[Soc-2016-dev] Weekly Report #05, Mantaflow Liquids

[Sebastián Barschkis]

Hello everyone,

here is this weeks report for the Mantaflow fluids project:

https://wiki.blender.org/index.php/User:Sebbas/GSoC_2016/Reports/Week_5

If you can follow along then good, if not then don't worry. Things will
(hopefully) become more clear in the next week.

= Week 5 =

This week was a bit of a bummer. I did not find any good solutions for the
problems from last week, got some insights though.

=== What's been done ===

* Regarding the mesh issue in the viewport (missing faces): Since I am using the
old smoke domain mesh for the generated liquid mesh (the one from .bobj.gz
files), wireframe mode is used by default. Obviously, for smoke this makes
sense, for liquids we need to have the faces. I desperately tried to toggle this
behaviour and get faces for the liquids domain but did not succeed. 

* Why does the generated liquid mesh have to replace the domain mesh anyway
(like in Elbeem)? I figured that it might make more sense if a flow object was
used for the generated liquid mesh: It leaves the domain size untouched
(important when later setting up a simulation that combines smoke and liquid
flow objects!) and you don't have to toggle mesh faces in the domain (liquids
with faces, smoke only wireframe). 

* I implemented a prototype for this approach but ended up with a problem in the
cache: Since I am using the pointcache functions to save the .bobj.gz surface
files, I can only access the domain settings. Later when loading the mesh, I can
therefore only set the domain mesh to the liquid mesh.

=== Next up ===

* Work on the liquid inflow. Next week I want to get something done again, so
I'll work on that first. Success rate should be higher there.

* Bang my head against the table until the meshes in the viewport work. No,
seriously, I need to decide how to proceed with the domain mesh issue. In my
view, for effects like "smoke on the water" you need to have a shared domain so
that smoke and liquids can interact. If I cannot figure out a working solution
for that, I will consider the 2-domain approach (one for smoke, one for liquids)
again. 

=== Questions ===

* Which grids (in Mantaflow) cannot be shared in a combined smoke and liquid
simulation?

* Benefits of using the flow meshes for generated liquid mesh (instead of using
domain mesh as done in Elbeem)?

Best wishes,
Sebastián