feat: multiple --concurrency values. #1012 #1082

Author: nedbat

CHANGES.rst

@@ -22,6 +22,10 @@ This list is detailed and covers changes in each pre-release version.
 Unreleased
 ----------
 
+- Feature: Now the ``--concurrency`` setting can have a list of values, so that
+  threads and another lightweight threading package can be measured together.
+  Closes `issue 1012`_ and `issue 1082`_.
+
 - Fix: A module specified as the ``source`` setting is imported during startup,
   before the user program imports it.  This could cause problems if the rest of
   the program isn't ready yet.  For example, `issue 1203`_ describes a Django
@@ -49,6 +53,8 @@ Unreleased
   works, to allow for command-line options in the future.
 
 .. _issue 989: https://github.com/nedbat/coveragepy/issues/989
+.. _issue 1012: https://github.com/nedbat/coveragepy/issues/1012
+.. _issue 1082: https://github.com/nedbat/coveragepy/issues/1802
 .. _issue 1203: https://github.com/nedbat/coveragepy/issues/1203
 
 

coverage/cmdline.py

@@ -17,6 +17,7 @@
 from coverage import Coverage
 from coverage import env
 from coverage.collector import CTracer
+from coverage.config import CoverageConfig
 from coverage.data import combinable_files, debug_data_file
 from coverage.debug import info_formatter, info_header, short_stack
 from coverage.exceptions import _BaseCoverageException, _ExceptionDuringRun, NoSource
@@ -39,16 +40,12 @@ class Opts:
         '', '--branch', action='store_true',
         help="Measure branch coverage in addition to statement coverage.",
     )
-    CONCURRENCY_CHOICES = [
-        "thread", "gevent", "greenlet", "eventlet", "multiprocessing",
-    ]
     concurrency = optparse.make_option(
-        '', '--concurrency', action='store', metavar="LIB",
-        choices=CONCURRENCY_CHOICES,
+        '', '--concurrency', action='store', metavar="LIBS",
         help=(
             "Properly measure code using a concurrency library. " +
             "Valid values are: {}."
-        ).format(", ".join(CONCURRENCY_CHOICES)),
+        ).format(", ".join(sorted(CoverageConfig.CONCURRENCY_CHOICES))),
     )
     context = optparse.make_option(
         '', '--context', action='store', metavar="LABEL",
@@ -570,6 +567,11 @@ def command_line(self, argv):
         debug = unshell_list(options.debug)
         contexts = unshell_list(options.contexts)
 
+        if options.concurrency is not None:
+            concurrency = options.concurrency.split(",")
+        else:
+            concurrency = None
+
         # Do something.
         self.coverage = Coverage(
             data_suffix=options.parallel_mode,
@@ -581,7 +583,7 @@ def command_line(self, argv):
             omit=omit,
             include=include,
             debug=debug,
-            concurrency=options.concurrency,
+            concurrency=concurrency,
             check_preimported=True,
             context=options.context,
             messages=not options.quiet,

coverage/collector.py

@@ -7,6 +7,7 @@
 import sys
 
 from coverage import env
+from coverage.config import CoverageConfig
 from coverage.debug import short_stack
 from coverage.disposition import FileDisposition
 from coverage.exceptions import ConfigError
@@ -55,7 +56,7 @@ class Collector:
     _collectors = []
 
     # The concurrency settings we support here.
-    SUPPORTED_CONCURRENCIES = {"greenlet", "eventlet", "gevent", "thread"}
+    LIGHT_THREADS = {"greenlet", "eventlet", "gevent"}
 
     def __init__(
         self, should_trace, check_include, should_start_context, file_mapper,
@@ -93,59 +94,28 @@ def __init__(
 
         `concurrency` is a list of strings indicating the concurrency libraries
         in use.  Valid values are "greenlet", "eventlet", "gevent", or "thread"
-        (the default).  Of these four values, only one can be supplied.  Other
-        values are ignored.
+        (the default).  "thread" can be combined with one of the other three.
+        Other values are ignored.
 
         """
         self.should_trace = should_trace
         self.check_include = check_include
         self.should_start_context = should_start_context
         self.file_mapper = file_mapper
-        self.warn = warn
         self.branch = branch
+        self.warn = warn
+        self.concurrency = concurrency
+        assert isinstance(self.concurrency, list), f"Expected a list: {self.concurrency!r}"
+
         self.threading = None
         self.covdata = None
-
         self.static_context = None
 
         self.origin = short_stack()
 
         self.concur_id_func = None
         self.mapped_file_cache = {}
 
-        # We can handle a few concurrency options here, but only one at a time.
-        these_concurrencies = self.SUPPORTED_CONCURRENCIES.intersection(concurrency)
-        if len(these_concurrencies) > 1:
-            raise ConfigError(f"Conflicting concurrency settings: {concurrency}")
-        self.concurrency = these_concurrencies.pop() if these_concurrencies else ''
-
-        try:
-            if self.concurrency == "greenlet":
-                import greenlet
-                self.concur_id_func = greenlet.getcurrent
-            elif self.concurrency == "eventlet":
-                import eventlet.greenthread     # pylint: disable=import-error,useless-suppression
-                self.concur_id_func = eventlet.greenthread.getcurrent
-            elif self.concurrency == "gevent":
-                import gevent                   # pylint: disable=import-error,useless-suppression
-                self.concur_id_func = gevent.getcurrent
-            elif self.concurrency == "thread" or not self.concurrency:
-                # It's important to import threading only if we need it.  If
-                # it's imported early, and the program being measured uses
-                # gevent, then gevent's monkey-patching won't work properly.
-                import threading
-                self.threading = threading
-            else:
-                raise ConfigError(f"Don't understand concurrency={concurrency}")
-        except ImportError as ex:
-            raise ConfigError(
-                "Couldn't trace with concurrency={}, the module isn't installed.".format(
-                    self.concurrency,
-                )
-            ) from ex
-
-        self.reset()
-
         if timid:
             # Being timid: use the simple Python trace function.
             self._trace_class = PyTracer
@@ -163,6 +133,54 @@ def __init__(
             self.supports_plugins = False
             self.packed_arcs = False
 
+        # We can handle a few concurrency options here, but only one at a time.
+        concurrencies = set(self.concurrency)
+        unknown = concurrencies - CoverageConfig.CONCURRENCY_CHOICES
+        if unknown:
+            show = ", ".join(sorted(unknown))
+            raise ConfigError(f"Unknown concurrency choices: {show}")
+        light_threads = concurrencies & self.LIGHT_THREADS
+        if len(light_threads) > 1:
+            show = ", ".join(sorted(light_threads))
+            raise ConfigError(f"Conflicting concurrency settings: {show}")
+        do_threading = False
+
+        try:
+            if "greenlet" in concurrencies:
+                tried = "greenlet"
+                import greenlet
+                self.concur_id_func = greenlet.getcurrent
+            elif "eventlet" in concurrencies:
+                tried = "eventlet"
+                import eventlet.greenthread     # pylint: disable=import-error,useless-suppression
+                self.concur_id_func = eventlet.greenthread.getcurrent
+            elif "gevent" in concurrencies:
+                tried = "gevent"
+                import gevent                   # pylint: disable=import-error,useless-suppression
+                self.concur_id_func = gevent.getcurrent
+
+            if "thread" in concurrencies:
+                do_threading = True
+        except ImportError as ex:
+            msg = f"Couldn't trace with concurrency={tried}, the module isn't installed."
+            raise ConfigError(msg) from ex
+
+        if self.concur_id_func and not hasattr(self._trace_class, "concur_id_func"):
+            raise ConfigError(
+                "Can't support concurrency={} with {}, only threads are supported.".format(
+                    tried, self.tracer_name(),
+                )
+            )
+
+        if do_threading or not concurrencies:
+            # It's important to import threading only if we need it.  If
+            # it's imported early, and the program being measured uses
+            # gevent, then gevent's monkey-patching won't work properly.
+            import threading
+            self.threading = threading
+
+        self.reset()
+
     def __repr__(self):
         return f"<Collector at 0x{id(self):x}: {self.tracer_name()}>"
 
@@ -244,13 +262,6 @@ def _start_tracer(self):
 
         if hasattr(tracer, 'concur_id_func'):
             tracer.concur_id_func = self.concur_id_func
-        elif self.concur_id_func:
-            raise ConfigError(
-                "Can't support concurrency={} with {}, only threads are supported".format(
-                    self.concurrency, self.tracer_name(),
-                )
-            )
-
         if hasattr(tracer, 'file_tracers'):
             tracer.file_tracers = self.file_tracers
         if hasattr(tracer, 'threading'):

coverage/config.py

@@ -334,6 +334,8 @@ def copy(self):
         """Return a copy of the configuration."""
         return copy.deepcopy(self)
 
+    CONCURRENCY_CHOICES = {"thread", "gevent", "greenlet", "eventlet", "multiprocessing"}
+
     CONFIG_FILE_OPTIONS = [
         # These are *args for _set_attr_from_config_option:
         #   (attr, where, type_="")

coverage/control.py

@@ -448,7 +448,7 @@ def load(self):
     def _init_for_start(self):
         """Initialization for start()"""
         # Construct the collector.
-        concurrency = self.config.concurrency or ()
+        concurrency = self.config.concurrency or []
         if "multiprocessing" in concurrency:
             if not patch_multiprocessing:
                 raise ConfigError(                      # pragma: only jython

coverage/version.py

@@ -5,7 +5,7 @@
 # This file is exec'ed in setup.py, don't import anything!
 
 # Same semantics as sys.version_info.
-version_info = (6, 1, 3, "alpha", 0)
+version_info = (6, 2, 0, "alpha", 0)
 
 
 def _make_version(major, minor, micro, releaselevel, serial):

doc/cmd.rst

@@ -124,9 +124,9 @@ There are many options:
                             clean each time.
       --branch              Measure branch coverage in addition to statement
                             coverage.
-      --concurrency=LIB     Properly measure code using a concurrency library.
-                            Valid values are: thread, gevent, greenlet, eventlet,
-                            multiprocessing.
+      --concurrency=LIBS    Properly measure code using a concurrency library.
+                            Valid values are: eventlet, gevent, greenlet,
+                            multiprocessing, thread.
       --context=LABEL       The context label to record for this coverage run.
       --include=PAT1,PAT2,...
                             Include only files whose paths match one of these
@@ -152,7 +152,7 @@ There are many options:
       --rcfile=RCFILE       Specify configuration file. By default '.coveragerc',
                             'setup.cfg', 'tox.ini', and 'pyproject.toml' are
                             tried. [env: COVERAGE_RCFILE]
-.. [[[end]]] (checksum: 869a31153b3cf401c52523ae9b52c7ab)
+.. [[[end]]] (checksum: 072cccad7f8ad3e7b72c266305ef5e4a)
 
 If you want :ref:`branch coverage <branch>` measurement, use the ``--branch``
 flag.  Otherwise only statement coverage is measured.
@@ -174,13 +174,17 @@ but before the program invocation::
 
 
 Coverage.py can measure multi-threaded programs by default. If you are using
-more exotic concurrency, with the `multiprocessing`_, `greenlet`_, `eventlet`_,
-or `gevent`_ libraries, then coverage.py will get very confused.  Use the
+more other concurrency support, with the `multiprocessing`_, `greenlet`_,
+`eventlet`_, or `gevent`_ libraries, then coverage.py can get confused. Use the
 ``--concurrency`` switch to properly measure programs using these libraries.
 Give it a value of ``multiprocessing``, ``thread``, ``greenlet``, ``eventlet``,
 or ``gevent``.  Values other than ``thread`` require the :ref:`C extension
 <install_extension>`.
 
+You can combine multiple values for ``--concurrency``, separated with commas.
+You can specify ``thread`` and also one of ``eventlet``, ``gevent``, or
+``greenlet``.
+
 If you are using ``--concurrency=multiprocessing``, you must set other options
 in the configuration file.  Options on the command line will not be passed to
 the processes that multiprocessing creates.  Best practice is to use the

doc/requirements.pip

@@ -10,7 +10,7 @@ babel==2.9.1
     # via sphinx
 certifi==2021.10.8
     # via requests
-charset-normalizer==2.0.7
+charset-normalizer==2.0.8
     # via requests
 cogapp==3.3.0
     # via -r doc/requirements.in

requirements/dev.pip

@@ -25,7 +25,7 @@ build==0.7.0
     # via check-manifest
 certifi==2021.10.8
     # via requests
-charset-normalizer==2.0.7
+charset-normalizer==2.0.8
     # via requests
 check-manifest==0.47
     # via -r requirements/dev.in

requirements/light-threads.in

@@ -6,4 +6,5 @@
 # The light-threads packages we test against
 
 eventlet
+gevent
 greenlet

requirements/light-threads.pip

@@ -8,9 +8,20 @@ dnspython==2.1.0
     # via eventlet
 eventlet==0.33.0
     # via -r requirements/light-threads.in
+gevent==21.8.0
+    # via -r requirements/light-threads.in
 greenlet==1.1.2
     # via
     #   -r requirements/light-threads.in
     #   eventlet
+    #   gevent
 six==1.16.0
     # via eventlet
+zope.event==4.5.0
+    # via gevent
+zope.interface==5.4.0
+    # via gevent
+
+# The following packages are considered to be unsafe in a requirements file:
+setuptools==59.2.0
+    # via gevent

tests/test_cmdline.py

@@ -610,7 +610,7 @@ def test_run(self):
             cov.save()
             """)
         self.cmd_executes("run --concurrency=gevent foo.py", """\
-            cov = Coverage(concurrency='gevent')
+            cov = Coverage(concurrency=['gevent'])
             runner = PyRunner(['foo.py'], as_module=False)
             runner.prepare()
             cov.start()
@@ -619,27 +619,23 @@ def test_run(self):
             cov.save()
             """)
         self.cmd_executes("run --concurrency=multiprocessing foo.py", """\
-            cov = Coverage(concurrency='multiprocessing')
+            cov = Coverage(concurrency=['multiprocessing'])
+            runner = PyRunner(['foo.py'], as_module=False)
+            runner.prepare()
+            cov.start()
+            runner.run()
+            cov.stop()
+            cov.save()
+            """)
+        self.cmd_executes("run --concurrency=gevent,thread foo.py", """\
+            cov = Coverage(concurrency=['gevent', 'thread'])
             runner = PyRunner(['foo.py'], as_module=False)
             runner.prepare()
             cov.start()
             runner.run()
             cov.stop()
             cov.save()
             """)
-
-    def test_bad_concurrency(self):
-        self.command_line("run --concurrency=nothing", ret=ERR)
-        err = self.stderr()
-        assert "option --concurrency: invalid choice: 'nothing'" in err
-
-    def test_no_multiple_concurrency(self):
-        # You can't use multiple concurrency values on the command line.
-        # I would like to have a better message about not allowing multiple
-        # values for this option, but optparse is not that flexible.
-        self.command_line("run --concurrency=multiprocessing,gevent foo.py", ret=ERR)
-        err = self.stderr()
-        assert "option --concurrency: invalid choice: 'multiprocessing,gevent'" in err
 
     def test_multiprocessing_needs_config_file(self):
         # You can't use command-line args to add options to multiprocessing