In [1] it was a good move to remove the periodic barrier.
Filtering spheres is another independent question and I don't see a
clear reason for that (testing isDynamic was maybe a bit hacky but less
restrictive finally).
Making split=0 and split=1 return the same thing [2] sounds good, but
the problem was to filter spheres with split=0 in the first place [1].
Before, it was possible to get fabric in a periodic packing of cylinders
for instance. Now it is not. Not a progress overall.
What is the problem if non spherical objects are kept in the loop? We
should solve this problem without regression.
Bruno
On 06/06/2016 10:47 PM, Jerome Duriez wrote:
I'm replying to
http://www.mail-archive.com/yade-dev@lists.launchpad.net/msg11970.html
(sorry to break the thread, but there has been a major email shutdown
at my university last week, and I just discover now this message,
browsing the archives)
So, in fact I started to introduce such kind of spherical shape-test
in a previous commit [1], where I let fabricTensor() accept
non-periodic simulations (before this first commit [1], fabricTensor()
crashed in such non-periodic cases).
At that time, this shape test was intended to let fabricTensor()
disregard any boundary effects, by comparison with the previous
behavior restricted to periodic simulations.
However, I introduced in [1] this shape test only in the main workflow
of fabricTensor() code, which has a role when split = 0 (default). For
the special case split=1, other lines of code do the job, where I did
not introduce any shape test.
Then, considering classical dry simulations (with interactions at
geometrical contact only) of spherical packings loaded by plates, you
may get slightly different results between
- fabricTensor()[0] that disregards boundary interactions
- and fabricTensor(splitTensor = 1,thresholdForce =0)[0] that included
boundary interactions
Even though, both should apply to the same contact interactions
network, from my point of view
This second commit [2] aimed thus to correct this mistake, introducing
the shape test in the "split=1 code part" as well, and reconciling the
two results offabricTensor()[0] and fabricTensor(splitTensor =
1,thresholdForce =0)[0] for such classical dry simulations.
As you see, all this arises from my current point of view that
non-spherical bodies are always used as boundary elements in
non-periodic simulations. I can introduce changes (in addition to the
ClassIndex suggestion, thanks) if you think other behaviors are
required / meaningfull
[1]
https://github.com/yade/trunk/commit/562d4c952f4b7f67a88ed954caa20b68a041e207
[2]
<https://github.com/yade/trunk/commit/562d4c952f4b7f67a88ed954caa20b68a041e207>https://github.com/yade/trunk/commit/e063ea12479a56f85ca456aef8f52be19cbed84d
<https://github.com/yade/trunk/commit/e063ea12479a56f85ca456aef8f52be19cbed84d>
fabricTensor(): unify the behavior regarding boundary interactions wh…
· yade/trunk@e063ea1
<https://github.com/yade/trunk/commit/e063ea12479a56f85ca456aef8f52be19cbed84d>
github.com
…ether split=0 or 1: they are now disregarded in both cases
<https://github.com/yade/trunk/commit/562d4c952f4b7f67a88ed954caa20b68a041e207>
fabricTensor() now ok for non-periodic simulations. revertSign attrib…
· yade/trunk@562d4c9
<https://github.com/yade/trunk/commit/562d4c952f4b7f67a88ed954caa20b68a041e207>
github.com
…ute removed as well
--------------------------------------------------
Jerome Duriez, Research Associate
University of Calgary, Dpt of Civil Engineering
+1 403 220 7367
_______________________________________________
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help : https://help.launchpad.net/ListHelp
_______________________________________________
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help : https://help.launchpad.net/ListHelp
