On Fri, Sep 23, 2016 at 11:26:31AM +0200, Jean-Marc Lasgouttes wrote: > > Actually, most of the used of getPixmap are like > QIcon(getPixmap(...)).pixmap(somesize); > > It looks like a strange code path to me. What does getPixmap return in the > case of a svg? Some rasterized version of the svg? Then how does it help to > go through a QIcon and then back to QPixmap?
I am not really an expert, but QIcon takes care of loading, scaling and post-processing of the image. Moreover, QIcon may contain multiple images with different sizes (most probably some common used sizes). So, when you request a certain size, it probably gives you a scaled version of the nearest precomputed size. This may explain the better quality. > It would look better to me to have a getIcon function which returns a QIcon, > so that we can directly render the svg at the right size. Or did I > misunderstand how this works? Seemingly, QIcon already provides multiple images with different sizes, such that the best fit is chosen depending on the size to be used. But, I repeat, I am not really an expert and mine are speculations rather than facts. -- Enrico