add base.context.dump to save experiment config as metadata

CHANGES.txt

@@ -1,7 +1,9 @@
 Next release
 ============
 
-- ...
+Features
+--------
+- New base.context.dump task (thanks @vlap)
 
 
 ScriptEngine 1.2.0

docs/sphinx/base-tasks.rst

@@ -8,10 +8,10 @@ with just the base task package.
 
 The base task package contains the following tasks, described in more detail below::
 
-    base.echo, base.chdir, base.command, base.context, base.context.from,
-    base.copy, base.exit, base.find, base.getenv, base.include, base.link,
-    base.make_dir, base.move, base.remove, base.setenv, base.task_timer,
-    base.template, base.time, base.unsetenv
+    base.echo, base.chdir, base.command, base.context, base.context.load,
+    base.context.dump, base.copy, base.exit, base.find, base.getenv,
+    base.include, base.link, base.make_dir, base.move, base.remove,
+    base.setenv, base.task_timer, base.template, base.time, base.unsetenv
 
 
 ``base.echo``
@@ -162,6 +162,47 @@ possibly nested, dictionary (i.e. single values or lists are not allowed).
     This task has been renamed to ``base.context.load`` (the old name was ``base.context.from``).
 
 
+``base.context.dump``
+^^^^^^^^^^^^^^^^^^^^^
+Dumps the current ScriptEngine context data, or a subset of context keys, to a YAML file::
+
+    base.context.dump:
+        file: <FILE_NAME>  # required
+        keys: <KEY_OR_LIST_OF_KEYS>  # optional
+        root: <ROOT_KEY>  # optional
+
+This task exports the evaluated runtime state of the context. Note that it dumps the context data
+dictionary, not the ScriptEngine script definitions or execution structure. Dynamic constructs
+(like ``base.include`` or task logic) and internal engine parameters (the ``se`` namespace) are not
+part of the output.
+
+By default (if ``keys`` is not specified), ``base.context.dump`` dumps the full context data to the
+given ``file``::
+
+    - base.context.dump:
+        file: all_context.yml
+
+To dump only specific context keys, use the ``keys`` argument::
+
+    - base.context.dump:
+        file: experiment-config.yml
+        keys:
+          - experiment
+          - model_config
+
+The ``root`` argument wraps the dumped dictionary under an optional top-level key::
+
+    - base.context.dump:
+        file: experiment-config.yml
+        root: "ic_meta"
+        keys:
+          - experiment
+          - model_config
+
+.. versionadded:: 1.3
+    Add ``base.context.dump`` task.
+
+
 Control flow
 ------------
 
@@ -331,7 +372,7 @@ will set the environment variables ``$LD_LIBRARY_PATH`` to
    Allow dotted keys for nested context parameters.
 
 ``base.unsetenv``
-^^^^^^^^^^^^^^^
+^^^^^^^^^^^^^^^^^
 Unsets one or more environment variables::
 
     - base.unsetenv:

pyproject.toml

@@ -40,6 +40,7 @@ se = "scriptengine.cli.se:main"
 "base.command" = "scriptengine.tasks.base.command:Command"
 "base.context" = "scriptengine.tasks.base.context:Context"
 "base.context.load" = "scriptengine.tasks.base.context:ContextLoad"
+"base.context.dump" = "scriptengine.tasks.base.context:ContextDump"
 "base.copy" = "scriptengine.tasks.base.file.copy:Copy"
 "base.echo" = "scriptengine.tasks.base.echo:Echo"
 "base.exit" = "scriptengine.tasks.base.exit:Exit"

src/scriptengine/tasks/base/context.py

@@ -70,7 +70,7 @@ def run(self, context):
         elif file_arg:
             self.log_info(f"Load context update from file: {file_arg}")
             try:
-                with open(file_arg) as f:
+                with open(str(file_arg)) as f:
                     dict_from_file = yaml.load(f, Loader=yaml.SafeLoader)
             except (FileNotFoundError, PermissionError, IsADirectoryError) as e:
                 self.log_error(e)
@@ -93,3 +93,52 @@ def run(self, context):
             raise ScriptEngineTaskError
 
         return SEContext(context_update_d)
+
+
+class ContextDump(Task):
+    """
+    This task dumps the context, or a subset of context keys, to a YAML file.
+
+    Examples:
+    - base.context.dump:
+        file: saveic/experiment-config.yml
+        root: "ic_meta"
+        keys:
+          - experiment
+          - model_config
+
+    - base.context.dump:
+        file: all_context.yml
+    """
+
+    _required_arguments = ("file",)
+
+    @timed_runner
+    def run(self, context):
+        file_arg = self.getarg("file", context)
+        keys_arg = self.getarg("keys", context, default=None)
+        root_arg = self.getarg("root", context, default=None)
+
+        self.log_info(f"Dump context to file: {file_arg}")
+
+        if keys_arg is not None:
+            keys_list = [keys_arg] if isinstance(keys_arg, str) else keys_arg
+            if not isinstance(keys_list, list):
+                self.log_error(
+                    f"The 'keys' argument must be a string or list (was a '{type(keys_arg).__name__}')"
+                )
+                raise ScriptEngineTaskRunError
+            data = {k: context[k] for k in keys_list if k in context}
+        else:
+            # Dump full context excluding internal 'se' namespace
+            data = {k: v for k, v in context.items() if k != "se"}
+
+        if root_arg is not None:
+            data = {str(root_arg): data}
+
+        try:
+            with open(str(file_arg), "w") as f:
+                yaml.dump(data, f, sort_keys=False)
+        except (FileNotFoundError, PermissionError, IsADirectoryError, OSError) as e:
+            self.log_error(e)
+            raise ScriptEngineTaskRunError

src/scriptengine/yaml/parser.py

@@ -49,6 +49,19 @@ def rrule_constructor(loader, node):
 yaml.add_constructor("!rrule", rrule_constructor)
 
 
+def string_class_representer(tag):
+    return lambda dumper, node: dumper.represent_scalar(tag, str(node))
+
+
+yaml.add_representer(NoParseString, string_class_representer("!noparse"))
+yaml.add_representer(NoParseYamlString, string_class_representer("!noparse_yaml"))
+yaml.add_representer(NoParseJinjaString, string_class_representer("!noparse_jinja"))
+yaml.add_representer(
+    dateutil.rrule.rrule,
+    lambda dumper, node: dumper.represent_scalar("!rrule", str(node), style="|"),
+)
+
+
 def parse(data):
     """Recursively parses data and returns a ScriptEngine Task or Job, or a list of those.
     The data is supposed to come from YAML-parsing a ScriptEngine script."""

tests/tasks/base/test_context.py

@@ -212,3 +212,126 @@ def test_context_load_file_not_dict(tmp_path):
     )
     with pytest.raises(ScriptEngineTaskRunError):
         t.run(SEContext())
+
+
+def test_context_dump_file(tmp_path):
+    f = tmp_path / "f.yml"
+    t1 = from_yaml(
+        """
+        base.context:
+            foo: 1
+            bar: 2
+        """
+    )
+    t2 = from_yaml(
+        f"""
+        base.context.dump:
+            file: {f}
+        """
+    )
+    ctx = SEContext()
+    ctx += t1.run(ctx)
+    t2.run(ctx)
+    assert f.exists()
+    assert yaml.safe_load(f.read_text()) == {"foo": 1, "bar": 2}
+
+
+def test_context_dump_keys(tmp_path):
+    f = tmp_path / "f.yml"
+    t1 = from_yaml(
+        """
+        base.context:
+            foo: 1
+            bar: 2
+            baz: 3
+        """
+    )
+    t2 = from_yaml(
+        f"""
+        base.context.dump:
+            file: {f}
+            keys:
+                - foo
+                - bar
+        """
+    )
+    ctx = SEContext()
+    ctx += t1.run(ctx)
+    t2.run(ctx)
+    assert f.exists()
+    assert yaml.safe_load(f.read_text()) == {"foo": 1, "bar": 2}
+
+
+def test_context_dump_root(tmp_path):
+    f = tmp_path / "f.yml"
+    t1 = from_yaml(
+        """
+        base.context:
+            foo: 1
+            bar: 2
+        """
+    )
+    t2 = from_yaml(
+        f"""
+        base.context.dump:
+            file: {f}
+            root: ic_meta
+            keys:
+                - foo
+        """
+    )
+    ctx = SEContext()
+    ctx += t1.run(ctx)
+    t2.run(ctx)
+    assert f.exists()
+    assert yaml.safe_load(f.read_text()) == {"ic_meta": {"foo": 1}}
+
+
+def test_context_dump_no_args():
+    t = from_yaml(
+        """
+        base.context.dump:
+        """
+    )
+    with pytest.raises(ScriptEngineTaskError):
+        t.run(SEContext())
+
+
+def test_context_dump_keys_string(tmp_path):
+    f = tmp_path / "f.yml"
+    t1 = from_yaml(
+        """
+        base.context:
+            foo: 1
+            bar: 2
+        """
+    )
+    t2 = from_yaml(
+        f"""
+        base.context.dump:
+            file: {f}
+            keys: foo
+        """
+    )
+    ctx = SEContext()
+    ctx += t1.run(ctx)
+    t2.run(ctx)
+    assert f.exists()
+    assert yaml.safe_load(f.read_text()) == {"foo": 1}
+
+
+def test_context_dump_keys_invalid_type(tmp_path):
+    f = tmp_path / "f.yml"
+    t = from_yaml(
+        f"""
+        base.context.dump:
+            file: {f}
+            keys:
+                a: 1
+        """
+    )
+    with pytest.raises(ScriptEngineTaskRunError):
+        t.run(SEContext())
+
+
+