Avoid useless or redundant db operations. Faster.

nedbat · nedbat · commit dade9db58673 · 2019-07-07T13:39:13.000-04:00
Moving operations into the "with self._connect" means less opening and
closing of the database.  Returning early if there is no data to write
avoids writing empty contexts.
diff --git a/CHANGES.rst b/CHANGES.rst
@@ -32,6 +32,8 @@ Unreleased
     information to each covered line.  Hovering over the "ctx" marker at the
     end of the line reveals a list of the contexts that covered the line.
 
+- Dynamic contexts with no data are no longer written to the database.
+
 - Added the classmethod :meth:`Coverage.current` to get the latest started
   `Coverage` instance.
 
diff --git a/coverage/control.py b/coverage/control.py
@@ -648,8 +648,9 @@ def _post_save_work(self):
             self._warn("No data was collected.", slug="no-data-collected")
 
         # Find files that were never executed at all.
-        for file_path, plugin_name in self._inorout.find_unexecuted_files():
-            self._data.touch_file(file_path, plugin_name)
+        if self._data:
+            for file_path, plugin_name in self._inorout.find_unexecuted_files():
+                self._data.touch_file(file_path, plugin_name)
 
         if self.config.note:
             self._data.add_run_info(note=self.config.note)
diff --git a/coverage/sqldata.py b/coverage/sqldata.py
@@ -255,8 +255,10 @@ def add_lines(self, line_data):
             ))
         self._start_using()
         self._choose_lines_or_arcs(lines=True)
-        self._set_context_id()
+        if not line_data:
+            return
         with self._connect() as con:
+            self._set_context_id()
             for filename, linenos in iitems(line_data):
                 file_id = self._file_id(filename, add=True)
                 data = [(file_id, self._current_context_id, lineno) for lineno in linenos]
@@ -279,8 +281,10 @@ def add_arcs(self, arc_data):
             ))
         self._start_using()
         self._choose_lines_or_arcs(arcs=True)
-        self._set_context_id()
+        if not arc_data:
+            return
         with self._connect() as con:
+            self._set_context_id()
             for filename, arcs in iitems(arc_data):
                 file_id = self._file_id(filename, add=True)
                 data = [(file_id, self._current_context_id, fromno, tono) for fromno, tono in arcs]
@@ -307,6 +311,10 @@ def add_file_tracers(self, file_tracers):
         `file_tracers` is { filename: plugin_name, ... }
 
         """
+        if self._debug.should('dataop'):
+            self._debug.write("Adding file tracers: %d files" % (len(file_tracers),))
+        if not file_tracers:
+            return
         self._start_using()
         with self._connect() as con:
             for filename, plugin_name in iitems(file_tracers):
@@ -667,6 +675,12 @@ def run_infos(self):
 
 
 class SqliteDb(SimpleReprMixin):
+    """A simple abstraction over a SQLite database.
+
+    Use as a context manager to get an object you can call
+    execute or executemany on.
+
+    """
     def __init__(self, filename, debug):
         self.debug = debug if debug.should('sql') else None
         self.filename = filename
@@ -679,7 +693,7 @@ def connect(self):
         filename = os.path.relpath(self.filename)
         # It can happen that Python switches threads while the tracer writes
         # data. The second thread will also try to write to the data,
-        # effectively causing a nested context. However, given the indempotent
+        # effectively causing a nested context. However, given the idempotent
         # nature of the tracer operations, sharing a conenction among threads
         # is not a problem.
         if self.debug:
diff --git a/tests/test_api.py b/tests/test_api.py
@@ -326,21 +326,6 @@ def test_two_getdata_only_warn_once(self):
         with self.assert_warnings(cov, []):
             cov.get_data()
 
-    def test_two_getdata_only_warn_once_nostop(self):
-        self.make_code1_code2()
-        cov = coverage.Coverage(source=["."], omit=["code1.py"])
-        cov.start()
-        import_local_file("code1")                                     # pragma: nested
-        # We didn't collect any data, so we should get a warning.
-        with self.assert_warnings(cov, ["No data was collected"]):     # pragma: nested
-            cov.get_data()                                             # pragma: nested
-        # But calling get_data a second time with no intervening activity
-        # won't make another warning.
-        with self.assert_warnings(cov, []):                            # pragma: nested
-            cov.get_data()                                             # pragma: nested
-        # Then stop it, or the test suite gets out of whack.
-        cov.stop()                                                     # pragma: nested
-
     def test_two_getdata_warn_twice(self):
         self.make_code1_code2()
         cov = coverage.Coverage(source=["."], omit=["code1.py", "code2.py"])
diff --git a/tests/test_debug.py b/tests/test_debug.py
@@ -136,16 +136,18 @@ def test_debug_callers(self):
         frames = re_lines(out_lines, frame_pattern).splitlines()
         self.assertEqual(len(real_messages), len(frames))
 
-        # The last message should be "Writing data", and the last frame should
-        # be _write_file in data.py.
         last_line = out_lines.splitlines()[-1]
+
+        # The details of what to expect on the stack are empirical, and can change
+        # as the code changes. This test is here to ensure that the debug code
+        # continues working. It's ok to adjust these details over time.
         from coverage.data import STORAGE
         if STORAGE == "json":
             self.assertRegex(real_messages[-1], r"^\s*\d+\.\w{4}: Writing data")
             self.assertRegex(last_line, r"\s+_write_file : .*coverage[/\\]data.py @\d+$")
         else:
-            self.assertRegex(real_messages[-1], r"^\s*\d+\.\w{4}: Creating data file")
-            self.assertRegex(last_line, r"\s+_create_db : .*coverage[/\\]sqldata.py @\d+$")
+            self.assertRegex(real_messages[-1], r"^\s*\d+\.\w{4}: Adding file tracers: 0 files")
+            self.assertRegex(last_line, r"\s+add_file_tracers : .*coverage[/\\]sqldata.py @\d+$")
 
     def test_debug_config(self):
         out_lines = self.f1_debug_output(["config"])
