and generate almost zero garbage during the loading process (unlike the
ImageIO loaders which practically generate an entire image worth of
garbage during the load).
git-svn-id: svn+ssh://src.earth.threerings.net/narya/trunk@2474 542714f4-19e9-0310-aa3c-eee0fc999fb1
- Reworked colorization repository such that we now have arbitrary named
"colorization classes" which can be used by various and sundry entities
to colorize themselves.
- Added support for individual object colorizations as well as "spots"
that go along with objects (will be used to automatically create portals
into buildings and automatically position users properly when at a
"station").
- Repackaged things in miso to more closely mimic what we do everywhere
else (no more miso.scene, now we have miso.data and miso.client).
- Fixed up miso scene XML representation so that objects can more easily
be expanded to have yet more stuff if we think of more stuff that they
might aught to have in the future. Structured the miso scene model so
that "uninteresting" objects (those that simply sit somewhere and don't
do anything) take up a lot less memory than "interesting" objects (those
that have action strings, "spots", colorizations and the works). I may
want to roll colorizations into the "uninteresting" realm, but that
remains to be seen.
- Made it possible for object tilesets to specify default render
priorities for objects in that tileset. This will hopefully allow us to
get by without any "custom" render priorities that are specified on
scene objects, instead relying on the default priorities to resolve
common conflicts.
There are surely other cleanups in there, but I think that was the major
thrust of it.
git-svn-id: svn+ssh://src.earth.threerings.net/narya/trunk@2228 542714f4-19e9-0310-aa3c-eee0fc999fb1
convert whatever its internal format is into ARGB for our inspection. It's
arguably a smidge slower (though we eliminated the multiple method calls
per pixel to determine whether it is non-transparent), but will work with
any source and destination image format.
git-svn-id: svn+ssh://src.earth.threerings.net/narya/trunk@2220 542714f4-19e9-0310-aa3c-eee0fc999fb1
createImageMask() not to freak out if the source image is not RGBA, Java
automatically unpacks pixels appropriately for 8-bit colormapped images,
so there's no need to be so picky. In createTracedImage() we were
converting the image to a screen optimized image anyway and thus the
source image format had absolutely no bearing on what we did with the
image later.
git-svn-id: svn+ssh://src.earth.threerings.net/narya/trunk@2218 542714f4-19e9-0310-aa3c-eee0fc999fb1
number of bytes used rather than a certain image count.
git-svn-id: svn+ssh://src.earth.threerings.net/narya/trunk@2173 542714f4-19e9-0310-aa3c-eee0fc999fb1
strange slow rendering problems we see when we allow mixed BITMASK and
TRANSLUCENT images to be rendered.
git-svn-id: svn+ssh://src.earth.threerings.net/narya/trunk@2140 542714f4-19e9-0310-aa3c-eee0fc999fb1
- All images are loaded through the image manager
- Architected to allow the use of prepared or unprepared images (and
volatile images when the day comes that they support alpha)
- Tunable caches for images and tiles
- Resource manager caches unpacked resources on the filesystem
- Various and sundry other cleanups
git-svn-id: svn+ssh://src.earth.threerings.net/narya/trunk@2116 542714f4-19e9-0310-aa3c-eee0fc999fb1
we're done loading images from them.
Always load through a BufferedInputStream because that seems to fix the
bogosity with ImageIO.load() not working without a cache.
git-svn-id: svn+ssh://src.earth.threerings.net/narya/trunk@2098 542714f4-19e9-0310-aa3c-eee0fc999fb1
classpath. Removed getImageSource() as calling through to the resource
manager is simpler.
git-svn-id: svn+ssh://src.earth.threerings.net/narya/trunk@1972 542714f4-19e9-0310-aa3c-eee0fc999fb1
Michael Bayne