addthis / stream-lib

Stream summarizer and cardinality estimator.
Apache License 2.0
2.26k stars 558 forks source link

Provide information on tuning/size parameters #146

Open thecoop opened 6 years ago

thecoop commented 6 years ago

There are several classes which require sizing or tuning parameters on construction (for example, HyperLogLog has got rsd or log2m parameters, CountMinSketch has got depth, width, and seed). There's no information on what those parameters do, and what you set them to get certain behaviours.

It would be really useful to have a table for each provided type, with information on the resulting size of the datastructures, and how many items they can be expected to scale to before becoming inaccurate. Even very rough guidelines would be useful; currently there's no information on what behaviour you can expect.