This is the repository of the v1 and v2 version of BIM Surfer. It is not maintained anymore. Find the most recent version on https://github.com/opensourceBIM/BIMsurfer
This can be seen when taking multiple snapshots of the camera standing still. There should be no change in the camera up vector. (might be due to a normalization error somewhere possibly for calculating the camera quaternion)
The exported geometry for WallStandardCase covers the windows of the house model. (Something to do with the IFC geometry engine, or something about WallStandardCase that needs to be taken into consideration? What's the difference between Wall and WallStandardCase?)
The rotation speed of the camera still feels a little awkward - it rotates at a different speed when viewing from the top to viewing from the side, probably this calculation is still not 100%
When turning the model, the shadowside comes to the front. It's pretty dark....
Always having the sun on the back of the camera makes sure that the viewable area is always light.
For the name nodes that are involved in mouse picking (and later the object hierarchy) we're using the geometry id's, but these names appear to be codes rather than proper names. Find out if there's a better way...
Possibly limit the camera's orbit at (0,0,1) and (0,0,-1) (or possibly at the floor?) to avoid turning it upside down - this is a bit of a trade-off. Locking the camera doesn't feel very smooth to the user, but turning the camera upside down is also a little unnatural - especially when viewing buildings since these can't usually be viewed from these strange angles.
When snapshots exceed the size hieght of the window a scroll bar should be created in their div.
At the moment the help text moves down when the snapshots go below the bottom of the screen (probably the viewport div becomes larger and absolute positioning for help text is relative to the bottom of this div)
The active working area for the mouse should be the entire screen (and must not be hampered by e.g. help icons etc)
In addition the canvas should be either fullscreen or 1024x1024 depending on the performance setting in the top menu
(It is said that full screen canvas can be slow on slow machines - so giving people the option of resizing... should test this hypothesis first though)
I believe it may be more intuitive for users to perform orbit manuevers as well as selection with the left mouse button and then use middle mouse only for panning.
The test model AC11-FZK-Haus-IFC looks a little bit strange
There's no lighting on the model (are the normals broken?)
The colors are pure white and pure black which is hard to comprehend
Also there appears to be some strange backface culling?
Still thinking about this... might be better to fix a fov on the x-axis since the y-axis is more likely to change. Possibly we could even use a diagonal fov
Some browsers do not use smooth sampling to scale down images in src/canvas.thumbnails.coffee. Would be nice if they did at least some linear sampling.