fset - Fast set/relational library

fset is a faster implementation of the functions in Clojure core.set.

All functions in fset are as compatible as possible with clojure.set. There are minor differences, for example requiring input arguments to actually implement IPersistentSet (while clojure.set also allows other collection types as input with unpredictable results, so this is sort of a feature). The library uses generative tests to verify compatibility with core functions (run with lein test).

The library also includes functions with a different interface to those in clojure.set to further improve speed. These functions are clearly identified by ending with "*". For example, fset contains both index and index*: index is already faster than clojure.set/index and has the same interface. fset/index* is even faster than fset/index but it breaks compatibility with clojure.set/index by returning a mutable java.util.HashMap instead of clojure.lang.PersistentHashMap.

fset also implements other non-set related functions from clojure.core when they are instrumental to improve performance for those in clojure.set (this is the case of fset/select-keys for example).

Usage

Add [fset 0.1.1] to your dependencies and then:

(require '[tech.droit.fset :as fset]
         '[clojure.set :as cset])

(def s1 (set (range 10)))
(def s2 (set (range 20)))

(= (cset/intersection s1 s2)
   (fset/intersection s1 s2))
;; true

Perf showcase

Some comparison with clojure.core/set functions as measured in test/fset/bench.clj with the Criterium library. Smaller numbers mean faster computation. All results below are obtained using small sets (of 10-100 items), but the improvement should be bigger on larger sets.

Set functions

Union (~35% speedup):

2.07E-6 cset
1.33E-6 fset

Intersection (~50% speedup):

6.98E-6 cset
3.64E-6 fset

Difference (~30% speedup):

5.03E-6 cset
3.42E-6 fset

Subset? (~80% speedup):

6.95E-7 cset
1.48E-7 fset

Superset? (~80% speedup):

7.11E-7 cset
1.49E-7 fset

Relation functions

Relation functions operate on sets of maps, where all maps have the same set of keys.

Select (~25% speedup):

2.44E-6 cset
1.85E-6 fset

Project (~50% speedup):

1.03E-5 cset
0.55E-5 fset

Join (~70% speedup):

2.77E-5 cset
0.79E-5 fset

Index (~33% speedup):

1.37E-5 cset
0.91E-5 fset

Other functions

clojure.set/rename-keys (~38% speedup)

3.00E-6 cset
1.88E-6 fset

fset/maps (map over sets). There is no corresponding single function in clojure.set for this, but closest clojure.core expression is shown below (~40% speedup):

5.30E-6 (into #{} (map #(assoc % :new 0) xrel))
3.25E-6 (fset/maps #(assoc % :new 0) xrel)

clojure.core/select-keys has been re-implemented in fset (~ 55% speedup):

2.48E-6 core
1.09E-6 fset (select-keys)

kset is also new: given a relation (a set of maps with the same keys) find the set of keys repeating in each map:

6.31E-7 (set (keys (first rel)))
3.57E-7 (fset/kset rel)

License

Copyright © 2021 Droit Financial Technologies

This program and the accompanying materials are made available under the terms of the Eclipse Public License 2.0 which is available at http://www.eclipse.org/legal/epl-2.0.