Merge pull request datalad#6864 from mih/bf-6863

Support unsetting config via `datalad -c :<name>`

datalad/cli/common_args.py

@@ -34,10 +34,12 @@
         ('-c',),
         dict(action='append',
              dest='cfg_overrides',
-             metavar='KEY=VALUE',
-             help="""configuration variable setting. Overrides any
-             configuration read from a file, but is potentially overridden
-             itself by configuration variables in the process environment.
+             metavar='(:name|name=value)',
+             help="""specify configuration setting overrides. They override any
+             configuration read from a file. A configuration can also be
+             unset temporarily by prefixing its name with a colon (':'), e.g. ':user.name'.
+             Overrides specified here may be overridden themselves by
+             configuration settings declared as environment variables.
              """)),
     change_path=(
         ('-C',),

datalad/cli/helpers.py

@@ -255,17 +255,23 @@ def _parse_overrides_from_cmdline(cmdlineargs):
     # errors: we need a section, a variable, and a value at minimum
     # otherwise we break our own config parsing helpers
     # https://github.com/datalad/datalad/issues/3451
-    noassign_expr = re.compile(r'[^\s]+\.[^\s]+=[\S]+')
+    assign_expr = re.compile(r'[^\s]+\.[^\s]+=[\S]+')
+    unset_expr = re.compile(r':[^\s]+\.[^\s=]+')
     noassign = [
         o
         for o in cmdlineargs.cfg_overrides
-        if not noassign_expr.match(o)
+        if not (assign_expr.match(o) or unset_expr.match(o))
     ]
     if noassign:
         lgr.error(
             "Configuration override without section/variable "
-            "or value assignment (must be 'section.variable=value'): %s",
+            "or unset marker or value assignment "
+            "(must be '(:section.variable|section.variable=value)'): %s",
             noassign)
         sys.exit(3)
-    overrides = dict(o.split('=', 1) for o in cmdlineargs.cfg_overrides)
+    overrides = dict(
+        [o[1:], None] if o.startswith(':')
+        else o.split('=', 1)
+        for o in cmdlineargs.cfg_overrides
+    )
     return overrides

datalad/cli/tests/test_main.py

@@ -269,11 +269,12 @@ def test_script_shims(script):
                  get_numeric_portion(version))
 
 
-@slow  # 11.2591s
 @with_tempfile(mkdir=True)
 def test_cfg_override(path=None):
     with chpwd(path):
-        cmd = ['datalad', 'wtf', '-s', 'some']
+        # use 'wtf' to dump the config
+        # should be rewritten to use `configuration`
+        cmd = ['datalad', 'wtf', '-S', 'configuration', '-s', 'some']
         # control
         out = Runner().run(cmd, protocol=StdOutErrCapture)['stdout']
         assert_not_in('datalad.dummy: this', out)
@@ -305,6 +306,27 @@ def test_cfg_override(path=None):
                            protocol=StdOutErrCapture)['stdout']
         assert_in('datalad.dummy: this', out)
 
+        # set a config
+        run_main([
+            'configuration', '--scope', 'local', 'set', 'mike.item=some'])
+        # verify it is successfully set
+        assert 'some' == run_main([
+            'configuration', 'get', 'mike.item'])[0].strip()
+        # verify that an override can unset the config
+        # we cannot use run_main(), because the "singleton" instance of the
+        # dataset we are in is still around in this session, and with it
+        # also its config managers that we will not be able to post-hoc
+        # overwrite with this method. Instead, we'll execute in a subprocess.
+        assert '' == Runner().run([
+            'datalad', '-c', ':mike.item',
+            'configuration', 'get', 'mike.item'],
+            protocol=StdOutErrCapture)['stdout'].strip()
+        # verify the effect is not permanent
+        assert 'some' == Runner().run([
+            'datalad',
+            'configuration', 'get', 'mike.item'],
+            protocol=StdOutErrCapture)['stdout'].strip()
+
 
 def test_incorrect_cfg_override():
     run_main(['-c', 'some', 'wtf'], exit_code=3)