Use code files for testing #5

Signed-off-by: Jono Yang <jyang@nexb.com>

src/matchcode_toolkit/fingerprinting.py

@@ -10,10 +10,6 @@
 import binascii
 import re
 
-from commoncode import filetype
-from licensedcode.tokenize import ngrams
-from typecode.contenttype import get_type
-
 from matchcode_toolkit.halohash import BitAverageHaloHash
 
 
@@ -211,6 +207,10 @@ def get_file_fingerprint_hashes(location, ngram_length=8, **kwargs):
 
     Return an empty mapping if `location` is not a text file
     """
+    from commoncode import filetype
+    from licensedcode.tokenize import ngrams
+    from typecode.contenttype import get_type
+
     # Do not process `location` if it's not a text file
     ft = get_type(location)
     if not (filetype.is_file(location) and ft.is_text):

tests/test_fingerprinting.py

@@ -20,6 +20,7 @@
 from matchcode_toolkit.fingerprinting import create_structure_fingerprint
 from matchcode_toolkit.fingerprinting import get_file_fingerprint_hashes
 from matchcode_toolkit.fingerprinting import split_fingerprint
+from matchcode_toolkit.halohash import byte_hamming_distance
 
 
 class Resource():
@@ -126,16 +127,28 @@ def test_do_not_compute_fingerprint_for_empty_dirs(self):
         self.assertEqual({}, empty_dir_2.extra_data)
         self.assertEqual({}, empty_dir_2.extra_data)
 
-    def test_get_file_fingerprint_hashes_same_halo1_with_exact_repetition(self):
-        """
-        The test files contain the string 'z1 z2 z3 z4 z5 z6 z7 z8' repeated
-        some number of times.
-
-        In the case of test_file1, the string is repeated twice in a row. In the
-        case of test_file2, the string is repeated 8 times.
-        """
-        test_file1 = self.get_test_loc('16words')
-        test_file2 = self.get_test_loc('64words')
+    def test_get_file_fingerprint_hashes_one_line_removed(self):
+        test_file1 = self.get_test_loc('inflate.c')
+        test_file2 = self.get_test_loc('inflate-mod.c')
+        result1 = get_file_fingerprint_hashes(test_file1)
+        result2 = get_file_fingerprint_hashes(test_file2)
+        result1 = result1.get('halo1')
+        result2 = result2.get('halo1')
+        expected_result1 = 'a23a49e4cd40718d1297be719e6564a4'
+        expected_result2 = 'aa3a49e4cd40718d1297be519e6564a4'
+        assert result1 == expected_result1
+        assert result2 == expected_result2
+        assert byte_hamming_distance(result1, result2) == 2
+
+    def test_get_file_fingerprint_hashes_one_line_added(self):
+        test_file1 = self.get_test_loc('inflate.c')
+        test_file2 = self.get_test_loc('inflate-mod2.c')
         result1 = get_file_fingerprint_hashes(test_file1)
         result2 = get_file_fingerprint_hashes(test_file2)
-        assert result1.get('halo1') == result2.get('halo1')
+        result1 = result1.get('halo1')
+        result2 = result2.get('halo1')
+        expected_result1 = 'a23a49e4cd40718d1297be719e6564a4'
+        expected_result2 = 'a23b49e4cd40708d1297be719c6564a4'
+        assert result1 == expected_result1
+        assert result2 == expected_result2
+        assert byte_hamming_distance(result1, result2) == 3