OpenEuphoria

1. SFML2 Demo Dilema - Solved

• Posted by Icy_Viking May 08, 2015
• 2716 views
• Last edited May 10, 2015

Hello,

I am having a bit of dilema trying to make a demo after wrapping SFML2. It involves trying to make a render window. Here is my euphoria code.

This is the C code I am trying wrap into Euphoria code.

sfVideoMode mode = {800,600,32}; 
sfRenderWindow* win; 
 
win = sfRenderWindow_create(mode,"SFML Win", sfResize | sfClose, NULL); 
 
if(!window) 
 return EXIT_FAILURE 
 
while (sfRenderWindow_isOpen(window)) 
{ 
  while(sfRenderWindow_pollEvent(window,&event)) 
  { 
   if(event.type == sfEvtClosed) 
   sfRenderWindow_close(window); 
  } 
 
  sfRenderWindow_clear(win,sfBlack); 
 
  sfRenderWindow_display(window); 
} 
 
sfRenderWindow_destroy(window); 
 
return EXIT_SUCCESS; 

--Flags 
public atom	sfNone = 0 
public atom	sfTitlebar = 1 
public atom     sfResize = 2 
public atom	sfClose = 4 
public atom	sfFullscreen = 8 
public atom	sfDefaultStyle = or_all({sfTitlebar,sfResize,sfClose}) 
 
public sequence sfWindowStyle = {sfNone,sfTitlebar,sfResize,sfClose,sfFullscreen,sfDefaultStyle} 
 
public enum type sfColor 
 
	sfBlack, 
	sfWhite, 
	sfRed, 
	sfGreen, 
	sfBlue, 
	sfYellow, 
	sfMagenta, 
	sfCyan, 
	sfTransparent 
	 
end type 
 
public enum type sfEventType 
 
	sfEvtClosed, 
	sfEvtResized, 
	sfEvtLostFocus, 
	sfEvtGainedFocus, 
	sfEvtTextEntered, 
	sfEvtKeyPressed, 
	sfEvtKeyReleased, 
	sfEvtMouseWheelMoved, 
	sfEvtMouseButtonPressed, 
	sfEvtMouseButtonReleased, 
	sfEvtMouseMoved, 
	sfEvtMouseEntered, 
	sfEvtMouseLeft, 
	sfEvtJoystickButtonPressed, 
	sfEvtJoystickButtonReleased, 
	sfEvtJoystickMoved, 
	sfEvtJoystickConnected, 
	sfEvtJoystickDisconnected, 
	sfEvtTouchBegan, 
	sfEvtTouchMoved, 
	sfEvtTouchEnded, 
	sfEvtSensorChanged, 
	sfEvtCount 
	 
end type 

--Graphics wrapper code 
public constant xsfRenderWindow_create = define_c_func(gfx,"sfRenderWindow_create",{C_POINTER,C_POINTER,C_UINT,C_POINTER},C_POINTER) 
 
public function sfRenderWindow_create(atom mode,sequence title,sfWindowStyle style,atom sett) 
 
 return c_func(xsfRenderWindow_create,{mode,allocate_string(title,1),style,sett}) 
	 
end function 

--Actual demo 
without warning 
 
include std/machine.e 
include EuSys2.ew 
include EuGfx2.ew 
 
include sfFlags.e 
 
sequence s 
 
--how to make the video mode sequence euphoria friendly? 
atom win = sfRenderWindow_create(s = {peeks(800),peeks(600),peeks(32)},"My Window", sfClose,0) 
 
if not win = 0 then 
	puts(1,"Could not create window!\n") 
	abort(0) 
end if 
 
while(sfRenderWindow_isOpen(win)) 
 
 while(sfRenderWindow_pollEvent(win,--what to put here?) 
 --key input 
end while 
 
sfRenderWindow_clear(win,sfBlack) 
 
sfRenderWindow_display(win) 
end while 
 
sfRenderWindow_destroy(win) 

Any help is greatly appericated. I'm mostly stuck with how to put the video mode in euphoria syntax and what to do about the event system using euphoria commands. I can post more code if it is needed.

3. Re: SFML2 Demo Dilema

• Posted by petelomax May 09, 2015
• 2681 views

sfVideoMode mode = {800,600,32}; 
win = sfRenderWindow_create(mode,"SFML Win", sfResize | sfClose, NULL); 

--how to make the video mode sequence euphoria friendly? 
atom win = sfRenderWindow_create(s = {peeks(800),peeks(600),peeks(32)},"My Window", sfClose,0) 

Oh no, definitely not that. Maybe constant mode=allocate(12); poke4(mode,{width,height,bpp}) but more likely you will have to wrap an sfVideoMode routine.

  while(sfRenderWindow_pollEvent(window,&event)) 

 while(sfRenderWindow_pollEvent(win,--what to put here?) 

My guess would be something like constant event = allocate(sizeof(sfEvent)) and use peek4s(event+??) to retrieve event.type, but you may well have to wrap a whole sfEvent class.

HTH, Pete

5. Re: SFML2 Demo Dilema

• Posted by Shian_Lee May 09, 2015
• 2632 views

I'm not a C programmer and I never had time to use it (some people still have to pick tomatoes...), so I've never used a C struct.

But long time ago, when memory was very limited, I had to create my own efficient data type in BASIC. So I've created a very short library with about 5 functions, that accepts a string argument, start_bit argument, and end_bit argument, for example:

data = "xyz"

data string was treated by my library simply as 24 bits (3 bytes * 8 bits).
if start_bit,end_bit arguments were, for example 6,12 then my library returned an integer of 7 bits (bits 6,7,8,9,10,11,12), which actually started at the first byte ('x') and ended at the second byte ('y').

This library was very short and efficient, using only a fast bit manipulation code.

In this way I could store in a string of any length, as many variables as I needed, including single bits, nibbles, integers, floating points (in IEEE format), other strings, actually any type of variable that exists on earth.

Creating such library is very easy, you need only few utility functions to create, modify, retrieve, and even search for specific data within a string (string in this case is treated at the bit level of course).

I don't know how complex is your C struct. But with that library I did magics, in memory and on disk.

7. Re: SFML2 Demo Dilema

• Posted by Shian_Lee May 09, 2015
• 2606 views

It looks too easy.

I don't know what sfEventType means (is it a WORD or what?), but for my old library it's a piece of cake... it's all about how many bits are in each data type.

9. Re: SFML2 Demo Dilema

• Posted by jimcbrown (admin) May 09, 2015
• 2594 views

It's an int. It's just a regular C++ enum. So is sfKeyCode. sfBool is also an int.

11. Re: SFML2 Demo Dilema

• Posted by jimcbrown (admin) May 09, 2015
• 2603 views

  while(sfRenderWindow_pollEvent(window,&event)) 

 while(sfRenderWindow_pollEvent(win,--what to put here?) 

Assuming that this is the sfKeyEvent struct, it'd look like this (4.0.5 compatible code):

-- sfKeyEvent has 6 members, all of them 4-byte ints 
constant event = allocate(4*6) 
-- note that allocate() returns a C pointer, which is the same thing that the &event stuff in C/C++ does. 
... 
 while(sfRenderWindow_pollEvent(win,event) 
... 
integer eventType = peek4s(event) 
integer code = peek4s(event+4) 
integer alt = peek4s(event+8) 
integer control = peek4s(event+12) 
integer shift = peek4s(event+16) 
integer system = peek4s(event+20) 

sfRenderWindow_create() looks more difficult to wrap, as it passes a struct directly (without a pointer). One possibility:

--Graphics wrapper code 
public constant xsfRenderWindow_create = define_c_func(gfx,"sfRenderWindow_create",{C_INT,C_INT,C_INT,C_POINTER,C_UINT,C_POINTER},C_POINTER) 
 
public function sfRenderWindow_create(integer width, integer height, integer bitsPerPixel,sequence title,sfWindowStyle style,atom sett) 
 
 return c_func(xsfRenderWindow_create,{width, height, bitsPerPixel,allocate_string(title,1),style,sett}) 
	 
end function 
 
atom win = sfRenderWindow_create(800,600,32,"My Window", sfClose,0) 
 

If that doesn't work, you might have no choice but to add an extra C wrapper to wrap sfRenderWindow_create() and then wrap the C wrapper in Euphoria.

13. Re: SFML2 Demo Dilema

• Posted by Shian_Lee May 09, 2015
• 2581 views

Well, that's what my old library did: not only storing variables, but also padding and aligning them, as fields in records.

With this kind of library you always deal with full struct, you don't really need to pass or accept individual variables.

15. Re: SFML2 Demo Dilema

• Posted by Shian_Lee May 09, 2015
• 2573 views

constant event = allocate(4 * 6) 
integer eventType = peek4s(event) 
integer code = peek4s(event+4) 
integer alt = peek4s(event+8) 
integer control = peek4s(event+12) 
integer shift = peek4s(event+16) 
integer system = peek4s(event+20) 

That's similar to what I was taking about, but using an intermediate library, with resolution of bits lets you write more generic code, and with maximum flexibility about any kind of data type.

The code above is suitable for one case, but would be better to write intermediate library for all cases (it takes time, but the logic is simple).

17. Re: SFML2 Demo Dilema

• Posted by Shian_Lee May 09, 2015
• 2574 views

I have to go now, so good luck.

Just remember that a regular string can be used as the most sophisticated struct that can be - if you simply treat it as a continuous stream of bits... But you must write the short library to store and extract variables from a string.

It might help if the data types are too complicated, and it gives you general solution for all cases.

19. Re: SFML2 Demo Dilema

• Posted by jimcbrown (admin) May 09, 2015
• 2593 views

You have a bit of code in the wrong place. Try this.

without warning 
 
include std/machine.e 
include EuSys2.ew 
include EuGfx2.ew 
 
include sfFlags.e 
 
atom win = sfRenderWindow_create(800,600,32,"My Window",sfClose,0) 
 
if win = -1 then 
	puts(1,"Could not create render window!\n") 
	abort(0) 
end if 
 
constant event = allocate(4 * 6) 
-- This is wrong. these shouldn't be here. No data is in event yet. 
-- These will be no more than random numbers. 
--integer eventType = peek4s(event) 
--integer code = peek4s(event+4) 
--integer alt = peek4s(event+8) 
--integer control = peek4s(event+12) 
--integer shift = peek4s(event+16) 
--integer system = peek4s(event+20) 
 
while sfRenderWindow_isOpen(win) do 
	 
	while sfRenderWindow_pollEvent(win,event) do 
	-- they belong down here instead 
integer eventType = peek4s(event) 
integer code = peek4s(event+4) 
integer alt = peek4s(event+8) 
integer control = peek4s(event+12) 
integer shift = peek4s(event+16) 
integer system = peek4s(event+20) 
		-- where did sfeventType come from? what is it's value? 
		-- it doesn't seem to be declared anywhere, let alone initialized 
		--if sfeventType = sfEvtClosed then 
		if eventType = sfEvtClosed then 
			sfRenderWindow_close(win) 
		end if 
	end while 
	 
	sfRenderWindow_clear(win,sfBlack) 
	 
	sfRenderWindow_display(win) 
 
end while 
 
sfRenderWindow_destroy(win) 

Note that I had the correct order in my original post:

constant event = allocate(4*6) 
... 
 while(sfRenderWindow_pollEvent(win,event) 
... 
integer eventType = peek4s(event) 
integer code = peek4s(event+4) 
integer alt = peek4s(event+8) 
integer control = peek4s(event+12) 
integer shift = peek4s(event+16) 
integer system = peek4s(event+20) 

21. Re: SFML2 Demo Dilema

• Posted by jimcbrown (admin) May 09, 2015
• 2569 views

Ah, I forgot about this. I think with C, enums start from 0, but they start from 1 with Euphoria by default. Try this:

public enum type sfEventType 
 
	sfEvtClosed = 0, 
	sfEvtResized, 
	sfEvtLostFocus, 
	sfEvtGainedFocus, 
	sfEvtTextEntered, 
	sfEvtKeyPressed, 
	sfEvtKeyReleased, 
	sfEvtMouseWheelMoved, 
	sfEvtMouseButtonPressed, 
	sfEvtMouseButtonReleased, 
	sfEvtMouseMoved, 
	sfEvtMouseEntered, 
	sfEvtMouseLeft, 
	sfEvtJoystickButtonPressed, 
	sfEvtJoystickButtonReleased, 
	sfEvtJoystickMoved, 
	sfEvtJoystickConnected, 
	sfEvtJoystickDisconnected, 
	sfEvtTouchBegan, 
	sfEvtTouchMoved, 
	sfEvtTouchEnded, 
	sfEvtSensorChanged, 
	sfEvtCount 
	 
end type 

23. Re: SFML2 Demo Dilema - Solved

• Posted by ghaberek (admin) May 10, 2015
• 2596 views

Here's the traditional way to wrap structures pre-4.1, which adds (will add?) memstruct support. Some knowledge of C is required to get this right, especially with type sizes and signage. Most structure elements are 4 bytes but some are more tightly packed. Unions and embedded structures can complicate matters. Adding to that, type sizes typically change between 32-bit and 64-bit (but not always), which is why we need memstruct support for 64-bit 4.1.

Basically, we define each offset as a constant, then we don't have to peek/poke with "magic numbers" in our code. This helps prevent bugs imposed by improper offsets in the structure, as well as adding to overall code clarity.

-- sfKeyEvent struct 
public constant 
    sfKeyEvent__event   =  0, -- sfEventType 
    sfKeyEvent__code    =  4, -- sfKeyCode 
    sfKeyEvent__alt     =  8, -- sfBool 
    sfKeyEvent__control = 12, -- sfBool 
    sfKeyEvent__shift   = 16, -- sfBool 
    sfKeyEvent__system  = 20, -- sfBool 
    SIZEOF_SFKEYEVENT   = 24, 
$ 
 
-- example: 
atom event = allocate_data( SIZEOF_SFKEYEVENT, 1 ) 
atom eventType = peek4s( event + sfKeyEvent__event ) 
atom code      = peek4s( event + sfKeyEvent__code ) 
-- etc. 

Here's how you'd do it with memstruct in 4.1.

public memtype 
    int as sfEventType, 
    int as sfKeyCode, 
    int as sfBool, 
$ 
 
public memstruct sfKeyEvent 
    sfEventType event 
    sfKeyCode code 
    sfBool alt 
    sfBool control 
    sfBool shift 
    sfBool system 
end memstruct 
 
-- example: 
atom event = allocate_data( sizeof(sfKeyEvent), 1 ) 
atom eventType = event.sfKeyEvent.event 
atom code      = event.sfKeyEvent.code 
-- etc. 

-Greg

25. Re: SFML2 Demo Dilema - Solved

• Posted by Shian_Lee May 10, 2015
• 2564 views

typedef struct 
{ 
  unsigned int width; 
  unsigned int height; 
  unsigned int bitsPerPixel; 
} sfVideoMode; 

-- struct sfVideoMode 
public constant 
    sfVideoMode__width        =  0, -- unsigned int 
    sfVideoMode__height       =  4, -- unsigned int 
    sfVideoMode__bitsPerPixel =  8, -- unsigned int 
    SIZEOF_SFVIDEOMODE        = 12, 
$ 
 
-- example 
atom mode = allocate_data( SIZEOF_SFVIDEOMODE, 1 ) 
atom width  = peek4u( mode + sfVideoMode__width ) 
atom height = peek4u( mode + sfVideoMode__heigth ) 
-- etc. 

public memstruct sfVideoMode 
    unsigned int width 
    unsigned int height 
    unsigned int bitsPerPixel 
end memstruct 
 
-- example: 
atom mode = allocate_data( sizeof(sfVideoMode) ) 
atom width  = mode.sfVideoMode.width 
atom height = mode.sfVideoMode.height 
-- etc. 

Greg,... I'm not a C programmer but I get the point. It's very sad actually, from my point of view.

The traditional method is Euphoria code, clear, easy, works just fine. 4.1 memstruct is certainly not a breakthrough, yet it's overloading the language with additional C syntax.

Instead, you could just add a short tutorial to the manual for doing it right, and how to avoid crash. Or just add some safer utility to the STD library, written in "traditional" Euphoria.

It may avoid some peeking errors, but come on. Why don't you add string type to the language as well?...

27. Re: SFML2 Demo Dilema - Solved

• Posted by Icy_Viking May 10, 2015
• 2534 views
• Last edited May 11, 2015

Once 4.1 becomes officially available, I will rewrite the wrapper using the memstructs. Right now though, it is written in Euphoria 4.0.5 and everything works so far. I really hope 4.1 becomes an official release soon, I believe the memstructs will help out alot when wrapping C code. Might even help with wrapping C++ code as well. Also, I'll be using the 32-bit versions for compatbility reasons. I also saw that SFML 2.3 came out recently, so once the C binding gets updated for 2.3 I'll need to update EuSFML2 once again.

Edit: Added tilde (~) to "C++" to fix formatting. -Greg

29. Re: SFML2 Demo Dilema - Solved

• Posted by Shian_Lee May 10, 2015
• 2560 views

What can I say... it's almost like coding in C directly, and for C programmers it's great, as well as for Euphoria users who need these wrappers.

But in general, Euphoria 4 is getting more and more stuffy and with too many options, i.e.: not reliable.

31. Re: SFML2 Demo Dilema - Solved

• Posted by ne1uno May 11, 2015
• 2551 views

totally agree. almost any other language is more likely to hang & crash on you or your users. though the silent exit problem can still exist with some as yet untested syntax, the reliability of the interpreter and compiler on debugged working code is unquestioned by all but a few who basically never build their own versions anyway so have only tried the earlier versions and even those not that much. it's impossible to make some people happy.

I would mention that it is 2015 and all the heavy lifting has been over for several years so fighting a "save developers time" angle is not really productive. people mostly work on what interests them. neither is is instructive to belabor how plain sequences and constants are just as effective as memstructs when you have never tried to fix win32lib or any large wrap and generally have no use for library written in other languages. we love syntactic sugar for a reason.

one other point: memstructs are in the menstruct branch and so far, the planned release of 4.1 won't include them nor do any of the alpha or beta. I have used memstruct quite a bit and never use any other euphoria except to build or test something.

33. Re: SFML2 Demo Dilema - Solved

• Posted by ghaberek (admin) May 13, 2015
• 2404 views

The memstruct feature opens the door to cross-architecture programming (i.e. 32-bit or 64-bit). The traditional methods of structure wrapping I posted above really fall apart in the 64-bit world, because almost everything shifts up from 4 to 8 bytes wide (but not everything!). Having to manage a litany of ifdef's to provide this compatibility without memstruct is going to be tedious, at best.

Well, memstruct certainly should be included in the 4.1 release. I've used it quite a lot and I know Matt's fixed a lot of outstanding issues. What else remains?

-Greg

35. Re: SFML2 Demo Dilema - Solved

• Posted by ghaberek (admin) May 13, 2015
• 2381 views

I vote that we vote.

-Greg

37. Re: SFML2 Demo Dilema - Solved

• Posted by jmduro May 13, 2015
• 2435 views

That is the reason why I'm switching to wxEuphoria. It's not that different from win32lib. Once I understood how it worked and corrected some documentation errors, porting was relatively easy comparing to EuGtK. Now I have portable code running on both Windows and Linux.

To get it work, I had to use binaries included in wxide because official versions don't work with it and I didn't find the problem (parameters on an event routine).

Regards

Jean-Marc

39. Re: SFML2 Demo Dilema - Solved

• Posted by Spock May 14, 2015
• 2454 views

Test Results using latest beta:

On my machine (Win7) Stability good, Startup slow
On a User machine (XP) program don't even start (or maybe it was starting but I shut it down after 5 mins)

Oh well, back to Dec 2010..

Yeah. You Devs are doing the right thing delaying the next release..

Spock

41. Re: SFML2 Demo Dilema - Solved

• Posted by irv May 14, 2015
• 2364 views

This may, or may not, be something you can use, but I'll throw it out there for discussion:

SFML looks to be about as complex as GTK3, with lots of stuff to wrap. When I started working on EuGTK, I tried to link all the functions at startup. Not a good idea. Took forever to start up. Since GTK is object-oriented (as is SFML), I thought 'there's no use in linking functions that my program won't ever use', and so I link only the functions (methods) required for a given object that is declared in a program.

IOW, if your program creates a window, then the window methods (and inherited methods) are linked. If your program uses a button - or 100 buttons - the button class methods are linked (once). Not many programs are going to use more than a couple dozen object classes.

Perhaps - or not - the same approach could be used with SFML?

43. Re: SFML2 Demo Dilema - Solved

• Posted by ghaberek (admin) May 14, 2015
• 2354 views

Win32Lib already uses delayed-loading for system functions. See registerw32Function() and w32Func().

-Greg

45. Re: SFML2 Demo Dilema - Solved

• Posted by Spock May 15, 2015
• 2303 views
• Last edited May 17, 2015

ATM I distribute a consolidated source file (41,609 net LOC) over our office network to update the main program. Compiling & distributing an exe doesn't really appeal to me for various reasons but no doubt it will be faster. I'll do some more tests. For our XP users I'd rather leave the Status Quo as is for them since 4.1 beta seems to be a non-starter anyway. (Our IT guy wants to upgrade all XP boxes but change doesn't happen fast for us due to time/budget/personalities..)

Spock

47. Re: SFML2 Demo Dilema

• Posted by Icy_Viking May 15, 2015
• 2275 views

I've updated the wrappers to make them more Euphoria friendly. I've also updated and fixed the demos. The update should be on the site within a few hours or so.