From 173ab0ea96b4969b51f4d23f033a45242fe7e80a Mon Sep 17 00:00:00 2001
From: Isidro Arias <isidroariass@hotmail.es>
Date: Thu, 6 Apr 2023 10:59:59 +0200
Subject: [PATCH] Bloom filter with tests

---
 data_structures/hashing/bloom_filter.py | 103 ++++++++++++++++++++++++
 1 file changed, 103 insertions(+)
 create mode 100644 data_structures/hashing/bloom_filter.py

diff --git a/data_structures/hashing/bloom_filter.py b/data_structures/hashing/bloom_filter.py
new file mode 100644
index 000000000..c34dd4eea
--- /dev/null
+++ b/data_structures/hashing/bloom_filter.py
@@ -0,0 +1,103 @@
+"""
+See https://en.wikipedia.org/wiki/Bloom_filter
+"""
+from hashlib import sha256, md5
+from random import randint, choices
+import string
+
+
+class Bloom:
+    def __init__(self, size=8):
+        self.bitstring = 0b0
+        self.size = size
+
+    def add(self, value):
+        h = self.hash(value)
+        self.bitstring |= h
+        print(
+            f"""\
+[add] value =      {value}
+      hash =       {self.format_bin(h)}
+      filter =     {self.format_bin(self.bitstring)}
+"""
+        )
+
+    def exists(self, value):
+        h = self.hash(value)
+        res = (h & self.bitstring) == h
+
+        print(
+            f"""\
+[exists] value =   {value}
+         hash =    {self.format_bin(h)}
+         filter =  {self.format_bin(self.bitstring)}
+         res =     {res}
+"""
+        )
+        return res
+
+    def format_bin(self, value):
+        res = bin(value)[2:]
+        return res.zfill(self.size)
+
+    def hash(self, value):
+        res = 0b0
+        for func in (sha256, md5):
+            b = func(value.encode()).digest()
+            position = int.from_bytes(b, "little") % self.size
+            res |= 2**position
+        return res
+
+
+def test_movies():
+    b = Bloom()
+    b.add("titanic")
+    b.add("avatar")
+
+    assert b.exists("titanic")
+    assert b.exists("avatar")
+
+    assert b.exists("the goodfather") in (True, False)
+    assert b.exists("interstellar") in (True, False)
+    assert b.exists("Parasite") in (True, False)
+    assert b.exists("Pulp fiction") in (True, False)
+
+
+def random_string(size):
+    return "".join(choices(string.ascii_lowercase + " ", k=size))
+
+
+def test_probability(m=64, n=20):
+    b = Bloom(size=m)
+
+    added = {random_string(10) for i in range(n)}
+    for a in added:
+        b.add(a)
+
+    # number of hash functions is fixed
+    k = 2
+
+    n_ones = bin(b.bitstring).count("1")
+    expected_probability = (n_ones / m) ** k
+
+    expected_probability_wikipedia = (1 - (1 - 1 / m) ** (k * n)) ** k
+
+    not_added = {random_string(10) for i in range(1000)}
+    fails = 0
+    for string in not_added:
+        if b.exists(string):
+            fails += 1
+    fail_rate = fails / len(not_added)
+
+    print(f"total = {len(not_added)}, fails = {fails}, fail_rate = {fail_rate}")
+    print(f"{expected_probability=}")
+    print(f"{expected_probability_wikipedia=}")
+
+    assert (
+        abs(expected_probability - fail_rate) <= 0.05
+    )  # 5% margin calculated experiementally
+
+
+if __name__ == "__main__":
+    test_movies()
+    test_probability()