doubles and various (un)signed integer lengths

[cgloeckner]

Hi, is there any support planned for doubles or various (un)signed integer lengths?

[ocornut]

Hello,
Which widgets were you thinking? Probably not soon for all widgets, but if you want to input them with keyboard like InputInt you can create a function that does sprintf + InputText + scanf to work with any type you want.

[cgloeckner]

I was thinking about InputInt as guessed. Have you got an example for that sprintf + InputText + scanf? I cannot imagine how oO

[ocornut]

Print your number into a string, call InputText, when it return true scanf back from string to number. Literally 4 lines of code. Look up how those functions work.

[cgloeckner]

My problem is to avoid the additional string handling, because (1st) an extra variable is necessary to keep the string state and (2nd) the imgui API doesn't support std::string for writing purposes. Or did I get something wrong?

[ocornut]

bool InputDouble(const char* label, double* mydouble)
char buf[64];
sprintf(buf, "%g", *mydouble);
if (ImGui::InputText("....."))
    sscanf(buf, "%s", mydouble);

[cgloeckner]

I'm a bit confused: Text (like in the example) or InputText like above?

[ocornut]

Sorry I meant InputText, was a typo..

[cgloeckner]

Ok. Once I need more then plain keyboard input (like +/- buttons), I'd need to workaround-build this, too.. right?

[ocornut]

Yes. But for integer than that are smaller than int you can also just copy your value into an int and call any Int function.

[cgloeckner]

True, but the overflow/underflow won't be handled correctly. I'd prefer an imgui-internal solution ;-)
Have you done any research on pitfalls for changing the current implementation towards that feature?

[ocornut]

I'd prefer an imgui-internal solution ;-)

Well I'd prefer a full-time job working on ImGui!

What overflow?
You can set limits to InputMin, SliderMin, etc.
It's really low priority, I've got millions of things to do on ImGui and already overwhelmed with tasks that are arguably more important. This is sort-of rare use case and there are perfectly usable workaround so it is unlikely I would want to work on a better solution anytime soon.

[cgloeckner]

Ok thx :)

[ocornut]

FYI, I have added functions in a branch to work with float, double, 32/64 bits, signed and unsigned (and in a proper lossless manner, as previously some functions didn't work great with big integers).
See #643 for details.