Quote:

1. macros have many, many uses.

Quote:

Macro EQ == Macro NEQ !=

Quote:

Macro Set =

Quote:

Macro isint(X) (cint(x) eq x) Macro issquare(x) isint((z set sqr(y)))

integer r = 5;if (r.isint()) ...if (r.issquare()) ...

Quote:

Macro msgbox(x) eventsend:Msgbox(cstring(x))

gui.msgbox()

Quote:

Also try to make a simple vm first, its less then an hours work to make a real good one, and it'll help you later. (when your trying to code for something which doesn't exist, it is Not Nice.)

Quote:

Most of it would happen at compile time. Sometimes it would happen at run-time (but rarely would you need to). Doing so is quite easy, Get code for function, Allocate new memory (for each var in function) Copy function to new spot on the object array Change vars in function. All in all quite a fast job.

Quote:

Also, What i try and do, is keep a smallish array (or vector, in c++), which contains the jump points. These are points where your code should go to. Something like jmp GUID_OF_FUNCTION, GUID_OF_OBJECT, Line number -1 (or NULL) should be used to signify My. (current object). If your not using guids, then shame on you!

Object   0 init()   1 istype(class type)Integer : Object   0 init() (inherited from object - can be overridden)   1 istype() (inherited from object - can be overridden, not a goot idea to though)       - new methods start here -   2 issquare()   3 multiply(integer i)Decimal : Object   0 init()   1 istype()       - new methods start here -   2 convert_to_integer()   3 multiply(decimal d)NewInteger : Integer : Object   0 init() (inherited from object - can be overridden)   1 istype() (inherited from object - can be overridden, not a goot idea to though)       - integer methods start here -   2 issquare()   3 multiply(integer i)       - newinteger methods start here -   4 do_wacky(string s)   

Quote:

With lists, Make it a data type, and make the list, a list of type datatype. That would allow you to have lists of lists of lists of lists, ect. maybe have a few methods in list, like concatinate, sort, split-at, ect. to make things easier.

integer i[] = [1, 2, 3]

integer[] i = [1, 2, 3]

Quote:

Have namespaces, and allow them to be changed at runtime (which would make things a lot easier).

Quote:

Function pointers! (you do not know how good these are, until your using a language which doesn't support them!)

integer i = 5method(multiply(integer)) m = method(i.multiply())i = i.call(m(i))

Quote:

Also, it would be nice to have a flexable grammer and/or reprogrammable grammer. (currently my parser, (with a bit of extra taggery) will accept C style for loops, basic style for loops, and with a bit of persuading, probably a few more types too!) Maybe something like a grammer define tag? GRAMMER ; NEWLINE

Quote:

We would have to sort some things out, but it sounds like a good idea. As I think I have said, I've done a bit of work on the parsing side of things, and it sounds like you have done a bit of work on the other side of things (xslt is greek to me) so we could work together.

Quote:

I also have to implement a few things, but they are mostly tied to the virtual machine (lists / maps / ranges).

Quote:

Original post by cmp that's perfekt, because i'am not interested in making a parser ;) currently i need someone who translates sourcecode to the xml syntax, wich itselft isn't that hard. the xml syntax itself is quite finished, but i have to figure some things out and things may change (for example: i'm still thinking about if i should use "name" or "select" for an attribute name).

Quote:

Quote: I also have to implement a few things, but they are mostly tied to the virtual machine (lists / maps / ranges). i would choose a different approach, in my understanding of a 'good' language a language itself should be as simple and clean as possible, so array, lists, maps, etc are all implemented in the language not in the runtime system.

Quote:

as for your integer array exmaple in my planned language you would write: i : Kernel.Array<Integer> = [[1, 2, 3]] where [[ e1, e2, ..., en ]] defines an array with the elements e1 to en.. similar to [[ ]] is << >> wich defines a tuple: <<e1, e2 .. en>>. ...

Quote:

i choose to use name : type as the definition of a variable becuase, this way the modifiers of of a method's return-type cannot conflict with the modifier of the method itself. for example: // a const method does not change it's class' contents const get(index : Integer) : const String { return array_object @ index }

Quote:

you may argue why not use the c++ syntax const String get(Integer index) const? but in c++ the last const is an method modifier and normally method mods like virtual, extern, etc are written in front of the method. and in my language there are much more modifiers planned than in c++, so this keeps the language logical.

Quote:

another interesting aspect of my langauge is, that every class inherits Kernel.Any so any class has method likle same_type(), equal, depp_equal, copy, deep_equal ... as you may have noticed many features of 'my' language are borrowed from eiffel, wich itself is a very clean language.

Quote:

from what i have got our projects are very similar, but you target only scripting, while i try to make a scripting and native language. as i said i'am currently writing an xml to c++ 'compiler' wich itself should be written in my little xml language, so it's portable (in theory you could even use your interpreter to run the compiler, if you would support loading xml source code).

Quote:

Original post by umbrae Quote: 1. macros have many, many uses. Yeah they do. But I haven't really had many situations that I felt needed macros, I believe that they subtract from readability - it is harder to understand the code. From the examples that you have given, I think each one would be better implemented in the language. Quote: Macro EQ == Macro NEQ != There is no good reason to ever do an assignment in a conditional expression within a scripting language. C++ maybe for optimisation purposes, but scripting language - no. So why don't we just use "="? It makes more sense than two capital letters. Quote: Macro Set = Why? There is already a perfectly good operator that means assignment. I would even choose Pascal's ":=" over a set keyword. Quote: Macro isint(X) (cint(x) eq x) Macro issquare(x) isint((z set sqr(y))) integer r = 5;if (r.isint()) ...if (r.issquare()) ... Quote: Macro msgbox(x) eventsend:Msgbox(cstring(x)) gui.msgbox() (takes strings, overloaded to take integers, and other datatypes also user overrideable) Macros Compared to Functions Generic Programming Using Cee Macros If you have more arguments for the inclusion of macros I would like to hear them. Quote: Also try to make a simple vm first, its less then an hours work to make a real good one, and it'll help you later. (when your trying to code for something which doesn't exist, it is Not Nice.) Good idea, I have been playing around with parsing for quite some time now. Would be good to have something that works, even if I have to hand code the bytecode. Quote: Most of it would happen at compile time. Sometimes it would happen at run-time (but rarely would you need to). Doing so is quite easy, Get code for function, Allocate new memory (for each var in function) Copy function to new spot on the object array Change vars in function. All in all quite a fast job. I've always been wary when thinking about reallocating whole arrays. It doesn't sound good in my book. I like static. I like the idea that once some classes have been definied that they aren't going to change. Quote: Also, What i try and do, is keep a smallish array (or vector, in c++), which contains the jump points. These are points where your code should go to. Something like jmp GUID_OF_FUNCTION, GUID_OF_OBJECT, Line number -1 (or NULL) should be used to signify My. (current object). If your not using guids, then shame on you! Shame on me. I have some idea of what a guid is, and I think I am going to use something similar but I would like some explanation if possible. Globally Unique Identifier? Each of my methods has a unique offset based on the distance it is from the first method in object. Each sub-class of object increases this offset with the methods it adds. Object 0 init() 1 istype(class type)Integer : Object 0 init() (inherited from object - can be overridden) 1 istype() (inherited from object - can be overridden, not a goot idea to though) - new methods start here - 2 issquare() 3 multiply(integer i)Decimal : Object 0 init() 1 istype() - new methods start here - 2 convert_to_integer() 3 multiply(decimal d)NewInteger : Integer : Object 0 init() (inherited from object - can be overridden) 1 istype() (inherited from object - can be overridden, not a goot idea to though) - integer methods start here - 2 issquare() 3 multiply(integer i) - newinteger methods start here - 4 do_wacky(string s) Also how do you keep these guid's over multiple compile sessions? What I want to do is have the ability to have multiple bytecode files and include them all in the virtual machine. Quote: With lists, Make it a data type, and make the list, a list of type datatype. That would allow you to have lists of lists of lists of lists, ect. maybe have a few methods in list, like concatinate, sort, split-at, ect. to make things easier. I am thinking about using the same syntax that Groovy uses. integer i[] = [1, 2, 3] or integer[] i = [1, 2, 3] Quote: Have namespaces, and allow them to be changed at runtime (which would make things a lot easier). The package system is a sort of namespace convention. I personally don't think namespace changing is good - it means that what you are doing probably isn't object orientated. Quote: Function pointers! (you do not know how good these are, until your using a language which doesn't support them!) Methods are just another object, like classes and objects. They all can be stored in a "type" container. There will be a way to do something like: integer i = 5method(multiply(integer)) m = method(i.multiply())i = i.call(m(i)) Or something similar. Need to work out calling conventions (ie what it returns, the arguments etc). Quote: Also, it would be nice to have a flexable grammer and/or reprogrammable grammer. (currently my parser, (with a bit of extra taggery) will accept C style for loops, basic style for loops, and with a bit of persuading, probably a few more types too!) Maybe something like a grammer define tag? GRAMMER ; NEWLINE I like the simplistic approach. Cluttering up a language with extra syntax seems too much. If there are multiple ways to do something then people will use these different ways. I like to be able to look at some code and be able to work out what it does easily. This is a similar argument to the one against macros. I'm a bit of an advocate for Self Documenting Code.

Quote:

1. i see you do not lilke macros... I like them, i use them, when named after what they do Its quite readable (although if not, i start agreeing with you on this point). Please note that the egsamples are actually quite comical, and were not good egsamples on how macros should be used...

#define null NULL

Quote:

Simple guids ...

Quote:

(the offset idea is good, but is still lacking, for egsample: what would be the offsett for: Where o0-9 are objects o8.o7.o6.o5.o4.o3.o2.o1.add o9 ??

load $1, o8     // not the 'proper' way to load a variable - just to show that $1 is o8obj $1, $1, 7   // offset by 7 to get the 'pointer' to the next object then follow that pointerobj $1, $1, 6obj $1, $1, 5obj $1, $1, 4obj $1, $1, 3obj $1, $1, 2obj $1, $1, 1   // now $1 = o1load $2, o9     // again this is actually differentpush $2         // o9 is now on the stackcall $1, 12     // calls the add method which is offset 12 from the object o1pop $2          // 2 now contains the answer

Quote:

Nice integer syntax, and a nice idea also. I would prefer integer i = [1,2,3,4,5] tho. Also i = [x,y,z,2,3,4] should also be acceptable (and handy)

integer[] i = [1..5, 8, 9]integer[] j = [11, 12, 16..20, i]

Quote:

With the grammer tags, i can see your point. Its just that it would allow for more expansion of the language (see lisp, and how easily it allow you to do things, simply be changine the grammer).

remote class HelloWorld {  void put(){     std_io.put("hello world!")  }; }class Client{  void run(){     server : remote HelloWorld = get_root_by_ip("127.0.0.1")     server.put()  }}

• sperate - for threads, every call on an seperate object will execute in another thread and the calling thread won't wait, until the result is used


• remote - for remote objects distributed in a network


• extern( language/plugin ) - automatic wrapper generation, but the source has to be written in my langauge (so you could compile the game object to lua or your bytecode and use it directly in your native application without the need to write any wrapper object by hand)


• const - for the const'ness of objects. it's a very good feature of c++ in my eyes.


• reference - for defining pointers to pointers (else you would have to write wrapper objects, wich sucks)

Quote:

Original post by cmp concerning your dislike of keywords: to me you're right to some extend, there should be as little keywords as possible, but for some very high level constructs you need keywords. another way of achiving so high abstraction would be to use external files describing these constructs, but this dosen't seem to be the right way in my eyes.

class a : b : c : d  // name is a, super is b, implemented interface is c, viewable package is d    - void method()        - const        - remote                integer i = 5;        if ( ...

Quote:

as an example: remote procedure call, i plan to use the keyword "remote". then you could just write: *** Source Snippet Removed ***

// a normal integer createdinteger i = integer.alloc().init(5)// orinteger i = 5// a networked integer 'subscribed'integer i = integer.subscribe(connection, "new integer")// a network integer 'published'integer i = 5;connection.publish(i, "new integer")

Quote:

all togehter i will try to support the following keywords: sperate - for threads, every call on an seperate object will execute in another thread and the calling thread won't wait, until the result is used remote - for remote objects distributed in a network extern( language/plugin ) - automatic wrapper generation, but the source has to be written in my langauge (so you could compile the game object to lua or your bytecode and use it directly in your native application without the need to write any wrapper object by hand) const - for the const'ness of objects. it's a very good feature of c++ in my eyes. reference - for defining pointers to pointers (else you would have to write wrapper objects, wich sucks)

integer i = integer.async().init() // separate threadinteger i = integer.subscribe(connection, "new integer") // remote connection

Quote:

every other keyword as inline or virtual should be guessable by the compiler. but in my current stage i still need the virtual keyword, as the xslt language isn't powerfull enough to determine these things. another thing, where you would need virtual, is the compilation of code into a library, because sometimes you have to guarantee that a method will be virtual.

Quote:

concerning your array-syntax: i think it's very wired that keywords bloat a language from your point of view, but an extra syntax for defining arrays does it not.

integer[] i = [1, 2, 3]

integer[] i = array.alloc(integer, 1, 2, 3)

Quote:

btw: i try to upload some code in the weekend, but by now i'm very consumed with real life (school, etc)

Quote:

Original post by umbrae With your example code - does the keyword mean that the class has to be over the network, or does it mean that it can be?

Quote:

And is it important to use the keyword when you declare a variable (server : remote HelloWorld)? Can't one of these keywords at least be taken out?

Quote:

If you take out the one in the declaration it means that you don't have to know that you are talking to a remote object - it is handled transparently.

Quote:

This way is would be possible to have different types of connections also one connection can have multiple objects communicating through it - and this can be defined in code, and changed. In this situation networking is done through objects not language.

Quote:

Of course the connection has to have been setup beforehand. Not sure if I should use a string to identify the published object or something else (or not identify it?).

Quote:

integer i = integer.async().init() // separate threadinteger i = integer.subscribe(connection, "new integer") // remote connection

Quote:

As for const - I haven't used it much, and when I did I had no errors due to changing something that was const. My idea is that they are for compile time only, and I have never needed them - I have never been saved from an error because of them.

Quote:

I personally don't use pointers to pointers much at all. I have programmed in Java for a while, I don't think you can have them in Java and I got along fine. They are only really useful when you are doing something nasty in C or C++, and even then there are Design Patterns that help you out of those holes.

Quote:

It sounds like your language is restricted because of the language you are turning it into. [...] If you made every method virtual, and always defined a method - even if it just calls it's parent's method (can you do this in C++?) I think it might work.

Quote:

I would consider that this: integer[] i = [1, 2, 3] is less bloated than: integer[] i = array.alloc(integer, 1, 2, 3)

Quote:

Yeah, I have an exam tomorrow, I've had 3 in the last week - real life is annoying sometimes.

Quote:

cmp, do you have a bnf of your language? If you have, or can create one then I will be able to write a parser for it. Is flex and bison a good option? - it's what I have used.

file:		definition*definition	: [mod_list] 'class' [ '<' ident_list '>'] [ ':' object_list ] '{' [member*] '}'		| 'namespace' identifier '{' [definition*] '}'object_list : object_list, object            | objectident_list : ident_list , identifier            | identifiermember		: attribute | methodattribute	: var_def [assignment]method		: [mod*] identifier ':' [type] ['(' parameter_list ')'] method_bodyvar_def		: identifier ':' typemethod_body	: '{' [statements] '}' 		| 'deferred'			type		: [mod*] ex_typeex_type		: ex_type '.' identifier | identifierstatement	: identifier assignment		| var_def [assignment]		| epxression		| 'if' '(' expression ')' '{' [statement*] '}' [elseif*] [else] 		| 'while' '(' expression ')' '{' [statement*] '}'		| 'throw' expression		| 'try' '{' [statement*] '}' [catch*]elseif		: 'elseif' '(' expression ')' '{' [statement*] '}'else		: 'else' '{' [statement*] '}'catch		: 'catch' '(' var_def ')' '{' [statement*] '}'expression	: expression infix expression		| prefix expression		| '(' expression ')'		| literal		| access [ '(' [argument_list] ')']		| 'new' type '(' [argument_list] ')'		| 'lamda' expression		| objectargument_list: argument_list ',' expression             | expressionliteral		: string | integer | real | boolean | character | array_literal | tuple_literalarray_literal	: '[[' exp_list ']]'tuple_literal	: '<<' exp_list '>>'infix		: '+' | '-' | '*' | '/' | '%' | '|' | '&' | '||' | '&&' | '@' | '==' | '!=' | '<=' | '>='prefix		: '^' | '!'mod		: 'extern' '(' string ')' 			// example: extern("lua") automaticly creates wrapper code		| 'extern'		| 'public' | 'protected' | 'private' 		| 'virtual' 		| 'const' 		| 'remote'		| 'seperate'object		: identifier | "( '?' | '?'( [1-9] | [1-9][0-9]* ))"access		: access '.' object | objectexp_list	: exp_list ',' expression | expressionparameter_list	: parameter_list ',' var_def | var_defassignment	: '=' expression

Quote:

Original post by cmp Quote: If you take out the one in the declaration it means that you don't have to know that you are talking to a remote object - it is handled transparently. this is one of the ideas, that the compiler knows the object is not located on this machine, but the programmer does not have to worry about it.

Quote:

@your proxy object idea: i think this is the only way to deal with this problem, i would just subclass the class wich should be located in the network and now simply through polymorphism the user of an object can't distinguish between a remote and a local object.

Quote:

Quote: This way is would be possible to have different types of connections also one connection can have multiple objects communicating through it - and this can be defined in code, and changed. In this situation networking is done through objects not language. yeah, but i think this way you have to write the proxyobjects everytime for every object. my idea was to let the compiler do this unneccasrry rask and this way the problem is not a runtime, but compile time problem. another aspect, wich is imporant, if the compiler knows that a problem may be remote, it won't let the main thread of execution wait for the method called or use similar techniques.

Quote:

Quote: Of course the connection has to have been setup beforehand. Not sure if I should use a string to identify the published object or something else (or not identify it?). i had the idea to just make the root object of an application visible via the network, but as each object may have subobjects you could reach every object through one, wich itself could be specified by the ip address or something else (of course the compiler would have to have at least the source code for the server's superclass when compiling the client.). Quote: integer i = integer.async().init() // separate threadinteger i = integer.subscribe(connection, "new integer") // remote connection i really don't understand, if asyinc is now a method or not, and if so init should be also one, but how do you actually allocate an object (or are async and init compile time generated) and if async is a method what is it returning wich itself could generate an object's instance? from what i get it seems that you are not removing keywords from the language, but just bluring the difference between a keyword and a method.

unit_group g = goblins.spawn(world).init()// normal callg.moveto(waypoint)// async callg.moveto.async(waypoint)

Quote:

Quote: As for const - I haven't used it much, and when I did I had no errors due to changing something that was const. My idea is that they are for compile time only, and I have never needed them - I have never been saved from an error because of them. you always have to remember that your source code may be reused by some else, and now it does matter, wheter an object/mehtod is const/private/public. as you said you won't do any mistake, but someone else could break your whole application when writing a plugin, just because he has changed something, wich shouldn't have been changed.

- void do_stuff_const (integer.const amount, string.const bob)

• a super object (has the name of the class - one instance ever)


• the object (a normal instance)


• an hooked object (a special instance)

Quote:

yes, i have to rethink it, because if you may have a pointer to pointer, why not a pointer to pointer, wich points to a pointer (wow so many points).

Quote:

Quote: It sounds like your language is restricted because of the language you are turning it into. [...] If you made every method virtual, and always defined a method - even if it just calls it's parent's method (can you do this in C++?) I think it might work. this is not the problem, but i want a language wich is fast and making a function virtual alway implies some runtime overhead (memory and speed), wich isn't the way i want to restrict the user's of my langauge.

Quote:

Quote: I would consider that this: integer[] i = [1, 2, 3] is less bloated than: integer[] i = array.alloc(integer, 1, 2, 3) no the [1,2,3] is not my problem, this is something i have planned myself. the problem is the integer[] wich i would have been rewritten as Array<Integer>, where Array is an object implemented in my language itslef (atcually i'm currently writing this object, but my stylesheet lack template support at this time, so everything brakes.)

Quote:

Quote: Yeah, I have an exam tomorrow, I've had 3 in the last week - real life is annoying sometimes. normally exams (btw i was really searching for this word when writing my last reply :D) aren't that hard, but maths and physics are special (linear algebra and einstein's theory of relativity).

Quote:

Quote: cmp, do you have a bnf of your language? If you have, or can create one then I will be able to write a parser for it. Is flex and bison a good option? - it's what I have used. actually i had one parser written in c++ (with flex and bison), but then i realised that i would be really cool to write the parser in my language, so currently i will write everthing as xml, so i can recompile it later to c++ and my langauge.

Quote:

currently this my bnf (i don't know if this is valid bnf, somthing enclosed by [] means it is optional something with a following * means it can be placed as many times as you want) *** Source Snippet Removed ***

tag:    other_tag    | tag other_tag

tag:    /* nothing */    | tag other_tag