feat: dump_settings management command (#36162)

This command dumps the current Django settings to JSON for
debugging/diagnostics. The output of this command is for *humans*... it
is NOT suitable for consumption by production systems.

In particular, we are introducing this command as part of a series of
refactorings to the Django settings files lms/envs/* and cms/envs/*.
We want to ensure that these refactorings do not introduce any
unexpected breaking changes, so the dump_settings command will both help
us manually verify our refactorings and help operators verify that our
refactorings behave expectedly when using their custom python/yaml
settings files.

Related to: https://github.com/openedx/edx-platform/pull/36131

openedx/core/djangoapps/util/management/commands/dump_settings.py

@@ -0,0 +1,93 @@
+"""
+Defines the dump_settings management command.
+"""
+import inspect
+import json
+import re
+
+from django.conf import settings
+from django.core.management.base import BaseCommand
+
+
+SETTING_NAME_REGEX = re.compile(r'^[A-Z][A-Z0-9_]*$')
+
+
+class Command(BaseCommand):
+    """
+    Dump current Django settings to JSON for debugging/diagnostics.
+
+    BEWARE: OUTPUT IS NOT SUITABLE FOR CONSUMPTION BY PRODUCTION SYSTEMS.
+    The purpose of this output is to be *helpful* for a *human* operator to understand how their settings are being
+    rendered and how they differ between different settings files. The serialization format is NOT perfect: there are
+    certain situations where two different settings will output identical JSON. For example, this command does NOT:
+
+    disambiguate between lists and tuples:
+    * (1, 2, 3)  # <-- this tuple will be printed out as [1, 2, 3]
+    * [1, 2, 3]
+
+    disambiguate between sets and sorted lists:
+    * {2, 1, 3}  # <-- this set will be printed out as [1, 2, 3]
+    * [1, 2, 3]
+
+    disambiguate between internationalized and non-internationalized strings:
+    * _("hello") # <-- this will become just "hello"
+    * "hello"
+
+    Furthermore, objects which are not easily JSON-ifiable will stringified using their `repr(...)`, e.g.:
+    * "Path('my/path')"                                                # a Path object
+    * "<lms.myapp.MyClass object at 0x704599fa2fd0>"                   # some random class instance
+    * "<_io.TextIOWrapper name='<stderr>' mode='w' encoding='utf-8'>"  # sys.stderr
+
+    and lambdas are printed by *roughly* printing out their source lines (it's impossible in Python to get the *exact*
+    source code, as it's been compiled into bytecode).
+    """
+
+    def handle(self, *args, **kwargs):
+        """
+        Handle the command.
+        """
+        settings_json = {
+            name: _to_json_friendly_repr(getattr(settings, name))
+            for name in dir(settings)
+            if SETTING_NAME_REGEX.match(name)
+        }
+        print(json.dumps(settings_json, indent=4))
+
+
+def _to_json_friendly_repr(value: object) -> object:
+    """
+    Turn the value into something that we can print to a JSON file (that is: str, bool, None, int, float, list, dict).
+
+    See the docstring of `Command` for warnings about this function's behavior.
+    """
+    if isinstance(value, (type(None), bool, int, float, str)):
+        # All these types can be printed directly
+        return value
+    if isinstance(value, (list, tuple, set)):
+        if isinstance(value, set):
+            # Print sets by sorting them (so that order doesn't matter) into a JSON array.
+            elements = sorted(value)
+        else:
+            # Print both lists and tuples as JSON arrays.
+            elements = value
+        return [_to_json_friendly_repr(element) for ix, element in enumerate(elements)]
+    if isinstance(value, dict):
+        # Print dicts as JSON objects
+        for subkey in value.keys():
+            if not isinstance(subkey, (str, int)):
+                raise ValueError(f"Unexpected dict key {subkey} of type {type(subkey)}")
+        return {subkey: _to_json_friendly_repr(subval) for subkey, subval in value.items()}
+    if proxy_args := getattr(value, "_proxy____args", None):
+        if len(proxy_args) == 1 and isinstance(proxy_args[0], str):
+            # Print gettext_lazy as simply the wrapped string
+            return proxy_args[0]
+    try:
+        qualname = value.__qualname__
+    except AttributeError:
+        pass
+    else:
+        if qualname == "<lambda>":
+            # Handle lambdas by printing the source lines
+            return "lambda defined with line(s): " + inspect.getsource(value).strip()
+    # For all other objects, print the repr
+    return repr(value)

openedx/core/djangoapps/util/tests/test_dump_settings.py

@@ -0,0 +1,64 @@
+"""
+Basic tests for dump_settings management command.
+
+These are moreso testing that dump_settings works, less-so testing anything about the Django
+settings files themselves. Remember that tests only run with (lms,cms)/envs/test.py,
+which are based on (lms,cms)/envs/common.py, so these tests will not execute any of the
+YAML-loading or post-processing defined in (lms,cms)/envs/production.py.
+"""
+import json
+
+from django.core.management import call_command
+
+from openedx.core.djangolib.testing.utils import skip_unless_lms, skip_unless_cms
+
+
+@skip_unless_lms
+def test_for_lms_settings(capsys):
+    """
+    Ensure LMS's test settings can be dumped, and sanity-check them for certain values.
+    """
+    dump = _get_settings_dump(capsys)
+
+    # Check: something LMS-specific
+    assert dump['MODULESTORE_BRANCH'] == "published-only"
+
+    # Check: tuples are converted to lists
+    assert isinstance(dump['XBLOCK_MIXINS'], list)
+
+    # Check: objects (like classes) are repr'd
+    assert "<class 'xmodule.x_module.XModuleMixin'>" in dump['XBLOCK_MIXINS']
+
+    # Check: nested dictionaries come through OK, and int'l strings are just strings
+    assert dump['COURSE_ENROLLMENT_MODES']['audit']['display_name'] == "Audit"
+
+
+@skip_unless_cms
+def test_for_cms_settings(capsys):
+    """
+    Ensure CMS's test settings can be dumped, and sanity-check them for certain values.
+    """
+    dump = _get_settings_dump(capsys)
+
+    # Check: something CMS-specific
+    assert dump['MODULESTORE_BRANCH'] == "draft-preferred"
+
+    # Check: tuples are converted to lists
+    assert isinstance(dump['XBLOCK_MIXINS'], list)
+
+    # Check: objects (like classes) are repr'd
+    assert "<class 'xmodule.x_module.XModuleMixin'>" in dump['XBLOCK_MIXINS']
+
+    # Check: nested dictionaries come through OK, and int'l strings are just strings
+    assert dump['COURSE_ENROLLMENT_MODES']['audit']['display_name'] == "Audit"
+
+
+def _get_settings_dump(captured_sys):
+    """
+    Call dump_settings, ensure no error output, and return parsed JSON.
+    """
+    call_command('dump_settings')
+    out, err = captured_sys.readouterr()
+    assert out
+    assert not err
+    return json.loads(out)