Using GEM from the C Language

Graphics and Random Numbers in the third part of Steve Pedler's series

 

Issue 30

Nov/Dec 87

Next Article >>

<< Prev Article

 

 


 

This is the third part of this series which looks at the use of the ST's GEM interface from C. In it we shall take a third (and for the moment final) look at the graphics functions available from GEM and add another demonstration to the program we have been developing.

The accompanying listing is for the fourth demo. You will see that there are two additional global declarations for functions returning WORDs and a few lines added to main() to allow the new demo to be called. Once you have got this typed in and successfully compiled, pressing any key (other than the space bar) at the end of the third demo will call the function demo4(). To exit this demo, press either mouse button to rerun the whole program, or both buttons together to exit back to the desktop.

DEMONSTRATION 4

The purpose of this demonstration is to draw a series of 'wireframe' globes having a 3D effect. To make things more interesting the globes will be of different sizes scattered randomly over the screen. It starts off with the declaration and initialisation of some auto variables. The variables color and max_color limit the colors used in drawing the globes, so that only the color indices 2 to 7 inclusive are used. This is solely out of personal preference, and you can use the full range of colours (1 to 15) if you wish. Variables max_rad, high_x and high_y impose limits on the size and position of the globes and again can be changed if you wish.

The next five lines should be familiar by now as we have used these functions before. The function vsLcolor() is one of the attribute functions, and sets line (as opposed to text or fill) colour. It has the same parameters as the related functions vst_color() or vsf_color(). The next group of lines use a function we have not met so far. The problem with this demonstration is that we don't know exactly where a globe is going to be drawn, since we are using random numbers to determine size and position. The result is that a globe drawn near (for example) the left or right edge of the screen might 'spill over' the edge. In this case part of the globe will be drawn on the opposite side of the screen, which doesn't look very impressive. Alternatively a globe at the bottom of the screen would overwrite the title of the demo. We could get round this by ensuring that the positioning of the globes was such that they could never approach the edge, but then they would tend to all be concentrated in the centre of the screen. What is needed then is a method of restricting – or clipping –graphic output to a defined area of the screen. It won't come as a surprise to find that GEM provides exactly that facility.

The function vs_clip() sets up a graphic clipping rectangle. It is called with the following parameters:

vs_clip( device handle, clipping flag, rectangle coordinates array )

The coordinates array is an array of four sixteen-bit words. The first two words contain the x and y coordinates of the upper left corner of the clipping rectangle, while the second two words contain the coordinates of the diagonally opposite corner. When the function is called, it will restrict graphics output to within this defined rectangle. The clipping flag simply indicates if clipping is enabled or disabled. If it has the value 1, clipping is enabled; if 0, clipping is disabled. Note that on first opening a virtual workstation clipping is disabled. The clipping rectangle used in this demo uses the whole screen except for 10 pixels at the bottom, which is where the title is printed.

We now enter the main loop of the function. Each pass through the while loop draws another globe, then checks to see if the user has pressed one of the mouse buttons. If a button has been pressed, the while loop is terminated and the function waits in another loop until the button is released again. It then disables the clipping rectangle in case it interferes with the output of one of the other demos and finally returns the number of the button(s) pressed to main().

THE MAIN LOOP OF THE DEMO

This is straightforward. Each pass through the loop, the size of the globe and its position on the screen is determined by three calls to calc_num(), discussed further below. The fill colour is then set to zero (background colour) and the function v_ellipse() is called, which draws a filled ellipse on the screen. Since the fill colour is the background colour, the result is to clear an area of the screen where the next globe is going to be drawn. If this isn't done, the new globe overwrites the old one which tends to look a bit of a mess. In fact, if I remember my geometry correctly, what v_ellipse() draws is not an ellipse (a shape which has two centres, not one as a circle does) but an oval, which can be thought of as a circle which has different radii in the x and y planes. Be that as it may, the function is called with these parameters:

v_ellipse( device handle, centre x coordinate, centre y coordinate, x radius, y radius )

As indicated above, this function uses fill attributes. Clearly, if the x and y radii are the same you get a circle — or do you? The answer is that you do in low resolution (and probably also in high resolution, though I haven't been able to try it) but definitely not in medium resolution. You may have been wondering why we aren't using v_circle() for this purpose if we want to draw a circle. I originally wrote this demo as a stand alone version to run in any resolution, and I found that in medium resolution v_circle() didn't clear the correct area for the new globe. This is due to the shape of the pixels which are more or less square in low and high resolutions but are significantly taller than they are wide in medium resolution. If you use v_ellipse() with identical x and y radii in this mode you get a pronounced oval, not a circle. The people who implemented GEM for the ST must have realised this problem with medium resolution, and ensured that v_circle() would draw a true circle, not an oval. Consequently, v_circle() does not clear a sufficient screen area in this mode and v_ellipse() must be used instead. Although this demo is intended for use in low resolution, I have left it as v_ellipse() in case you want to try converting the demo to run in other resolutions.

Having cleared the way for the new globe, the line colour is set to the current value in the variable color and temp_radius is initialised. The function v_ellarc() is then called to draw the outline of the new globe. This function is analogous to v_ellipse() except that it uses line rather than fill attributes, and is supplied with two extra parameters to enable part of an ellipse (i.e. an elliptical arc) to be drawn if desired. These two parameters (starta and stopa in this program) are the beginning and end angles for the arc, and are expressed in tenths of a degree. By setting them to 0 and 3600 respectively, we get a complete circle (this is similar to v_pieslice() which we used in the first demo in this series). There is of course a similar function which uses fill rather than line attributes; this is v_ellpie() which has exactly the same parameters as v_ellarc().

We now enter a second while loop in which the interior lines of the globe are drawn. Function v_ellarc() is called twice each pass through the loop to draw a series of ovals with steadily decreasing radii in both the x and y planes. It is this which gives the wireframe 3D appearance. The loop terminates when temp_radius becomes zero or less than zero. The value of color is then incremented, and if it exceeds max_color is reset to 2.

USING RANDOM NUMBERS

It only remains to discuss the workings of the function calc_num().

Two parameters are passed to the function, which are the minimum and maximum values of the random number to be returned. To generate the random number I have used the function rand() which is part of the ANSI standard for C and therefore should be part of the standard function library in all compilers. It returns a random number in the range zero to the maximum positive integer value for the compiler. This is an important concept, since it means that with Lattice C (used to write this program) which uses 32 bit integers the range of numbers returned by rand() will be 0 to 2,147,483,647 - hence vaLmax is set to this figure. If you use a compiler which uses 16 bit integers, you should change this figure to the maximum positive integer value obtainable, which is 32,767.

If we divide the number returned by the call to rand() by the maximum positive integer value, we obtain a number in the range 0 to 1. This number must be a floating point number, otherwise the decimal part will be discarded. Therefore vaLtemp and vaLmax are declared as doubles. Multiplying this figure by the upper limit and adding the lower limit gives us a figure in the range we want. Since this is still a double, it is cast to a WORD before returning it to the calling function.

I don't claim this is the best way of generating random numbers, but it works for this program. If you use it in other programs, one thing to beware of is that the maximum value returned by calc_num() is not the upper limit you pass to it, but the upper limit plus the lower limit value. You may need to adjust the upper limit parameter accordingly.

LOOKING AHEAD

In these three articles we have now looked at a reasonable selection of the graphics functions available from GEM. Other than the number and variety of different functions there is nothing particularly remarkable about any of them - though the GEMinterface does make them easy to use. In the next article in this series, we should start to look at those functions which make GEM really stand out, namely the construction and use of windows, menus and dialogues.

  Source Code Listing

top