Here you can find the full developer API for the pyprobables project. pyprobables provides a suite of probabilistic data-structures to be used in data analytics and data science projects.
Bloom Filters are a class of probabilistic data structures used for set operations. Bloom Filters guarantee a zero percent false negative rate and a predetermined false positive rate. Once the number of elements inserted exceeds the estimated elements, the false positive rate will increase over the desired amount.
probables.BloomFilter
probables.BloomFilterOnDisk
For more information of all methods and properties, see BloomFilter.
probables.ExpandingBloomFilter
probables.RotatingBloomFilter
probables.CountingBloomFilter
Cuckoo filters are a space efficient data structure that supports set membership testing. Cuckoo filters support insertion, deletion, and lookup of elements with low overhead and few false positive results. The name is derived from the cuckoo hashing strategy used to resolve conflicts.
probables.CuckooFilter
probables.CountingCuckooFilter
Count-Min Sketches, and its derivatives, are good for estimating the number of occurrences of an element in streaming data while not needing to retain all the data elements. The result is a probabilistic count of elements inserted into the data structure. It will always provide the maximum number of times a data element was encountered. Notice that the result may be more than the true number of times it was inserted, but never fewer.
probables.CountMinSketch
probables.CountMeanSketch
For more information of all methods and properties, see CountMinSketch.
probables.CountMeanMinSketch
For more information of all methods and properties, see CountMinSketch.
probables.HeavyHitters
For more information of all methods and properties, see CountMinSketch.
probables.StreamThreshold
For more information of all methods and properties, see CountMinSketch.
Quotient filters are an aproximate membership query filter (AMQ) that is both space efficient and returns a zero false negative rate and a probablistic false positive rate. Unlike Bloom filters, the quotient filter only requires a single hash of the element to insert. The upper q bits denote the location within the filter while the lower r bits are stored in the filter.
Quotient filters provide some useful benifits over Bloom filters including:
- Merging of two filters (not union)
- Resizing of the filter
- Ability to remove elements
probables.QuotientFilter
probables.utilities.Bitarray
probables.exceptions
probables.hashes
homequickstartgenindexmodindexsearch