fix schema validation prohibiting job execution with empty string input

CHANGES.rst

@@ -23,6 +23,8 @@ Changes:
 Fixes:
 ------
 - Fix broken `OpenAPI` schema link references to `OGC API - Processes` repository.
+- Fix `Job` creation failing when submitting an empty string as input for a `Process` that allows it due
+  to schema validation incorrectly preventing it.
 - Fix ``GET /providers/{provider_id}`` response using ``$schema`` instead of ``$id`` to provide its content schema.
 
 .. _changes_4.30.1:

tests/functional/test_wps_app.py

@@ -20,6 +20,7 @@
     setup_config_with_celery,
     setup_config_with_mongodb,
     setup_config_with_pywps,
+    setup_mongodb_jobstore,
     setup_mongodb_processstore
 )
 from weaver import xml_util
@@ -45,6 +46,7 @@ def setUp(self):
         config = setup_config_with_pywps(config)
         config = setup_config_with_celery(config)
         self.process_store = setup_mongodb_processstore(config)
+        self.job_store = setup_mongodb_jobstore(config)
         self.app = get_test_weaver_app(config=config, settings=settings)
 
         # add processes by database Process type
@@ -146,6 +148,24 @@ def test_execute_allowed_demo(self):
         resp.mustcontain("<wps:ProcessAccepted")
         resp.mustcontain(f"PyWPS Process {HelloWPS.identifier}")
 
+    def test_execute_allowed_empty_string(self):
+        template = "service=wps&request=execute&version=1.0.0&identifier={}&datainputs=name="
+        params = template.format(HelloWPS.identifier)
+        url = self.make_url(params)
+        with contextlib.ExitStack() as stack_exec:
+            for mock_exec in mocked_execute_celery():
+                stack_exec.enter_context(mock_exec)
+            resp = self.app.get(url)
+        assert resp.status_code == 200  # FIXME: replace by 202 Accepted (?) https://github.com/crim-ca/weaver/issues/14
+        assert resp.content_type in ContentType.ANY_XML
+        resp.mustcontain("<wps:ExecuteResponse")
+        resp.mustcontain("<wps:ProcessAccepted")
+        resp.mustcontain(f"PyWPS Process {HelloWPS.identifier}")
+        loc = resp.xml.get("statusLocation")
+        job_id = loc.rsplit("/", 1)[-1].split(".", 1)[0]
+        job = self.job_store.fetch_by_id(job_id)
+        assert job.inputs[0]["data"] == ""
+
     def test_execute_deployed_with_visibility_allowed(self):
         headers = {"Accept": ContentType.APP_XML}
         params_template = "service=wps&request=execute&version=1.0.0&identifier={}&datainputs=test_input=test"

tests/wps_restapi/test_processes.py

@@ -1951,6 +1951,21 @@ def test_execute_process_no_json_body(self):
         assert resp.status_code == 400
         assert resp.content_type == ContentType.APP_JSON
 
+    def test_execute_process_valid_empty_string(self):
+        """
+        Ensure that a process expecting an input string parameter can be provided as empty (not resolved as "missing").
+        """
+        path = f"/processes/{self.process_public.identifier}/jobs"
+        data = self.get_process_execute_template(test_input="")
+
+        with contextlib.ExitStack() as stack:
+            for exe in mocked_process_job_runner():
+                stack.enter_context(exe)
+            resp = self.app.post_json(path, params=data, headers=self.json_headers)
+            assert resp.status_code == 201, "Expected job submission without inputs created without error."
+            job = self.job_store.fetch_by_id(resp.json["jobID"])
+            assert job.inputs[0]["data"] == "", "Input value should be an empty string."
+
     def test_execute_process_missing_required_params(self):
         """
         Validate execution against missing parameters.

weaver/processes/convert.py

@@ -2861,10 +2861,10 @@ def wps2json_job_payload(wps_request, wps_process):
         for input_value in input_list:
             input_data = input_value.get("data")
             input_href = input_value.get("href")
-            if input_data:
-                data["inputs"].append({"id": iid, "data": input_data})
-            elif input_href:
+            if input_href:  # when href is provided, it must always be non-empty
                 data["inputs"].append({"id": iid, "href": input_href})
+            else:  # no check if value to allow possible empty string, numeric zero or explicit null
+                data["inputs"].append({"id": iid, "data": input_data})
     output_ids = list(wps_request.outputs)
     for output in wps_process.outputs:
         oid = output.identifier

weaver/utils.py

@@ -3488,7 +3488,7 @@ def clean_json_text_body(body, remove_newlines=True, remove_indents=True):
     """
     # cleanup various escape characters and u'' stings
     replaces = [(",\n", ", "), ("\\n", " "), (" \n", " "), ("\n'", "'"), ("\"", "\'"),
-                ("u\'", "\'"), ("u\"", "\'"), ("\'\'", "\'"), ("'. ", ""), ("'. '", ""),
+                ("u\'", "\'"), ("u\"", "\'"), ("'. ", ""), ("'. '", ""),
                 ("}'", "}"), ("'{", "{")]
     if remove_indents:
         replaces.extend([("\\", " "), ("  ", " ")])

weaver/wps_restapi/swagger_definitions.py

@@ -1314,8 +1314,6 @@ class BoundingBoxInputType(ExtendedMappingSchema):
     supportedCRS = SupportedCRSList()
 
 
-# FIXME: support byte/binary type (string + format:byte) ?
-#   https://github.com/opengeospatial/ogcapi-processes/blob/master/openapi/schemas/processes-core/binaryInputValue.yaml
 class AnyLiteralType(OneOfKeywordSchema):
     """
     Submitted values that correspond to literal data.
@@ -1326,10 +1324,29 @@ class AnyLiteralType(OneOfKeywordSchema):
         - :class:`AnyLiteralDefaultType`
     """
     _one_of = [
-        ExtendedSchemaNode(Float(), description="Literal data type representing a floating point number."),
-        ExtendedSchemaNode(Integer(), description="Literal data type representing an integer number."),
-        ExtendedSchemaNode(Boolean(), description="Literal data type representing a boolean flag."),
-        ExtendedSchemaNode(String(), description="Literal data type representing a generic string."),
+        ExtendedSchemaNode(
+            Float(),
+            title="LiteralDataFloat",
+            description="Literal data type representing a floating point number.",
+        ),
+        ExtendedSchemaNode(
+            Integer(),
+            title="LiteralDataInteger",
+            description="Literal data type representing an integer number.",
+        ),
+        ExtendedSchemaNode(
+            Boolean(),
+            title="LiteralDataBoolean",
+            description="Literal data type representing a boolean flag.",
+        ),
+        ExtendedSchemaNode(
+            # FIXME: support byte/binary type (string + format:byte) ?
+            #   https://github.com/opengeospatial/ogcapi-processes/blob/master/openapi/schemas/processes-core/binaryInputValue.yaml
+            #   see if we can use 'encoding' parameter available for below 'String' schema-type to handle this?
+            String(allow_empty=True),  # valid to submit a process with empty string
+            title="LiteralDataString",
+            description="Literal data type representing a generic string.",
+        ),
     ]
 
 
@@ -3490,14 +3507,9 @@ class ExecuteInputFile(AnyOfKeywordSchema):
 #   https://github.com/opengeospatial/ogcapi-processes/blob/master/openapi/schemas/processes-core/binaryInputValue.yaml
 # FIXME: does not support bbox
 #   https://github.com/opengeospatial/ogcapi-processes/blob/master/openapi/schemas/processes-core/bbox.yaml
-class ExecuteInputInlineValue(OneOfKeywordSchema):
+class ExecuteInputInlineValue(AnyLiteralType):
+    title = "ExecuteInputInlineValue"
     description = "Execute input value provided inline."
-    _one_of = [
-        ExtendedSchemaNode(Float(), title="ExecuteInputValueFloat"),
-        ExtendedSchemaNode(Integer(), title="ExecuteInputValueInteger"),
-        ExtendedSchemaNode(Boolean(), title="ExecuteInputValueBoolean"),
-        ExtendedSchemaNode(String(), title="ExecuteInputValueString"),
-    ]
 
 
 # https://github.com/opengeospatial/ogcapi-processes/blob/master/openapi/schemas/processes-core/inputValue.yaml