Fixes the black screen issue on macOS, fixed HiDPI rendering, add keyboard interaction

[NaivelyWritten]

This PR fixes #15, added keyboard interactions, and also fixed the wrong mouse interaction under HiDPI environments such as on macOS retina screen.

[rajkundu]

This looks great! Unfortunately haven't had a chance to test this PR since I switched jobs a few years ago.
One question: how come the math for xpos and ypos are so dissimilar? I.e., why is there no viewportSize.x term for xpos, and why is io.MousePos positive for xpos and negative for ypos, and screenPos positive for ypos and negative for xpos?

ImVec2 screenPos = ImGui::GetWindowPos();
ImVec2 viewportSize = ImGui::GetWindowSize();

auto dpiScale = ImGui::GetWindowDpiScale();
double xpos = (io.MousePos[0] - screenPos.x) * dpiScale;
double ypos = (screenPos.y + viewportSize.y - io.MousePos[1]) * dpiScale;

fa38496#diff-86bc0a1a2a83bc219e04944ce2a5cd6f9fccecb6b1f993ac58e00453427d30f7R54-R55

[ProfFan]

Hi @rajkundu it's because the mouse position is in the ImGui coordinate system and the x and y poses are in the VTK coordinate system, they are defined on different corners

[rajkundu]

Ah, okay. The different corners makes sense for the inversion, but why is viewportSize.x not used in the xpos calculation?

[ProfFan]

@rajkundu because they both share the x axis, viewport size is used to convert the y:

+ (top left screen corner) - - - - - -
|
|
+ ( top left window corner)
|
|
+ (bottom left window corner)
|
+

[rajkundu]

Got it! Makes perfect sense, and thanks for drawing it out. I will test this when I can, likely in the next few months, but in the meantime, I'm glad that this PR is here for others to see and check out. Thanks for your contribution!

[rajkundu]

I still haven't had a chance to test this :(

@NaivelyWritten have you been able to test on anything outside of macOS/HiDPI?