I have recently implemented something similar, in Rust, for work. I think there are significant changes around Shape and the Hasher classes.
Shape can be reduced to numberOfBits and numberOfHashFunctions. Hasher can be reduced to 2 Longs. A compound hasher can be created by multiple hashers in a list. The Rust implementation forced an intense look at the architecture and a significant simplification. I think the simplification refactoring should be applied before the Bloom filters are released. I would also note that the current implementation is too slow. It significantly changes the timings of indexes, so much so that I stopped working on a paper because the differences in the index timings were overwhelmed by the change in Bloomfilter performance. Claude On Mon, Aug 30, 2021 at 2:11 PM Alex Herbert <alex.d.herb...@gmail.com> wrote: > Hi Claude, > > I previously raised the issue of moving the Bloom filters to another branch > while still under development. However this would reduce the visibility of > the code and discourage potential contributors. To maintain visibility for > contributors I would suggest that a collections release could avoid > releasing this package. > > IIRC the package was fairly mature in its current form with a few > exceptions such as changes to move to Splitterators in-place of Iterators. > Other discussions we had were about more major reengineering of the type > not possible while maintaining binary compatibility. Is a major > reengineering what you are suggesting? > > Regards, > > Alex > > On Mon, 30 Aug 2021 at 13:37, Claude Warren <cla...@xenei.com> wrote: > > > Greetings, > > > > I see that the Bloom filter implementation has not been released. It > would > > be in V4.5. I have not had time to come back and clean it up as it > should > > be to make is simpler and faster. I am concerned that there may be an > > upcoming release of 4.5 which would lock the implementation and freeze > many > > of the discussions we had before. > > > > Is it possible to remove the Bloom filter impelmentations from > > Commons-Collections. I will work on them when I get some time and > > resubmit. We should also probably have a discussion about whether or not > > collections is the proper place for them. > > > > Claude > > > > -- > > I like: Like Like - The likeliest place on the web > > <http://like-like.xenei.com> > > LinkedIn: http://www.linkedin.com/in/claudewarren > > > -- I like: Like Like - The likeliest place on the web <http://like-like.xenei.com> LinkedIn: http://www.linkedin.com/in/claudewarren
