Thread: How are GUI APIs made?

Just out of interest really, as I've never been able to picture in my mind how one would be made (Buttons, text boxes etc.).

From an object-oriented perspective, you make an object that is capable of responding to events. Then you extend that and make an object capable of displaying an image. Then you extend that and you make an object capable of storing some input from the user. You keep extending these basic components, adding little bits of functionality at a time.

A button, for instance, is just an object that responds to clicks. It displays an image and maybe some text. When you click it, it might change the image and play a sound, then perform some action. When it's done, it returns to it's original state.

A test box is just an image, but when you click in it and start typing, it displays and stores your text, too.

Then people can declare instances of these objects, change their basic properties, and integrate them into their app.

edit:

I recently did this in Flex - we were trying to create a web-based front-end to an already made Java app, so we had to replicate the old look exactly. All we did we define an object that responded to clicks, and then we defined a button, and sliders, and menus, etc... that extended from that. When we needed a "PlayBack" button - we just extended the base button, and added a little more functionality. Each file ended up only being like 15 or 20 lines - really simple way to look at it.

If I were to make an API in C, I would look at it more in terms of just data. I would maintain a list of all objects on the screen, and provide methods to add objects, move them forward or back etc... Each object would be a struct that had their x-y coordinates, their status (enabled vs. disabled), and "type". Then depending on the "type", I would call a certain method to display that type, and based on it's x-y coordinates, size, status, and other data, it could display it on screen.

Then any time someone clicked, I could go through the list, see if the click occurred on that object, and then execute any commands that were associated with a "click" for that object.

There are made using existing libraries.

Originally Posted by sean

From an object-oriented perspective, you make an object that is capable of responding to events. Then you extend that and make an object capable of displaying an image. Then you extend that and you make an object capable of storing some input from the user. You keep extending these basic components, adding little bits of functionality at a time.

A button, for instance, is just an object that responds to clicks. It displays an image and maybe some text. When you click it, it might change the image and play a sound, then perform some action. When it's done, it returns to it's original state.

A test box is just an image, but when you click in it and start typing, it displays and stores your text, too.

Then people can declare instances of these objects, change their basic properties, and integrate them into their app.

edit:

I recently did this in Flex - we were trying to create a web-based front-end to an already made Java app, so we had to replicate the old look exactly. All we did we define an object that responded to clicks, and then we defined a button, and sliders, and menus, etc... that extended from that. When we needed a "PlayBack" button - we just extended the base button, and added a little more functionality. Each file ended up only being like 15 or 20 lines - really simple way to look at it.

If I were to make an API in C, I would look at it more in terms of just data. I would maintain a list of all objects on the screen, and provide methods to add objects, move them forward or back etc... Each object would be a struct that had their x-y coordinates, their status (enabled vs. disabled), and "type". Then depending on the "type", I would call a certain method to display that type, and based on it's x-y coordinates, size, status, and other data, it could display it on screen.

Then any time someone clicked, I could go through the list, see if the click occurred on that object, and then execute any commands that were associated with a "click" for that object.

But how would you go about putting the image for a button on the screen? Or even turning the console into a window that could have more than text put in it?

Originally Posted by MK27

sometimes just hair and nails and puppy dog tails.

I guess stalks don't like Microsoft then...

But how would you go about putting the image for a button on the screen?

Using a graphics library, the same as any game, like OpenGL or something would work. However, I would imagine that more professional GUIs have their own libraries. At some point, you have to use assembly and place your pixels - anything above that is really just a graphics library of sorts.