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

To allow the console to be selected at runtime, I have un-singleton'd it.

--
Mats

I've submitted a few updates.

Some small bug-fixes, such as entering just a line-number should REMOVE that line, not end up with a line with just a number on it.

Loading a text file with an empty line in it would end the loading, whcih I don't think is what it should do.

There is a bug when loading text BASIC files by the way. Since the same scanner object is used for both reading the load command, and for inputting the text from the file, when we get back to the command loop, it fails the "scanner.empty()" check by an assert in the STL implementation (In VS 2008 Express Edition at least). I think it is because the string that is used for the input to scanner is "gone".

--
Mats

I'm pushing another refactoring, splitting out the functionality to run, list, load, save etc into separate functions. I will make it table-driven in a bit.

I have found a difference between the original minibasic and our implementation:
doing "list 1000" will list line 1000 only in our implementation. In the original code, it lists FROM line 1000.

--
Mats

Originally Posted by Paul Panks

How are things going?

Paul

I think we have something that is generally functional (and in my experiments, about 3x faster than the original code - albeit probably 10x bigger too... Can't win them all).

--
Mats

Originally Posted by phantomotap

I just took a few minutes to look through the changes and make a few notes.

1): Is the intent of `string awkward_save_helper()' to save with any file extension? If not, the `string awkward_save_helper()' implementation really should use something like `bool ends_with(string const&, string const&);' because most file systems and operating systems in use these days support internal '.' characters. If so, `void load(string name)' should probably drop the implied ".MBI" extension. (If both "test.123" and "text.123.mbi" is in the same directory "LOAD test.123" can't be interpreted as a potentially mistyped "text.123" with the implied extension.)

Indeed, I was just doing some more work on that bit of code, and I was thinking exactly the same thing. And I think we should have different defaults for different types of files - e.g. TSAVE should save as test.bas, where SAVE saves as MBI - not quite sure what BSAVE should use.

2): What is "direct.h" used for?! O_o

Apparently not needed - I just commented it out and it still compiles just fine. I will remove it.

3): After finally realizing the problem I had with extending the virtual character codes, do you think it is time to make a couple of conversion tables so we can get the bigger bang for our precious bits?

I'm quite a fan of table-driven programming, so yes.

4): Did you ever do anymore work with variable names?

Not really - I decided to do some more refactoring first, and haven't got back onto it.

--
Mats

Originally Posted by phantomotap

"not quite sure what BSAVE should use"

".MBB"?

Sounds good to me. I will work on that on the way home tonight. Should be doable if I don't get too sleepy on the train! [The advantage of sitting on a train for about 90 minutes each day is that small projects like this gets done pretty neatly].

On the way here, I moved the long "if token == RUN ... else if token == LIST ... " into a table with function pointers.

"I decided to do some more refactoring first, and haven't got back onto it."

I was thinking around hooking that implementation to virtual character codes. (So that a non-windows implementation wouldn't have to stick with whatever extended keys may be used.)

Soma

Not quite sure how those things are related... Maybe what you were asking had nothing to do with what I was replying to - my reply was referring to making long variable names and allowing multiple arrays. Which makes little sense to your follow on point. So perhaps we can clear up what the original question was meant to be...

--
Mats

I've submitted code to fix up the extension - at least in the basic cases. We probably need a better method, really. But at least we get appropriate names for the respective files.

Also submitted the table-driven parser of run, list, save etc.

And I FINALLY fixed the "load test.bas" causes an error!! I had to make a stack of the scanner, which isn't very good solution, but it's better than an error.

I also fixed the BSAVE command - it was broken if you just load a file and save again, because there was no compiled code. I just added a recompile.

I also shuffled a bit of code around, as per usual

--
Mats

Originally Posted by Paul Panks

How are things going?

Paul

Paul,

Can you get the latest code from
matsp / minibasic / downloads — bitbucket.org
and give it a whirl. If something isn't working right, please let us know. If it IS working, please let us know that too.

If you know someone else who wishes to try it out as well, please feel free to point them towards the link above.

--
Mats

Would anyone object to putting #ifdef _DEBUG around debugging code? A quick hack in interpreter.cpp shows that it's saving about 8KB there alone.

--
Mats

I made another bit of code-clean-up - got about 30KB less code by re-arranging a bunch of things. [That's gcc-code, which I feel is quite bulky - mostly to handle exceptions].

--
Mats