mats

**matsp** · 04-21-2009

And I've changed a few long if-else chains to switch statements. Makes the code easier to read, I think, and also possibly more efficient as we don't have to walk down a long chain of if-else statements (not that it's a big performance hit at the moment)

--
Mats

**Sebastiani** · 04-21-2009

Sounds good.

**phantomotap** · 04-21-2009

I finally decided to crack open the source to see what you two have been doing... fitting my work will require a sledge hammer... or a scalpel.

I'll try to work on that some more later, but for now... please... pull down the latest update before doing any more work. I somehow got a bad set when I tried to merge my few updates to Mats updates. It was a pain... and I think I lost a bug fix in the progress. (I worked my changes into commit 37 manually, and I'll be happy to remove the macro bits later.)

Anyway, the way the `get' helper routine(s) are implemented a simple approach to handling ESCAPE simply will not work. I needed this feature to test my theories, but I didn't want to break anything by trying to patch it; I've hacked a silly callback mechanism until it can be done right. (It can be pulled out later in a few seconds.)

Nearest I can tell, the approach we've all taken to the padding feature is totally, utterly, miserably wrong. I've done the mathematics, and I believe fixing this feature will fix the "PITMAN.MBI" display problems. (I recorded the offsets we generate and compared those results with what the original and source seem to expect.) I've left minimal notes under the "issues" section of the "BitBucket" project page. I can't find anything else at fault... even the "EDIT" mode of "PITMAN.MBI" seems to display properly. (I imagine that this is because it uses a different layout engine for the "EDIT" mode display.)

For whatever reason, interrupting "MBI" execution with ESCAPE followed by loading the same "MBI" back into memory will result in garbage. I'm thinking the compiler or virtual machine needs to be reset after an escape, but I didn't know if that would trash the loaded source as well, and I didn't have time to experiment any further.

Soma

**phantomotap** · 04-21-2009

I did lose a fix: after you compile the latest source type in something and hit backspace. (The cursor appears to be in the wrong position until a character is typed.)

Oh... yea... and... my editor strips trailing whitespace so my set pushed a lot of unnecessary changes; to add insult, anything I've touched will use spaces instead of tabs because I forgot to change the settings. Sorry.

Soma

**phantomotap** · 04-22-2009

Rather than make you type in code, I've added the "MBI" to the repository.

I grabbed another hour to experiment.

I've uploaded some more source. I don't know the source well enough to put the hacks where they best belong; I've only just punned a hack that may correct the padding issues. (This will have to be cleanly fixed by someone who knows the source better if it proves correct.) The "PITMAN.MBI" program is still not correct, but it looks much better. (That said, the one issue may not be an issue of `void minibasic::show_pad()' but an artifact of one of the below problems.)

I've also commented out the scrolling code because it behaves weirdly.

Before either of you get upset about the punning and the tracing stuff... I've found two pretty crazy bugs. I'm listing them here for now, but if I can't find a fix in the next 15-20 minutes I'll probably try and grab a minute to pin a note to the "issues" section.

1): The expression "64+1*3" is a silly way of saying "67" so how come the following code prints "Stage" and my little men starting at the same horizontal offset?

2): How come running the code twice without a reset puts "Stage" in a two different places?

Soma

Code:

10 CLS
15 LET L=4
20 LOCATE 70,1
25 PRINT "Stage ";#2;1
30 FOR A=1 TO 4
35 LET X=3:LET Y=64+A*3
37 LOCATE 0,0:PRINT X:LOCATE 2,0:PRINT Y
40 COLOR 2,0
45 LOCATE Y+1,X
50 PRINT " 0 "
55 LOCATE Y+1,X+1
60 PRINT "/X\"
65 LOCATE Y+1,X+2
70 PRINT "/ \"
75 NEXT A
80 FOR Z=1 TO 25
85 NEXT Z

**matsp** · 04-22-2009

Originally Posted by phantomotap

I did lose a fix: after you compile the latest source type in something and hit backspace. (The cursor appears to be in the wrong position until a character is typed.)

Oh... yea... and... my editor strips trailing whitespace so my set pushed a lot of unnecessary changes; to add insult, anything I've touched will use spaces instead of tabs because I forgot to change the settings. Sorry.

Soma

Speaking of which: The current code, I believe is a mix of tabs and spaces. We probably should use one or t'other.

--
Mats

**phantomotap** · 04-22-2009

GET THE LATEST SOURCE!

Paul:

Unless Mats removes it, the link below my name will take you to a page where you can download our latest push. Please download it and try it with a few programs. You've used the original, a lot apparently, and should know when something is misbehaving. Please report back any issues you discover.

Mats:

The "experimental" branch is now the default, and the old default branch is now "depreciated".

Sebastian:

Your `std::string operator & <???> (const std::string &, const <???> &)', while cute, breaks MSVC. (The compiler literally crashes.) I've not explored this, but I imagine I can fix it. If you would though, do you mind trying to separate it from the class proper(non-membe, non-friend operator)?

Everyone:

I think I've found and fixed a few more bugs.

As far as I can tell, I've fixed all of the bugs I've referenced. This naturally doesn't mean I've found them all; I'm sure to find more if I keep looking, but I've got other stuff to do for the rest of the week. (Which should explain my trying to get a lot done today.) I hope you guys, who should know more about the source, can look around at my changes and see where it will best fit into the design.

The "PITMAN.MBI" program displays perfectly as do several other programs. That said, the one Sabastini made doesn't work at all. (I accidentally "removed" "REIGN.MBI" because I thought it was something I created by accident.) I'm not entirely sure what is happening. My only thoughts here: Sabastini wrote his program under the false pretense of the documentation. (Naturally, I could have broke something in fixing the others, but I hope not.)

Soma

matsp / minibasic / downloads — bitbucket.org