Printable View
Say we were going to include 8bit ints. How would we go about this? THe C data types are as follows:
http://www.phim.unibe.ch/comp_doc/c_...ata_types.html
As you can see, there is no 8 bit ints. Only 16 bit. I can implement 16bit easy enough, but really, what is the point for a 8bit?
CornedBee, I figured you would have a good reason to keep those data types around. :) Thats good enough for me. Thanks again.
8 bit int is more commonly known as byte or char.
You may want to distinguish between a "char", which is a single ANSI/Unicode character, and a "byte", which is a binary octet.
Well all pointers are essentially the same - a data address. The difference typing a pointer makes however applies to how it can be used. You can declare an int pointer (int* myIntPointer), set it to a particular address and when you dereference it (*myIntPointer) you get an int out of it. In other words it reads the size of an int, from the address it holds. Now on the other hand if you set a char pointer (char*) to that same address, you would read a char instead of an int. It is the same difference typing regular variables makes, instead of working with raw chunks of memory (for which you could use the generic void*). Type safety means restrictions on how pointers can be used, which also includes casting to and from void pointers.Quote:
Originally Posted by eyeRmonkey
As for arrays, here are my preferences
Just as suggestions. Use of the "new" operator can be used to differentiate between stack/heap allocation, if you like.Code:// fixed-length array
char[2] trimChars = {" ", 0};
// dynamic array
char[] newArray = new char[newLength];
// unchecked dynamic array
unsafe char[] tempArray = new char[(EndPos - StartPos) + sizeof(int)];
Have you thought about how to do classes yet?
Your takes on how to implement this in LF?Code:class IntNumber
{
int _Value;
new (int Value)
{
_Value = Value;
}
void Add (int Value)
{
if (((_Value \ 2) + (Value \ 2)) > Maths.Power(2, 30))
throw new OutOfRangeException();
else
_Value += Value;
}
int cast
get
{
return _Value;
}
set (value)
{
_Value = value;
}
}
class FreakyIntNumber inherits IntNumber
{
overrides new (int Value)
{
base._Value = (int)((float)Value * 0.86);
}
}
I like how you did arrays. I know we discussed unsafe arrays earlier on and they sound reasonable to me. I'm not sure how easy they will be to implement in C, but they are our list. I do think safe arrays should be the default though.
That looks good for classes. Could you bring up the stuff you mentioned about how you thought properties should work? I think you mentioned no global variables or something? The discussion you and CornedBee got into went over my head.
Also, If you could expain what "overrides" and "base" do in that class example, that would be awesome. Other than that things look good.
Yeah, I guess it's just a personal preference, Str just looks a little Calculator-Basic to me :)Quote:
Originally Posted by eyeRmonkey
Well for situations such as dealing with characters (ranging from 0 - 255) or reading from binary files, etc., how can you say that there is no point on having it?Quote:
Originally Posted by Girvo911
Girvo, maybe you missed CornedBee's post on the bottom of the last page (page 24). He gives many good reasons for including 8-bit. One of the more notable points was that we need to be compatible with the Win-API structure.
I'm also not sure how we would convert the the 'char' data type to a 'byte' data type. Maybe we don't need to? Aren't letters treated as numbers if you store them in a byte normally? I know some languages do this. Like this in this statement:
The char is treated as a number, right? The only problem with not ACTUALLY doing something to make a 'byte' data type is if you did this:Code:If('A' < 95) {...}
Code:Byte x = 97
Str myStr = "Hello {x} World" // the {} is the LF way to insert variables into strings.
/* That should return "Hello 97 World" if you did it VB, but if
we just use char as out byte data type then that would return
"Hello a world" */
When you team members design the syntax with me, try not to make it look totally like C++. There's already other languages that look too much like it as it is.
the char data type is the same as a byte data type. no conversion needed. it is called char because that is the size that will store any ascii character.Quote:
Originally Posted by eyeRmonkey
That is what I mentioned in my post (#969). My other point was that bytes are treated as numbers by default and chars are treated as numbers by default? Is that correct? Or are they always treated as integers and you have to use CStr() if you want it to treat it as a char?
Plus if byte is the same as char then why do we need char if we are going to have a UInt8 data type?
Because char will be targeted more towards string characters.Quote:
Originally Posted by eyeRmonkey
Also I've been doing some thinking and I think the semicolons should be optional. So both the new line character that the enter key produces and semicolons can be supported for those who feel one over the other. The same should hold true for the () around the expressions in If statements. It too can be optional.
I know char is geared more towards charcters. I trying to make 2 points: 1) When Buyte (in LF) is converted to C will it be stored as byte? 2) If not then how will we convert it. If so, then will there be extra handeling to deal with it as a string and if not then how will we we store it?
As we discussed earlier in the thread, optional semicolons are almost useless. People will rarely choose to type them and then readability is decreased. I guess it is useful for people who want to keep readabilty and if you choose not to then thats up to you. If the IDE inserts them for you 99% of the time then I don't see why we shouldn't have them. Does anyone else anything to say on this subject?
As far as () around the expression of an IF statement, I see no reason not to include them. It is barely any extra typing and it makes parsing easier and increases readability. Especially since we plan to have a bunch of weird extra syntax rules like multiple condition expressions ( If (5 < a < 10) {...} or If (a < c, d, e) ).
Actually it would be easy for the parser no matter what. Anything between If and the new line character or {} is considered an expression. ;)Quote:
Originally Posted by eyeRmonkey
Sigh... Fine we might make them optional. I see some advantage to this, I guess. Girvo, what are your thoughts on making those 2 things optional? Would it make your job a lot harder because you have 2 deal with 2 possible situations every time?
Does anyone see any downsides to doing it this way?
It won't be hard to do this at all. A simple Or operator can take care of this situation in code ;)
So I don't see a downside.
Don't be ridiculous Jake. You have to realize that parsing isn't that easy. Even with the tools Girvo is using. There are a ton of ways its not as simple as a "Or". For example, the way we though of for doing variables in strings: "Hello {myVar} World". You obviously can't just look for the next '{' after an 'If'. Using () would help this, but isn't totally neccessary, because strings have to parsed seperatly anyway.
The () around if-conditions is extremely simple. The semicolon is not.
That's why I bought a particular book that teaches you how ;)Quote:
Originally Posted by eyeRmonkey
Simple to type or parse? I assume you mean parse. Are you reffering to them being complicated to parse if we make them optional or just in general?
No eyeR. What I'm saying is that it won't be difficult to do both at all.
Well since I doubt the book uses ParGen (flex/lexx/alexx & yacc/bison/ayacc), then that doesn't help us a lot.Quote:
Originally Posted by Jacob Roman
No. The book teaches you how to program your own parser, scanner, and compiler. ;)
Okay. :)Quote:
Originally Posted by Jacob Roman
Jacob, if you don't mind me asking, what book is it that you bought? And also, what tools will you be using? Or will you write your own lexical scanner, parser and compiler yourself?
The book is called Writing Compilers and Interpreters. It uses C++. I did plan on writing the compiler myself, but since I'm computerless and all since my computer died on me, I have to wait on the new one that's being built at a very slow rate :rolleyes:
I also planned on outputting SSE assembly rather than the typical x86 assembly, so there can be a major speed boost. Something your tools I don't think can do.
If we TRUELY want to output assembly (which at this point is more work than its worth) then we can use C as our front end and ASM as out backend. In fact, this might work prefectly. Girvo could write the parsing and stuff to get it to output perfect C. Then Jake could parse the C into ASM.
What do you guys think about that? That works because we can still continue development of the project until Jake gets up and running. We can compile the C code just fine until Jake gets to the point of a working ASM parser.
Jake. I hope you don't plan on making the compiler anymore. When I say compiler I mean anything that outputs .obj/.exe files. We are going to use a GNU compiler or assembler and we will just be making the parsing stuff. There is no way we can make one as good as them. MAYBE we could still make the linker, but I don't see much point in that either. Once again, it has been done for us in a much more sophisticated way.
Well in my own spare time (when I get my own computer of course) I can try to make a compiler/linker on my own as a learning experience. Maybe having it output SSE assembly might take out the competition. Who knows.
Yes, I suppose anything is possible. And I agree that it is an awesome learning experience. Plus the stuff you learn will definitely help us in this project.
I personally, don't think having the () or ; optional is a good idea. I think it will just make things sloppy. Some lines without semicolons, some lines without parenthesis around the if statements, and I assume loops and everything else that uses them. It will reduce redability in my opinion. I say pick one or the other.
Or better yet. Have an option in the IDE to have either semicolons or no semicolons, so all of your code stays consistant ;)Quote:
Originally Posted by {yak}
But the () can be optional like VB has it.
That would be alright. What are you going to do about multiline statements? Are they going to change when I have the semicolons 'on'? I think in VB you had to use the _ to tell it to continue onto the next line? But in other languages, you can continue onto the next line, and then when it hit's the semicolon it knows it's the end of the statement. What are going to be the rules with that?Quote:
Or better yet. Have an option in the IDE to have either semicolons or no semicolons, so all of your code stays consistant
I believe the () aren't really optional, they are just used for order of precedence, so when you use them in VB you're just saying do what's ever between these parenthesis first.
I believe in the char/byte argument you may have missed this post.Quote:
Originally Posted by penagate
My point being that a single character, in Unicode, requires two bytes to store, whereas an ANSI character requires only one byte to store. To that end, a char can differ in size depending on whether the application is compiled to Unicode or not, and a byte will always be 8 bits.
This behaviour may or may not be consistent with other languages, but as long as the difference is clearly defined, it shouldn't matter.
Jacob can do whatever he would like. I don't mind either way. If Jacob would like to work on the parser/wahtever by himself, then thats fine by me. I can handle rejection ;)
Anyways, that would give me more time to work on ModBASIC, which I think will turn out to be a good language. I am making a forum that will be about creating compilers/parsers for people from ALL projects will be able to come and talk.
So, if you don't need me, then if you don't mind you can choose to leave the compiler up to Jacob. I just hope he knows how hard it actually is to wrtie a FUNCTIONAL, not to mention good, compiler.
-Girvo911
Basically all I meant was that direct access to member variables not be allowed. To that end, that means Public member variables are replaced with properties, which do not necessarily need explicit get/set methods but they are there nontheless (as implicit methods) so the behaviour of Public member variables is not re-created.Quote:
Originally Posted by eyeRmonkey
In the IntNumber class example it would be like this:
Except in that example I used a cast property (which is something I thought of on the spot :p) I'm not sure how you would do this in another language but essentially that is called when you use the object like a value type e.g.Code:int property Value
get
{
return _Value;
}
set (value)
{
_Value = value;
}
Code:IntNumber myNumber = new IntNumber(5);
myNumber = 6 // calls the "cast set" method
int myValue = myNumber // calls the "cast get" method
Overrides is used with classes that inherit other classes, known as "derived" classes. If you inherit a class, all of its methods/variables (that aren't scoped as Protected) become part of your derived class. Usually you don't want a class that is exactly the same, so you can change the behaviour of inherited methods by "overriding" them with your own definition. In my example my derived class overrode the constructor of the base class.Quote:
Originally Posted by eyeRmonkey
"Base" refers to the inherited class. If you wish to refer to a variable/method that you inherited, although it doesn't appear in the code of the derived class you can refer to it by prepending "base.".
I was gonna do it for the sake of a learning experience and experiment a little, not use it in the main project that we are putting together, unless it ends up real good. So you still got the compiler job. Bear in mind, I have no computer to even program on yet :(Quote:
Originally Posted by Girvo911
I was just saying that, because I have only limited experince crafting compilers, and what you guys seem to want to be able to do for the FIRST prject is quite difficult.
I have no money to invest in programming books, (although I found quite a good price on Compilers: Desing, tools, and implemteation, which is a MUST read for anyone even CONSIDERING writing compilers) and i learn all my stuff through trial and error.
Anyways, I am not sure whether I can be of much use to everyone here, even though I have made languages before. Really, I think you should find someone more suited to the task. Don't forget, I'm 14.
-Girvo
Jacob, if I may ask. Are ants building your new PC?
chem
Simple to parse.Quote:
Originally Posted by eyeRmonkey
And making the semicolons optional is extremely complicated, because for each line that doesn't end with one, you have to guess whether the next line continues the statement or starts a new one, something that is not always clear and can lead to very, very subtle bugs.
"His" tools are for reading source code into an abstract in-memory representation and have nothing at all to do with the compiler back-end, which is in charge of outputting the next code form.Quote:
Originally Posted by Jacob Roman
And I believe you're strongly mistaken about the effectiveness of your assembly. SSE instructions aren't faster than normal floating point instructions. They merely can operate on more than one data set at once, which is great in some circumstances (matrix multiplication would be a good example), but will not help at all in most cases. I haven't done any SSE programming myself, so I don't know the next thing for sure, but the fact that these instructions are more complex could mean you actually lose time setting them up.