I don't know the first thing about STEP files, so I have no comments about the logic, but please don't use List<T>. It's just a bespoke implementation of std::vector<T> which doesn't offer any benefits other than implicit shallow copy for code that needs it (at the cost of requiring users to manually free the memory). The STEP code isn't such a case, so please switch to std::vector<int> and use ranged for loops.

Also, while SolveSpace is single threaded in that code path, please make the lists and the functions that operate on them members of StepFileWriter. There's no reason to add more global state.

For cross reference:
#206 (comment)
#1580

The "correct" STEP export is neither this nor #1580 but some "nonlinear combination" of both may be :-)

@ilguido would you mind sharing the file that you're using so I can use it as a reference?

I'm starting to think that this approach (discovering matching/duplicate points/lines/curves numerically at this level instead of trying to deduce it from the structure) may not be so bad...

@ruevs I'm also in favor of doing it "the right way", but this would be the 3rd attempt by non-core developers to resolve the issue (of step file quality). When there is that much interest I think it's time to do something, and if you or I don't have the time to work through a "more correct" solution I think it's time to accept one or more of these others. It may be worth pointing out that these "fixes" to the existing exporter do not impact the rest of the application - they are limited to the step file exporter.

Based on the image at the top of this PR it looks like this one handles the PWL curves as well, and to me that's important. I kind of like the snapping of #1580 but that's secondary and I'm not sure if that one handles PWL curves yet.

@ilguido I agree with @iscgar comments (above) on this PR and using standard data structures instead of the solvespace Lists where possible and not introducing globals.

You solution does "the right thing" when exporting the two samples I tried:
#206 (comment)
1mmCube.zip

unlike #1580

Exported STEPS files:
1mmCube_master_vs_ilguido_vs_rcarmo.zip

The refactored code should be OK now. I further tested it, but I invite you to do your checks anyway.

@ilguido you imported into gmsh. Do you do FEA work? If so you might have something to say under issue #866

I've merged the other version of this PR.