From 99fada5b52d4635b45bd8d0fba51baa0158badd8 Mon Sep 17 00:00:00 2001
From: Yohann D'ANELLO <ynerant@crans.org>
Date: Sun, 17 Oct 2021 10:50:50 +0200
Subject: [PATCH] Compute signature matrix using NumPy

---
 algods/algods.py | 58 ++++++++++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 21 deletions(-)

diff --git a/algods/algods.py b/algods/algods.py
index 169f277..4cab606 100644
--- a/algods/algods.py
+++ b/algods/algods.py
@@ -1,8 +1,13 @@
 import argparse
-import re
+import random
 import unicodedata
 import sys
-from typing import Generator
+
+import numpy as np
+
+
+SHINGLE_SIZE = 5  # Known as k
+PERMUTATIONS_COUNT = 3
 
 
 def parse_args(argv: dict = None) -> argparse.Namespace:
@@ -26,21 +31,38 @@ def normalize(doc: str) -> str:
     ).casefold().encode('ascii', 'ignore').decode('ascii')
 
 
-def compute_shingles(docs: list[str], single_size: int) -> Generator[set[int], any, None]:
+def compute_shingles(docs: list[str], single_size: int) -> np.ndarray:
+    shingle_matrix = np.zeros((2, len(docs)))
     shingle_id = {}
-    id_shingle = []
-    ids = 0
 
-    for d in docs:
-        char_shing = [d[i:i + single_size] for i in range(len(d) - single_size + 1)]
-        sid = set()
+    for doc_id, doc in enumerate(docs):
+        char_shing = [doc[i:i + single_size] for i in range(len(doc) - single_size + 1)]
         for sh in char_shing:
             if sh not in shingle_id:
-                shingle_id[sh] = ids
-                id_shingle.append(sh)
-                ids = ids + 1
-            sid.add(shingle_id[sh])
-        yield sid
+                shingle_id[sh] = len(shingle_id)
+                if shingle_id[sh] >= len(shingle_matrix):
+                    # Extend matrix, double its size
+                    shingle_matrix = np.append(shingle_matrix, np.zeros(shingle_matrix.shape), axis=0)
+
+            shingle_matrix[shingle_id[sh], doc_id] = 1
+
+    shingle_matrix = shingle_matrix[:len(shingle_id)]
+
+    return shingle_matrix
+
+
+def min_hash(doc: str, perm: list[str]) -> str:
+    for d in perm:
+        if d in doc:
+            return d
+
+
+def compute_signature_matrix(shingles: np.ndarray, permutations_count: int) -> np.ndarray:
+    permutation_matrix = np.zeros((permutations_count, len(shingles)))
+    for i in range(permutations_count):
+        permutation_matrix[i] = np.random.permutation(len(shingles))
+
+    return permutation_matrix @ shingles
 
 
 def jaccard_similarity(doc1: set, doc2: set) -> float:
@@ -56,14 +78,8 @@ def parse(stream, similarity: float) -> None:
     docs = [line.rstrip('\n') for line in stream]
     docs = [normalize(doc) for doc in docs]  # Remove special characters and normalize accents
 
-    shingles = list(compute_shingles(docs, 5))
-
-    for i, doc1 in enumerate(shingles):
-        for j in range(i + 1, len(shingles)):
-            doc2 = shingles[j]
-            d = jaccard_similarity(doc1, doc2)
-            if d >= similarity:
-                print(f"{i} {j} {d:.06f}")
+    shingles = compute_shingles(docs, SHINGLE_SIZE)
+    signature = compute_signature_matrix(shingles, PERMUTATIONS_COUNT)
 
 
 def main():