Added joystick example.

[binary1248]

Originally used when testing #1326 but can be used with the existing code as well.

[MarioLiebisch]

If there's no controller connected it will still spit out basically placeholder text for the first controller. Maybe there should be some instruction like "No controllers could be found. Connect a gamepad or joystick to test it."?

[binary1248]

The text that is always output isn't placeholder text, they are simply the labels of the joystick items. When a joystick is connected, values will be displayed for the last joystick whose state was modified. When no joysticks are connected, no values are displayed.

[MarioLiebisch]

Yeah, but (at least to me) from a first look it's still not clear whether the program is stuck or just couldn't find any joystick. So maybe just rather than setting ID to nothing, set it to <Not connected>, if there's no controller? Similar not existing buttons/axes should possibly output "n/a" rather than nothing.

[MarioLiebisch]

I'm not really sure about the whole text output only. I'd prefer something visual, even if it just consists of simple shapes. Maybe something like this (potentially with more text):

[binary1248]

The point of this example is to demonstrate usage of the joystick API. The predecessor to this code would output purely to the console, however it would become unreadable when outputting all the possible values of a joystick's state hence the sfml-graphics output. The current text rendering already takes up almost half of the code. Making it any more sophisticated such as in the picture you posted will shift the focus of the code away from the joystick API to the drawable API which shouldn't be the case in a joystick example. This should stay as an example and not become some kind of debug tool.

[MarioLiebisch]

Yeah, probably true,

[MarioLiebisch]

Also, maybe just me, but I prefer to reserve i, j, etc. for numerical indices, while using a, b, etc. for iterators (even though [i]terator makes sense).

[LaurentGomila]

My personal convention, and thus most likely what's used in SFML, is to use it for iterators.

[MarioLiebisch]

While convenient and using existing classes, I'd really just define a small struct for the values here, even if it's just to identify the components as label and value or something like that. It's about readability and the chained expressions with first or second just feel a bit odd.

[LaurentGomila]

Agreed. std::pair is never a good choice.

[MarioLiebisch]

While it's unlikely the enum values for the axes might change, I'm not really a fan of just assuming a constant/fixed order here.

[LaurentGomila]

In core SFML code I would agree, but this is just an example, so it's a good thing if we keep this kind of boilerplate code as short as possible.

[MarioLiebisch]

As an added feature, maybe allow the user to increase/decrease the movement threshold with some keys, like Page Up/Down, visualizing what it does?

[eXpl0it3r]

There are still some open suggestions, are you going to change things @binary1248? I mean it's just an example, doesn't have to be the best thing ever.

[binary1248]

I still intend to work on this... yes.

[binary1248]

@MarioLiebisch, @LaurentGomila, @eXpl0it3r: All the mentioned issues should be resolved in the current version.

[eXpl0it3r]

I don't own a joystick, can someone test this?

[danieljpetersen]

I just tried this out on Linux Mint 18.3

• The Steam Controller identified as a 360 controller. It doesn't seem like this is working correctly here. Only three buttons have an effect -- The threshold value goes up when you press the X button, and it goes down when you press the Y button. Pressing the start button causes the program to exit, which I guess is to be expected.
• Everything works perfectly with the Logitech Extreme 3D (this thing)
• The PC version of the 360 controller also works perfectly for me.

[eXpl0it3r]

Thanks for testing!

The Steam Controller identified as a 360 controller. It doesn't seem like this is working correctly here. Only three buttons have an effect -- The threshold value goes up when you press the X button, and it goes down when you press the Y button. Pressing the start button causes the program to exit, which I guess is to be expected.

Sounds more like a potential driver issue. If you have time to track down the issue, maybe there's something we need to adjust on our side?

[danieljpetersen]

Perhaps it is a driver issue. Apparently the Steam Controller isn't mainlined into the Linux kernel yet, and you need to have Steam running in the background order for the controller to properly work.

I'll test it out on Windows.

[danieljpetersen]

Hm. I get the same result in Windows, except this time it's not reporting that any controller is connected. I still think you're right though, it looks like the controller is mapped to keyboard keys and not presenting itself as a joystick?

[binary1248]

This PR doesn't change the way joysticks are handled by SFML. It is merely an example of how to use the API. If wrong values are displayed, then any application making use of the SFML joystick implementation will also function incorrectly. As such, if you find and are able to reproduce any bugs in the current joystick implementation, you should discuss the issue on the forum and possibly open an issue against the actual SFML joystick implementation.

[eliasdaler]

Logitech Gamepad F310 on Linux - everything works as expected in DirectInput and XInput modes.

[mantognini]

Overall, it looks good and works nicely on macOS.

[mantognini]

What is the motivation for this global ss?

[binary1248]

Having to construct/destroy the ss everywhere where it is currently used would cause unnecessary overhead. I would have liked to use std::to_string but we aren't there yet. 😉

[LaurentGomila]

Having to construct/destroy the ss everywhere where it is currently used would cause unnecessary overhead.

Is it something that you measured? I've always seen people instantiating (or recommending to instantiate) this locally, only where it is needed.

[binary1248]

I always use stringstreams to stringify numerical values, also when getting a rough feel for performance (frame times) as you know. I noticed a significant delta when recreating them constantly, which is why I made it a habit to just resort to using long-lived ones. It is much faster to just empty the buffer and re-fill it constantly than always have to construct/destruct the object including its internal state etc.

[mantognini]

I don't know about the perf cost, never had to measure it. I guess one could always encapsulate this using a static variable in a function. Not saying you should do it, just a random thought.

[mantognini]

I don't have this key on my keyboard. Up seems more suitable.

[mantognini]

Idem with Down.

[binary1248]

@mantognini Fixed.

[eXpl0it3r]

There's a small conflict due to the iOS Demo being merged.

[binary1248]

@eXpl0it3r Fixed.