From 5c0ecf89637ed67d1c6a94de7c5bf0ddf69faf2c Mon Sep 17 00:00:00 2001
From: Al <albarrentine@gmail.com>
Date: Sat, 21 Oct 2017 10:34:12 -0400
Subject: [PATCH] [dedupe] Jaccard similarity

---
 src/jaccard.c | 27 +++++++++++++++++++++++++++
 src/jaccard.h | 11 +++++++++++
 2 files changed, 38 insertions(+)
 create mode 100644 src/jaccard.c
 create mode 100644 src/jaccard.h

diff --git a/src/jaccard.c b/src/jaccard.c
new file mode 100644
index 00000000..1f96c61f
--- /dev/null
+++ b/src/jaccard.c
@@ -0,0 +1,27 @@
+#include "jaccard.h"
+
+
+double jaccard_similarity(khash_t(str_set) *s1, khash_t(str_set) *s2) {
+    if (s1 == NULL || s2 == NULL) return -1.0;
+
+    size_t set_intersection = 0;
+    size_t set_union = 0;
+
+    khiter_t k;
+    const char *key;
+
+    kh_foreach_key(s1, key, {
+        k = kh_get(str_set, s2, key);
+        if (k != kh_end(s2)) {
+            set_intersection++;
+        } else {
+            set_union++;
+        }
+    });
+
+    // set_union contains all the keys that were in s1 but not s2
+    // so just add all the keys in s2 to complete the union
+    set_union += kh_size(s2);
+
+    return (double)set_intersection / set_union;
+}
\ No newline at end of file
diff --git a/src/jaccard.h b/src/jaccard.h
new file mode 100644
index 00000000..a6468078
--- /dev/null
+++ b/src/jaccard.h
@@ -0,0 +1,11 @@
+#ifndef JACCARD_H
+#define JACCARD_H
+
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "collections.h"
+
+double jaccard_similarity(khash_t(str_set) *s1, khash_t(str_set) *s2);
+
+#endif
\ No newline at end of file