DeepSOIC wrote:Can you please also check out this one:
git commit f41f0ac49a566d0525cd147486919ef300144019?
I got rid of arc's angle tracking constraints completely, which is beneficial in that there are less solver parameters required for ellipse arc. The same thing can be done for circular arcs and I think it is worth doing. Sorry for duplicating my own self
A code cleanup will be required after the commit, since ArcOfEllipse.****angle are not used anymore; the new constraint with that atan is not needed too...
I took a look at that commit and I understand the change in the code. However, I am still not convinced that the change would bring any satisfactory result. Unless the change, for example would solve Ulrich frozen sketch or any other problem... could you check this?
I explain where I stand:
1) When I code, I try to replicate the structures already in place, so that the code is uniform.
2) The arc of circle is coded in this way. Probably there was a reason to do it so. It is not that I think that what it is now is coded by powerful beings that know better (though probably it is neverthess true
) and that that code can not be changed, but I would like to have a sound reason to depart from their approach.
3) In principle, removing two parameters with a simple parameter dependency (they merely touch two columns) should not bring any advantage to the solver (unless the constraint is wrongly coded) . That is really peanuts for such a solver running in a modern computer.
Nevertheless, I have not tried the commit yet, can you tell me if you can appreciate any advantage in general? have you tried the problematic sketches Ulrich submitted with it, any improvement? If for example it fixes the frozen sketch of Ulrich
, I integrate it right away
BTW, I still had something for you from another post, about Ctrl-A used for alignment constraint. The shortcuts of the sketcher are getting complicated (mostly thanks to me). So after this project is finished, I plan to dedicate so time to try to improve the situation...