* Moved `UIComponent::GetViewport()` to `Scene` class.
* Simplified resource initialization of `UIComponent` and removed need of overriding `IsEnabled()`.
* Renamed `UIComponent::GetElement()` to `GetRoot()` to be more consistent with UI subsystem.
* Removed ability to set root UI element to `UIComponent`, instead to add UI elements should be added as children to element renturned by `UIComponent::GetRoot()`.
* Cleaned up debug rendering in HelloGUI sample. Debug data is now rendered after pressing F2.
* Separated batches and vertex buffers of normal UI and UI rendered to texture.
* Cleaned up `Render()` method adding additional `surface` parameter to render into. Passing null renders on to screen.
Support for rendering UI to texture.
This change adds support for rendering GUI into a texture. This is achieved by adding `UIComponent` component to `Node`. `UIComponent` overrides material of `Node`'s `StaticModel`. UI subsystem is changed so that batches belonging to UI elements owned by `UIComponent` are rendered into a texture that is also contained in `UIComponent`. Input is handled by UI subsystem, so anything that worked on normal UI will work on UI rendered on to 3D object. This required a minor change for input handling. Previously "screen" was considered surface of entire window. Now screen is considered area of root_element_position + root_element_size. This is fully backwards-compatible.
There is one issue i am unable to solve: since on OpenGL texture coordinates start from bottom-left corner a workaround is needed. Texture is flipped by projection matrix, however it breaks scissor test. I do not know how to properly fix it therefore scissor test is disabled when rendering into texture on OpenGL. Issue is clearly visible when scrolling a `ListView` in `HelloGUI` sample when UI is rendered on to a cube.
Other changes: lua and angelscript bindings, HelloGUI sample can toggle between rendering on screen or rendering on cube, added a ListView to HelloGUI sample window. `UI::DebugDraw()` support.
Subpixel-positioned text looks blurry, due to bilinear filtering of the
underlying texture. The basic idea here is to stretch the font glyphs
horizontally to increase the sharpness at subpixel positions. The
stretched images need to be smoothed to avoid aliasing artifacts; this
is done in FontFaceFreeType::BoxFilter().
Glyphs are always pixel-aligned vertically, so no vertical oversampling
is needed.
To make this feature comprehensible (I hope!) I've removed the
'subpixelGlyphPositions' flag and replaced it with a couple of values:
'fontSubpixelThreshold' sets the point at which subpixel positioning
kicks on, and 'fontOversampling' controls the amount of stretching.
The default values are 12pt text and 2x oversampling. These are fairly
conservative settings, which should improve small text without wasting
a lot of memory.
Note that when the font hint level is NORMAL (the default), subpixel
positioning and oversampling are both disabled. So, this feature doesn't
change any default behavior, and applies some hopefully sensible values
if the hint level is set to LIGHT or NONE.
When this option is enabled, text will be formatted with subpixel
(fractional) positions on the x-axis. Positions on the y-axis are
still pixel-aligned.
Note that this option has no effect if the hinting level is set
to FONT_HINT_LEVEL_NORMAL, as each glyph is rounded to an integral
pixel size by the hinter. It only makes a different if the hinting
mode is LIGHT (vertical hinting only) or NONE.
With subpixel positioning, the output will look blurrier due to
texture filtering. TODO: Add horizontal oversampling to improve
sharpness. That needs extra memory so it should be configurable.
This commit changes the 'pointSize' parameter in Font, Text
and Text3D from an int to a float, allowing e.g. 14.5pt text.
Note that when hinting is enabled, font metrics are snapped
to pixel boundaries, so the effect may be hard to see unless
you also set UI.FontHintLevel to LIGHT or NONE.
This is a change to the public API, but existing code (including
scripts) should compile and run as before.
There are three levels: NORMAL (the default), LIGHT and NONE.
The LIGHT level makes FreeType align font outlines to pixel
boundaries vertically, but not horizontally.
Displays text in various sizes, with checkboxes to play with settings
that affect the way text is rendered:
- Color (white-on-black versus black-on-white)
- Autohinter
- sRGB output conversion (not available on all platforms)
