OpenVSP 3.52.0

This one almost went out the door as a .4 bugfix-only release made from trimmings on the floor resulting from a handful of big features I am working on. Those big features aren’t quite ready to release yet, but by the time I was done sweeping up the mess, it looked more like a .0 than a .4 release. In addition to the bycatch, there are a group of fixes for problems identified by some users.

The tidy up was initiated because I was getting annoyed dealing with fixes in four separate piles, so I decided to stack them all together and release them. Now I can rebase all my work on top of this and keep going. The upcoming features are really cool, but you’ll just have to be patient. If you like fewer bugs in your tools, you should still update.

If you rummage around in the commit messages here, you’ll find breadcrumbs about the in-progress features. I’m not saying what they are, but I’m also too lazy / not paranoid enough for the level of op-sec to go back and scrub everything. On the other hand, I am aware of the leaks and I don’t want you to think you got one over on me.

The most ‘feature’ part of this release is a handful of improvements to FitModel. Most importantly, FitModel has gained API support for all your automated fitting needs. In the GUI, things are a little nicer, with a few small tweaks – most noticable, you can now ‘undo’ an optimization step.

The most scary thing here is a change to how ID’s are remapped when collisions happen. This is a substantial change to something that is very core, so I don’t take this lightly. OpenVSP uses unique ID’s for a lot of entities (namely Parms and Geoms). Sometimes (when you insert a file into itself, or when you paste a duplicate geometry) collisions occur — the new entity has the same ID as the preexisting entity. This is bad. So, previously when a collision is detected (during the paste operation), a new ID is created and any later (during the paste) reference to the colliding ID is replaced with the new ID on the fly.

The problem arises when an ID within the pasted set of entities referrs to an entity outside the pasted set of entities. In that situation, a new ID is created, but no new entity is created with that ID. This results in an invalid link. This was safe for about a decade because the problem never came up. However, the introduction of Routing Geometry (3.43.0) made it possible. If you copy/paste a Routing Geom without including everything the Routing geom, then the route points will point to nowhere.

So now the ID remapping is a two-pass process. On the first pass, new ID’s are created for entities within the paste set (but not for references). On the second pass, things are created with the new ID’s. Any ID that referrs to an entity outside the paste set will be preserved. So, if you copy a Routing geom without any of the things it points at, it will still work.

If your Routing geom points at GeomA and GeomB, and then you copy/paste the Routing geom and GeomA (but not GeomB), the new Routing Geom will now point at GeomACopy and GeomB (the original). It is pretty slick.

I say this is scary because it is really core code that is pretty tricky and hasn’t been touched in ~10 years. Any time I go near it, I get a bit nervous. That said, I’ve beaten on this quite a bit and I feel good about it.

A few of the bugs were user reported. Keenan noticed that a recent fix for how planar slice automatic bounds were calculated had changed the bounds on a test case — hopefully the new bounds will achieve the best of both worlds.

Mass Properties calculation is now faster and more accurate. There was a bug in the inertia calculation for solids that I had been compensating for by jacking up the tessellation of each slice. That made it slow, the fix was to find the bug and then not over-resolve. Thanks Mike R. for the test case.

The GUI controls for the N parameter used from attachment could not be used. They lie within a box that was too small.

Converting a zero height (or width) rounded rectangle to CEdit would crash. The CEdit screen would also crash if you deleted the XSec while the GUI was still open. Both fixed.

The *.vkey file written alongside *.vspgeom files was missing some surfaces if there were no triangles outside a subsurface. I.e. if you walpaper your entire Geom with SubSurfaces, then the surface would vanish from the *.vkey file. I don’t know why Dave insisted on doing this, but now it works.

There has been a problem using FindParm( container_id, name, group ) from the API — you couldn’t find all Parms. Hopefully that has been sorted and Parms from SubSurfacex, XSecs, Textures, Mesh Sources, Bogies, Routing Points, etc. should all be findable. Lights, control surface groups and FEA assembly connections were entirely unreachable. They should be reachable now.

VSPAERO control surfaces should now work on geom’s whose name contains an underscore. The Reynolds number was scaled wrong in VSPAERO, this has been fixed.

There was another pretty heavy pass over the API – auditing coverage and documentation. Hopefully things are in a better place and we won’t have another heavy pass like this in the future.

The ‘Scale’ feature in OpenVSP has been chronically plagued with errors of omission. The whole thing has been audited and hopefully we’re close to full coverage for now. If you try to scale a model and it doesn’t scale the way you’d like, let me know.

The process of resetting everything when a model is renewed has been simplified and hopefully made more complete. Parms now remember their initial value, so resetting to default is straightforward.

Bryan S. had AI help him fix a bunch of issues he was encountering with FEA Structures. Thanks

In certain situations of degenerate surfaces, normal vector calculation has been improved. This will remove some visual artifacts that have been bothering me for about a decade.

And a bunch more small fixes all over the place. There are so many fixes that even looking at the changes I don’t remember them all. They certainly aren’t all listed here.

Features:

• FitModel undo button
• FitModel GUI improvements
• FitModel API access
• Read point clouds as .csv files
• Read unformatted Plot3D and Cart3D .tri files
• Expose Parasite Drag flow conditions in Results – thanks Ojasvi

Bugs:

• Fix ID remapping when referencing an ID outside of the DecodeXml set
• Fix scaling of Reynolds number in VSPAERO
• Fix Mass Properties calculation – now better and faster
• API access to lights, control surface groups and FEA assembly connections
• AddMaterial now writes the material to the file
• Tag file names carry a . separator; CFDMesh writes them to a subdirectory
• Fix Parm lookup by container, name and group after a file load
• Fix NGonMesh contributing nothing to CompGeom and mass properties
• Fix WireGeom not moving when the Geom moves
• Fix WireGeom’s four patch types sharing one Parm name
• Fix VSPAERO control surface lookup for a Geom name containing an underscore
• Fix export crash on a model with no geometry
• Fix crashes and bad geometry from degenerate plates and zero area triangles
• Fix a polygon mesh being used after the sweep that deletes it
• Fix vspgeom key file dropping a part covered entirely by subsurfaces
• Fix Cobra exiting on a zero nose or aft length
• Fix Geom rename not reaching the rest of the model
• Register three GUI screen enums and the Background3D error code for scripts
• Fix the ‘N’ attachment buttons being outside the layout
• Fix colour picker deactivation