Organize into databases

[Jayd-1234]

Implementation of pairs function in numpy with triu_indices.

Returns a List by default. Can be changed to return a List of Tuples though through switch?

Caveats

• Tuple creation will unfortunately, require list comprehension. Possible numpy alternative is numba or numpy.ndarray.view. But it will require assumptions about dtype, while being potentially slower. Considering for future work.
• Currently returns double precision List or Tuple. Maybe have a way of inferring dtype from input List, and return same output dtype.
• Another possible speedup can be achieved, if numpy arrays are fed into the function. The current implementation requires converting the List or Tuple data to numpy array for processing, which is slow.

Notes

• Records may not be possible to add, as they will require unique (key, value) pairs, which is not feasible in this case. But of course, feel free to suggest a way to do it.

[jpivarski]

You closed it? I really was going to look at it! Are you doing this because you want to redesign? That would be okay. (Sorry that I haven't had time to look at it just yet...)

[Jayd-1234]

I closed it because I wanted the current operations.py which is under heavy development to be a bit stable. It function still works though, I can create a PR if you want.

[jpivarski]

operations.py is unstable. I'm actively working on it. That's why I suggested that you create a gist of the concise idea, rather than a fully working fork.

I think you updated operations.py by copying the file, which would be counted by git as a conflict.

Don't worry about making it production-quality. Remember that your project is about algorithm development, not packaging. Showing me the idea concisely, without all the caveats that make it production-quality, would also make it easier for me to review.

[Jayd-1234]

Thanks. I will keep that in mind from now on. :smiley: