Use JSON schema for automatic config validation

pyproject.toml

@@ -16,6 +16,7 @@ dependencies = [
     "appdirs==1.4.4",
     "argcomplete==3.0.5",
     "docx2python==1.27.1",
+    "jsonschema==4.19.0",
     "nltk==3.8.1",
     "pdfplumber==0.9.0",
     "protobuf~=3.19.0", # Used by Stanza; see https://github.com/spraakbanken/sparv-pipeline/issues/161

sparv/core/Snakefile

@@ -238,7 +238,10 @@ update_autocompletion_cache()
 sparv_config.validate_module_config()
 
 # Validate config
-sparv_config.validate_config()
+if not "schema" in selected_targets:
+    from sparv.core import schema
+    json_schema = schema.build_json_schema(sparv_config.config_structure)
+    schema.validate(sparv_config.config, json_schema)
 
 # Get reverse_config_usage dict for look-ups
 reverse_config_usage = snake_utils.get_reverse_config_usage()

sparv/core/config.py

@@ -223,25 +223,6 @@ def validate_module_config():
                     "are" if len(annotators) > 1 else "is", config_key), "sparv", "config")
 
 
-def validate_config(config_dict=None, structure=None, parent=""):
-    """Make sure the corpus config doesn't contain invalid keys."""
-    config_dict = config_dict or config
-    structure = structure or config_structure
-    for key in config_dict:
-        path = (parent + "." + key) if parent else key
-        if key not in structure:
-            if not parent:
-                raise SparvErrorMessage(f"Unknown key in config file: '{path}'. No module with that name found.",
-                                        module="sparv", function="config")
-            else:
-                module_name = parent.split(".", 1)[0]
-                raise SparvErrorMessage(f"Unknown key in config file: '{path}'. The module '{module_name}' "
-                                        f"doesn't have an option with that name.",
-                                        module="sparv", function="config")
-        elif not structure[key].get("_source"):
-            validate_config(config_dict[key], structure[key], path)
-
-
 def load_presets(lang, lang_variety):
     """Read presets files and return dictionaries with all available presets annotations and preset classes."""
     global presets

sparv/core/schema.py

@@ -351,21 +351,30 @@ def validate(cfg: dict, schema: dict) -> None:
     """Validate a Sparv config using JSON schema."""
     import jsonschema
 
+    def build_path_string(path):
+        parts = []
+        for part in path:
+            if isinstance(part, str):
+                parts.append(part)
+            elif isinstance(part, int):
+                parts[-1] += f"[{part}]"
+        return ".".join(parts)
+
     try:
         jsonschema.validate(cfg, schema)
     except jsonschema.ValidationError as e:
         msg = ["There was a problem trying to parse the corpus config file.\n"]
 
-        if e.validator == "unevaluatedProperties":
-            # This only happens for unexpected keys at the root level
-            prop = re.search(r"'(.+)' was unexpected", e.message)
-            if prop:
-                msg.append(f"Unexpected property at root level: {prop.group(1)!r}")
-            else:
-                msg.append(e.message)
+        # Rephrase messages about unexpected keys
+        unknown_key = re.search(r"properties are not allowed \('(.+)' was unexpected", e.message)
+        if unknown_key:
+            full_path = ".".join(list(e.absolute_path) + [unknown_key.group(1)])
+            msg.append(f"Unexpected key in config file: {full_path!r}")
         else:
-            if e.absolute_path:
-                msg.append(f"Offending config path: {'.'.join(e.absolute_path)}")
             msg.append(e.message)
+            if e.absolute_path:
+                msg.append(f"Offending config path: {build_path_string(e.absolute_path)}")
+                if "description" in e.schema:
+                    msg.append(f"Description of config key: {e.schema['description']}")
 
         raise SparvErrorMessage("\n".join(msg))