On Sun, Nov 3, 2024 at 4:05 AM Claude Warren <cla...@xenei.com> wrote:
> I believe I answered this elsewhere but just for clarity in the archive. > > I believe the MultidimensionalBloomFilters add a new level of complexity > that we do not want to address at this time. I also believe that > MultidimensionalBloomFilters will not break binary compatibility. They > should extend current functionality. > OK, thanks Claude, we can leave this for later then, if ever ;-) I want to release 4.5.0-M3 very soon. Gary > > On Sun, Oct 6, 2024 at 4:56 PM Gary Gregory <garydgreg...@gmail.com> > wrote: > > > Thank you Claude, the git master code now throws instead of returning > null. > > > > We can delay MultidimensionalBloomFilter unless introducing it later > would > > break binary compatibility. How beneficial would introducing this > interface > > for users? > > > > Gary > > > > > > On Sun, Oct 6, 2024, 10:58 AM Claude Warren <cla...@xenei.com> wrote: > > > > > This is starting tondelve into the realm of multidimensional bloom > > filters. > > > > > > I think throwing an exception on a null or zero length array is > > acceptable. > > > > > > We could define an interface MultidimensionalBloomFilter that could be > > > implemented by any Bloom filter that comprises multiple filters and > givr > > it > > > the method flatten(). But this is a can of worms I would like to keep > > > sealed for a bit longer. > > > > > > Claude > > > > > > On Sat 5 Oct 2024, 20:44 Gary Gregory, <garydgreg...@gmail.com> wrote: > > > > > > > On Fri, Oct 4, 2024 at 1:30 PM Alex Herbert < > alex.d.herb...@gmail.com> > > > > wrote: > > > > > > > > > You cannot return an empty BloomFilter when there is no array of > > > > > BloomFilters to flatten. The BloomFilter is tied to a Shape. > Without > > a > > > > > shape you cannot know what size filter to create. So we have to > > return > > > > > null or throw an exception. Whatever choice must be documented. > > > > > > > > > > > > > The code now throws an exception instead of returning null. I also > > > updated > > > > the Javadoc. > > > > > > > > > > > > > > > > > > Note we do not currently check if the Shapes are all equal when > > > > > creating the BloomFilterExtractor. This is not essential but when > > > > > merging two Bloom filters in flatten an error will be thrown if the > > > > > filters are not compatible. I say it is not essential as the > > > > > functionality of the BloomFilterExtract in processBloomFilters does > > > > > not mandate any prerequisites on the Predicate for the filters. So > > for > > > > > example you could have many filters of different sizes and a > > predicate > > > > > that tests if they all contain a given IndexProducer. > > > > > > > > > > > > > I'll leave that deeper topic alone for now, and for you and Claude to > > > > further discuss if warranted. > > > > > > > > TY, > > > > Gary > > > > > > > > > > > > > > > > > > Alex > > > > > > > > > > On Fri, 4 Oct 2024 at 18:02, Gary Gregory <garydgreg...@gmail.com> > > > > wrote: > > > > > > > > > > > > OK, but how do you create an empty filter? > > > > > > > > > > > > Gary > > > > > > > > > > > > On Fri, Oct 4, 2024, 12:54 PM Claude Warren <cla...@xenei.com> > > > wrote: > > > > > > > > > > > > > I think that it should return an empty filter as the javadoc > says > > > > > > > > > > > > > > > > > > > > > On Fri 4 Oct 2024, 15:53 Gary D. Gregory, <ggreg...@apache.org > > > > > > wrote: > > > > > > > > > > > > > > > See the new disabled test BloomFilterExtractorTest. > > > > > > > > > > > > > > > > Gary > > > > > > > > > > > > > > > > On 2024/10/04 14:51:55 "Gary D. Gregory" wrote: > > > > > > > > > Hi Claude and all, > > > > > > > > > > > > > > > > > > The method: > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > org.apache.commons.collections4.bloomfilter.BloomFilterExtractor.flatten() > > > > > > > > > > > > > > > > > > is documented as always returning a BloomFilter: > > > > > > > > > > > > > > > > > > /** > > > > > > > > > * Create a standard (non-layered) Bloom filter by > > merging > > > > all > > > > > of > > > > > > > > the layers. If > > > > > > > > > * the filter is empty this method will return an empty > > > Bloom > > > > > > > filter. > > > > > > > > > * > > > > > > > > > * @return the merged bloom filter. > > > > > > > > > */ > > > > > > > > > > > > > > > > > > But that's not how it's coded, it can return null, the > > simplest > > > > > > > > reproducer is: > > > > > > > > > > > > > > > > > > BloomFilterExtractor.fromBloomFilterArray(new > > > > > BloomFilter[0]).flatten() > > > > > > > > -> null > > > > > > > > > > > > > > > > > > Should we: > > > > > > > > > > > > > > > > > > - Change the Javadoc > > > > > > > > > - Change the code > > > > > > > > > -- How should this code be changed? > > > > > > > > > -- Should the method be moved to BloomFilter and use "this" > > as > > > > the > > > > > > > > default filter? > > > > > > > > > > > > > > > > > > TY, > > > > > > > > > Gary > > > > > > > > > > > > > > > > > > > > > > > > --------------------------------------------------------------------- > > > > > > > > > To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org > > > > > > > > > For additional commands, e-mail: > dev-h...@commons.apache.org > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > --------------------------------------------------------------------- > > > > > > > > To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org > > > > > > > > For additional commands, e-mail: dev-h...@commons.apache.org > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > --------------------------------------------------------------------- > > > > > To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org > > > > > For additional commands, e-mail: dev-h...@commons.apache.org > > > > > > > > > > > > > > > > > > > > > > -- > LinkedIn: http://www.linkedin.com/in/claudewarren >
