fix: allow to set all configs via tesseract build --config-override (#239)

#### Relevant issue or PR
Fixes #232 

#### Description of changes
- Parse values passed as YAML, so arguments that expect lists (like
`build_config.package_data`) can be set.
- Some more error checking and helpful error messages for good measure.
- Add more testing to ensure things work as intended.

#### Testing done
CI, old and new tests

Author: dionhaefner

tesseract_core/sdk/cli.py

@@ -17,6 +17,7 @@
 
 import click
 import typer
+import yaml
 from jinja2 import Environment, PackageLoader, StrictUndefined
 from pydantic import ValidationError as PydanticValidationError
 from rich.console import Console as RichConsole
@@ -193,27 +194,38 @@ def main_callback(
 
 def _parse_config_override(
     options: list[str] | None,
-) -> tuple[tuple[list[str], str], ...]:
+) -> dict[tuple[str, ...], Any]:
     """Parse `["path1.path2.path3=value"]` into `[(["path1", "path2", "path3"], "value")]`."""
     if options is None:
-        return ()
+        return {}
 
-    def _parse_option(option: str):
-        bad_param = typer.BadParameter(
-            f"Invalid config override {option} (must be `keypath=value`)",
-            param_hint="config_override",
-        )
-        if option.count("=") != 1:
-            raise bad_param
+    def _parse_option(option: str) -> tuple[tuple[str, ...], Any]:
+        if "=" not in option:
+            raise typer.BadParameter(
+                f'Invalid config override "{option}" (must be `keypath=value`)',
+                param_hint="config_override",
+            )
 
-        key, value = option.split("=")
-        if not key or not value:
-            raise bad_param
+        key, value = option.split("=", maxsplit=1)
+        if not re.match(r"\w[\w|\.]*", key):
+            raise typer.BadParameter(
+                f'Invalid keypath "{key}" in config override "{option}"',
+                param_hint="config_override",
+            )
+
+        path = tuple(key.split("."))
+
+        try:
+            value = yaml.safe_load(value)
+        except yaml.YAMLError as e:
+            raise typer.BadParameter(
+                f'Invalid value for config override "{option}", could not parse value as YAML: {e}',
+                param_hint="config_override",
+            ) from e
 
-        path = key.split(".")
         return path, value
 
-    return tuple(_parse_option(option) for option in options)
+    return dict(_parse_option(option) for option in options)
 
 
 @app.command("build")

tesseract_core/sdk/engine.py

@@ -371,7 +371,7 @@ def build_tesseract(
     image_tag: str | None,
     build_dir: Path | None = None,
     inject_ssh: bool = False,
-    config_override: tuple[tuple[list[str], str], ...] = (),
+    config_override: dict[tuple[str, ...], Any] | None = None,
     generate_only: bool = False,
 ) -> Image | Path:
     """Build a new Tesseract from a context directory.
@@ -397,11 +397,12 @@ def build_tesseract(
     config = get_config(src_dir)
 
     # Apply config overrides
-    for path, value in config_override:
-        c = config
-        for k in path[:-1]:
-            c = getattr(c, k)
-        setattr(c, path[-1], value)
+    if config_override is not None:
+        for path, value in config_override.items():
+            c = config
+            for k in path[:-1]:
+                c = getattr(c, k)
+            setattr(c, path[-1], value)
 
     image_name = config.name
     if image_tag:

tests/sdk_tests/test_cli.py

@@ -52,3 +52,75 @@ def test_bad_docker_executable_env_var():
     )
     assert result.returncode == 1
     assert "Executable `not-a-docker` not found" in result.stderr.decode()
+
+
+@pytest.mark.parametrize(
+    "arg_to_override",
+    [
+        "name",
+        "build_config.custom_build_steps",
+        "build_config.base_image",
+        "build_config.package_data",
+    ],
+)
+def test_config_override(
+    arg_to_override, cli_runner, mocker, dummy_tesseract_location, mocked_docker
+):
+    mocked_build = mocker.patch("tesseract_core.sdk.engine.build_tesseract")
+
+    def _run_with_override(key, value):
+        return cli_runner.invoke(
+            cli,
+            [
+                "build",
+                str(dummy_tesseract_location),
+                "--config-override",
+                f"{key}={value}",
+                "--generate-only",
+            ],
+            catch_exceptions=False,
+        )
+
+    if arg_to_override == "name":
+        argpairs = (
+            (
+                "my-tesseract",
+                {("name",): "my-tesseract"},
+            ),
+        )
+    elif arg_to_override == "build_config.custom_build_steps":
+        argpairs = (
+            (
+                "[RUN foo='bar']",
+                {("build_config", "custom_build_steps"): ["RUN foo='bar'"]},
+            ),
+            (
+                '[RUN echo "hello world"]',
+                {("build_config", "custom_build_steps"): ['RUN echo "hello world"']},
+            ),
+        )
+    elif arg_to_override == "build_config.base_image":
+        argpairs = (
+            (
+                "ubuntu:latest",
+                {("build_config", "base_image"): "ubuntu:latest"},
+            ),
+        )
+    elif arg_to_override == "build_config.package_data":
+        argpairs = (
+            (
+                '["data/file.txt:/app/data/file.txt"]',
+                {
+                    ("build_config", "package_data"): [
+                        "data/file.txt:/app/data/file.txt"
+                    ]
+                },
+            ),
+        )
+    else:
+        raise ValueError(f"Unknown arg_to_override: {arg_to_override}")
+
+    for value, expected in argpairs:
+        result = _run_with_override(arg_to_override, value)
+        assert result.exit_code == 0, result.stderr
+        assert mocked_build.call_args[1]["config_override"] == expected