Benjamin Winkel

Dear @brandon-rhodes, thanks so much for your detailed thoughts. It's great that you're interested in the package! In your example, the problem is with the interface. The `propagate_many` function needs...

> The documentation on PyPI thanks for providing a useful page with examples, by the way! does not include a dedicated link to the repository, which was a slight inconvenience....

> The mighty propagate_many() interface, which will be of the most interest to Skyfield folks, unfortunately does not cache the Satellite objects it creates but takes raw TLE lines that...

> All Skyfield needs is ECI coordinates. It would be magnificent if there were a way, after turning a list of TLE lines into a native array of Satellite structs,...

> The ideal interface would be (1) a function that takes TLE lines and returns an array of raw inlined SGP4 structs, and (2) a function that takes the structs...

> Just for fun, I'm going to look today at other options for a thin wrapper around SGP4 that might serve Skyfield without requiring you to shoe-horn extra calling modes...

Wow, that is amazing. For the record, I also ran the `cysgp4` on one core (using the `set_num_threads` function), and on my machine the results are 14.7 ms vs 8.6...

PS: If one wanted to parallelize the Vallado code, a deep copy of the `satrec` (`elsetrec`) struct will be necessary in each thread. The `sgp4` propagator function makes modifications to...

> I had not seen Wagner's implementation, and do not know its history. Reading through it, the variable names and so forth are different enough that it may be an...

> Good point, cython is easier to understand! But in this case I could find no equivalent for PyMemberDef, where a simple table of types and offsets automatically makes Python...