perf: more combine speed-ups

By avoiding writing metadata that differs but doesn't change the data,
we get a higher hitrate on the hash-checking when combining.

Use --debug=process to include these details for debugging.

5 files changed, 42 insertions, 13 deletions

diff --git a/CHANGES.rst b/CHANGES.rst
index 5e3fb44c..00aa97f0 100644
--- a/CHANGES.rst
+++ b/CHANGES.rst
@@ -31,8 +31,8 @@ Unreleased
 
 - Combining data files with ``coverage combine`` now quickly hashes the data
   files to skip files that provide no new information.  This can reduce the
-  time needed.  For coverage.py's own test suite, combining was about 17%
-  faster.
+  time needed.  Many details affect the results, but for coverage.py's own test
+  suite, combining was about 40% faster.
 
 - When searching for completely unexecuted files, coverage.py uses the presence
   of ``__init__.py`` files to determine which directories have source that
@@ -51,7 +51,7 @@ Unreleased
 - Fixed a mis-measurement of a strange use of wildcard alternatives in
   match/case statements, closing `issue 1421`_.
 
-- The ``[run] note`` setting has been completely removed.
+- The deprecated ``[run] note`` setting has been completely removed.
 
 .. _implicit namespace packages: https://peps.python.org/pep-0420/
 .. _issue 1383: https://github.com/nedbat/coveragepy/issues/1383
diff --git a/coverage/sqldata.py b/coverage/sqldata.py
index 2fbc53f5..ea6b1199 100644
--- a/coverage/sqldata.py
+++ b/coverage/sqldata.py
@@ -4,6 +4,7 @@
 """SQLite coverage data."""
 
 import collections
+import datetime
 import functools
 import glob
 import itertools
@@ -51,10 +52,11 @@ CREATE TABLE meta (
     key text,
     value text,
     unique (key)
-    -- Keys:
+    -- Possible keys:
     --  'has_arcs' boolean      -- Is this data recording branches?
     --  'sys_argv' text         -- The coverage command line that recorded the data.
     --  'version' text          -- The version of coverage.py that made the file.
+    --  'when' text             -- Datetime when the file was created.
 );
 
 CREATE TABLE file (
@@ -298,13 +300,18 @@ class CoverageData(SimpleReprMixin):
             self._debug.write(f"Initing data file {self._filename!r}")
         db.executescript(SCHEMA)
         db.execute("insert into coverage_schema (version) values (?)", (SCHEMA_VERSION,))
-        db.executemany(
-            "insert or ignore into meta (key, value) values (?, ?)",
-            [
+
+        # When writing metadata, avoid information that will needlessly change
+        # the hash of the data file, unless we're debugging processes.
+        meta_data = [
+            ("version", __version__),
+        ]
+        if self._debug.should("process"):
+            meta_data.extend([
                 ("sys_argv", str(getattr(sys, "argv", None))),
-                ("version", __version__),
-            ]
-        )
+                ("when", datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")),
+            ])
+        db.executemany("insert or ignore into meta (key, value) values (?, ?)", meta_data)
 
     def _connect(self):
         """Get the SqliteDb object to use."""
diff --git a/doc/cmd.rst b/doc/cmd.rst
index be323c0c..919b6d88 100644
--- a/doc/cmd.rst
+++ b/doc/cmd.rst
@@ -1025,7 +1025,8 @@ of operation to log:
 * ``plugin``: print information about plugin operations.
 
 * ``process``: show process creation information, and changes in the current
-  directory.
+  directory.  This also writes a timestamp and command arguments into the data
+  file.
 
 * ``pybehave``: show the values of `internal flags <env.py_>`_ describing the
   behavior of the current version of Python.
diff --git a/doc/dbschema.rst b/doc/dbschema.rst
index 42e616d9..b576acaa 100644
--- a/doc/dbschema.rst
+++ b/doc/dbschema.rst
@@ -66,10 +66,11 @@ This is the database schema:
         key text,
         value text,
         unique (key)
-        -- Keys:
+        -- Possible keys:
         --  'has_arcs' boolean      -- Is this data recording branches?
         --  'sys_argv' text         -- The coverage command line that recorded the data.
         --  'version' text          -- The version of coverage.py that made the file.
+        --  'when' text             -- Datetime when the file was created.
     );
 
     CREATE TABLE file (
@@ -115,7 +116,7 @@ This is the database schema:
         foreign key (file_id) references file (id)
     );
 
-.. [[[end]]] (checksum: 9d87794485a9aa6d9064b735972a3447)
+.. [[[end]]] (checksum: 6a04d14b07f08f86cccf43056328dcb7)
 
 
 .. _numbits:
diff --git a/tests/test_data.py b/tests/test_data.py
index b1a215e2..79c90420 100644
--- a/tests/test_data.py
+++ b/tests/test_data.py
@@ -904,6 +904,26 @@ class CoverageDataFilesTest(CoverageTest):
         assert_measured_files(covdata3, MEASURED_FILES_1_2)
         self.assert_file_count(glob.escape(basename) + ".*", 0)
 
+    def test_meta_data(self):
+        # The metadata written to the data file shouldn't interfere with
+        # hashing to remove duplicates, except for debug=process, which
+        # writes debugging info as metadata.
+        debug = DebugControlString(options=[])
+        covdata1 = CoverageData(basename="meta.1", debug=debug)
+        covdata1.add_lines(LINES_1)
+        covdata1.write()
+        with sqlite3.connect("meta.1") as con:
+            data = sorted(k for (k,) in con.execute("select key from meta"))
+        assert data == ["has_arcs", "version"]
+
+        debug = DebugControlString(options=["process"])
+        covdata2 = CoverageData(basename="meta.2", debug=debug)
+        covdata2.add_lines(LINES_1)
+        covdata2.write()
+        with sqlite3.connect("meta.2") as con:
+            data = sorted(k for (k,) in con.execute("select key from meta"))
+        assert data == ["has_arcs", "sys_argv", "version", "when"]
+
 
 class DumpsLoadsTest(CoverageTest):
     """Tests of CoverageData.dumps and loads."""