fix: generate flat file names differently

Fixes a few unusual issues with reports:

- #580: HTML report generation fails on too long path
- #584: File collisions in coverage report html
- #1167: Remove leading underscore in coverage html

CHANGES.rst

@@ -46,11 +46,20 @@ Unreleased
 
 - TOML parsing now uses the `tomli`_ library.
 
-- Use a modern hash algorithm when fingerprinting to speed HTML reports
-  (`issue 1189`_).
+- Some minor changes to usually invisible details of the HTML report:
+
+  - Use a modern hash algorithm when fingerprinting, for high-security
+    environments (`issue 1189`_).
+
+  - Change how report file names are generated, to avoid leading underscores
+    (`issue 1167`_), to avoid rare file name collisions (`issue 584`_), and to
+    avoid file names becoming too long (`issue 580`_).
 
 .. _Django coverage plugin: https://pypi.org/project/django-coverage-plugin/
+.. _issue 580: https://github.com/nedbat/coveragepy/issues/580
+.. _issue 584: https://github.com/nedbat/coveragepy/issues/584
 .. _issue 1150: https://github.com/nedbat/coveragepy/issues/1150
+.. _issue 1167: https://github.com/nedbat/coveragepy/issues/1167
 .. _issue 1168: https://github.com/nedbat/coveragepy/issues/1168
 .. _issue 1189: https://github.com/nedbat/coveragepy/issues/1189
 .. _tomli: https://pypi.org/project/tomli/

coverage/files.py

@@ -77,7 +77,7 @@ def canonical_filename(filename):
     return CANONICAL_FILENAME_CACHE[filename]
 
 
-MAX_FLAT = 200
+MAX_FLAT = 100
 
 @contract(filename='unicode', returns='unicode')
 def flat_rootname(filename):
@@ -90,11 +90,9 @@ def flat_rootname(filename):
     For example, the file a/b/c.py will return 'a_b_c_py'
 
     """
-    name = ntpath.splitdrive(filename)[1]
-    name = re.sub(r"[\\/.:]", "_", name)
-    if len(name) > MAX_FLAT:
-        h = hashlib.sha1(name.encode('UTF-8')).hexdigest()
-        name = name[-(MAX_FLAT-len(h)-1):] + '_' + h
+    fp = hashlib.new("sha3_256", filename.encode("UTF-8")).hexdigest()
+    name = re.sub(r"[\\/.:]", "_", filename)
+    name = fp[:16] + "_" + name[-(MAX_FLAT-17):]
     return name
 
 

tests/test_files.py

@@ -69,18 +69,20 @@ def test_canonical_filename_ensure_cache_hit(self):
 
 
 @pytest.mark.parametrize("original, flat", [
-    ("a/b/c.py", "a_b_c_py"),
-    (r"c:\foo\bar.html", "_foo_bar_html"),
-    ("Montréal/☺/conf.py", "Montréal_☺_conf_py"),
+    ("a/b/c.py", "9740bf4b0a33df83_a_b_c_py"),
+    ("/a/b/c.py", "055993932a538df4__a_b_c_py"),
+    (r"c:\foo\bar.html", "e402c17adf14f7d1_c__foo_bar_html"),
+    (r"d:\foo\bar.html", "85b26652b26cac26_d__foo_bar_html"),
+    ("Montréal/☺/conf.py", "1c7e19adc3b6fe6b_Montréal_☺_conf_py"),
     ( # original:
-        r"c:\lorem\ipsum\quia\dolor\sit\amet\consectetur\adipisci\velit\sed\quia\non"
-        r"\numquam\eius\modi\tempora\incidunt\ut\labore\et\dolore\magnam\aliquam"
-        r"\quaerat\voluptatem\ut\enim\ad\minima\veniam\quis\nostrum\exercitationem"
-        r"\ullam\corporis\suscipit\laboriosam\Montréal\☺\my_program.py",
+        r"c:\lorem\ipsum\quia\dolor\sit\amet\consectetur\adipisci\velit\sed" +
+        r"\quia\non\numquam\eius\modi\tempora\incidunt\ut\labore\et\dolore" +
+        r"\magnam\aliquam\quaerat\voluptatem\ut\enim\ad\minima\veniam\quis" +
+        r"\nostrum\exercitationem\ullam\corporis\suscipit\laboriosam" +
+        r"\Montréal\☺\my_program.py",
         # flat:
-        "re_et_dolore_magnam_aliquam_quaerat_voluptatem_ut_enim_ad_minima_veniam_quis_"
-        "nostrum_exercitationem_ullam_corporis_suscipit_laboriosam_Montréal_☺_my_program_py_"
-        "97eaca41b860faaa1a21349b1f3009bb061cf0a8"
+        "0e99262918baa54c__nostrum_exercitationem_ullam_corporis_suscipit_" +
+        "laboriosam_Montréal_☺_my_program_py"
      ),
 ])
 def test_flat_rootname(original, flat):