Source files for images in Magnum documentation
-----------------------------------------------

Compile and install Magnum with `Sdl2Application`, windowless application for
your platform and `magnum-distancefieldconverter` utility and any `PngImporter`
and `PngImageConverter` plugins from Magnum Plugins.

Create build dir, point CMake to this directory and compile the executables:

    mkdir build-doc
    cd build-doc
    cmake ../doc/generated
    cmake --build .

### Primitive images

Generated by the `primitives` executable. Run it in this directory, the output
is put into `doc/` directory. Apply `pngcrush` to them for smaller file sizes:

    for f in $(ls primitives-*.png); do pngcrush -ow $f; done

### Shader images

Generated by the `shaders` executable. Must be run in this directory, the
output is put into `doc/` directory. The executable requires two textures:

-   `vector.png`, generated as full-page PNG output at 96 DPI from `vector.svg`,
    converted to pure grayscale using imagemagick:

    ```bash
    mogrify -flatten -background '#ffffff' -colorspace gray vector.png
    ```

-   `vector-distancefield.png`, generated as full-page PNG output at 384 DPI
    (1024x1024) to `vector-src.png` and then processed through
    `magnum-distancefieldconverter`

    ```bash
    magnum-distancefieldconverter --importer PngImporter --converter PngImageConverter --output-size "64 64" --radius 16 vector-src.png vector-distancefield.png
    ```
Apply `pngcrush` to the result for smaller file sizes:

    for f in $(ls shaders-*.png); do pngcrush -ow $f; done

### Distance field vector images

Similarly as above, just with slightly different parameters. The
`distancefield-src.png` is generated as full-page PNG output at 192 DPI
(512x512) from `vector.svg`, the `distancefield-dst.png` is then converted with

```bash
magnum-distancefieldconverter distancefield-src.png distancefield-dst.png --output-size "128 128" --radius 24
```

This is chosen so it corresponds to the `TextureTools::DistanceFieldGL` doc
snippet where it takes a 256x256 image and converts it to 64x64 with a radius
12, but is scaled twice to look better on HiDPI displays.