Load in algorithms from external module

[carl-ellis]

This means you can have some business critical logic in a separate module, while still being able to maintain a public fork of skyline.

I have some functionality to define custom_analysers too (have specific algorithms to run on specific metrics) which I can make a pull request for too. It is all defined in the settings file, but I'm not sure if that's the direction that you want Skyline to go in.

[astanway]

Wondering - why would you need this as opposed to contributing algorithms to the public repo? Secret algorithms?

[astanway]

I'm not really sold.

This strikes me as overly complex. The algorithms logic as it is is very modular and easy to extend. What do you mean by "specific" and "plugins"?

On May 12, 2014, at 4:31 AM, Omer Katz [email protected] wrote:

@astanway Some might be proprietary, some might not fit the main repo as they are too specific. Some will just be installable as Plugins.

—
Reply to this email directly or view it on GitHub.

[hit9]

I think the ability to load algorithms from custom module is worth to be supported.

Some widely used algorithms are built-in, and some should be custom.

By the way, I think we can reuse many statistics (such as series.mean(), series.median) between algorithms, this will speed up skyline, well, this point is irrelevant to this topic...

[astanway]

Sure, but that functionality is already in Skyline - you can pick and choose the algos you want in settings.py

On May 13, 2014, at 1:30 AM, Omer Katz [email protected] wrote:

Specific as in relevant to a very narrow set of metrics (or even one). Some algorithms might not be useful to every skyline user but might be useful for someone, thus it should be able to used as a plugin.

—
Reply to this email directly or view it on GitHub.