Documentation of performance subjects

[WilliamParker]

Currently several significant subjects on performance optimization require knowledge of the Rete algorithm and/or Clara's internals to understand. Examples include

• The benefit of using hash-based joins versus Cartesian joins. This came up recently on the mailing list at https://groups.google.com/forum/#!topic/clara-rules/SbiZA4_3s1U
• The benefit of inserting facts in a batch all at once, rather than one at a time, particularly when using accumulators. I've seen a case where large numbers of facts that were all fed to the same accumulators were inserted one at a time, leading to significant performance penalties, and the problems with this approach aren't obvious IMO if one is not familiar with Clara's implementation.
• The ability to use a custom fact-type-fn to gain constant-time dispatch on particular fields of a fact that are hotspots.

I'd like to make this knowledge more accessible in some way. I'm currently leaning toward adding a "Performance Optimization" page under the Advanced Topics section in the right sidebar and linking to pages on these and other relevant topics from that page but I'm open to suggestions.

[WilliamParker]

One tension I see here is how to make this information available without overwhelming users who don't have such performance concerns with information on it. That is why I'm leaning toward adding a single page on performance that then links to further subpages for now, rather than a bunch of new entries under Advanced Topics.

[rbrush]

A single page under advanced topics, with whatever structure makes sense underneath that, makes sense to me.