Bug 2076540 - Add a skip_schema_validation option to bypass schema validation

CHANGELOG.md

@@ -5,6 +5,7 @@
 - JS/TS server: Add support for `server-deletion-request` ping. Server apps can now add `server-deletion-request` to `send_in_pings` on identifier metrics (e.g. user IDs) ([DENG-4079](https://mozilla-hub.atlassian.net/browse/DENG-4079))
 - JS: Ensure imports are sorted ([#865](https://github.com/mozilla/glean_parser/pull/865))
 - BREAKING CHANGE: Require pings defined requiring an "ohttp" uploader capability to have `metadata.include_info_sections: false` ([bug 2074263](https://bugzilla.mozilla.org/show_bug.cgi?id=2074263))
+- Add a `skip_schema_validation` option to `parse_objects` to skip schema validation of input files ([bug 2076540](https://bugzilla.mozilla.org/show_bug.cgi?id=2076540))
 
 ## 21.0.1
 

glean_parser/parser.py

@@ -95,9 +95,10 @@ def _load_file(
     if filetype not in ("metrics", "pings", "tags"):
         filetype = None
 
-    for error in validate(content, filepath):
-        content = {}
-        yield error
+    if not parser_config.get("skip_schema_validation", False):
+        for error in validate(content, filepath):
+            content = {}
+            yield error
 
     return content, filetype
 
@@ -486,6 +487,10 @@ def parse_objects(
           the input `filepaths` do not exist.
         - `interesting`: Contains an array of interesting metrics/ping files.
           Probes not included in these files will be marked as disabled.
+        - `skip_schema_validation`: Don't validate the input files against the
+          schema, including checking that `$schema` is a known schema. This is
+          only safe when they have been validated elsewhere: invalid input may
+          raise exceptions rather than produce errors.
     """
     if config is None:
         config = {}

tests/test_parser.py

@@ -727,6 +727,21 @@ def test_do_not_disable_expired():
     assert metrics["category"]["metric"].disabled is False
 
 
+def test_skip_schema_validation():
+    contents = [{"category": {"metric": {"type": "boolean", "description": 42}}}]
+    contents = [util.add_required(x) for x in contents]
+
+    all_metrics = parser.parse_objects(contents)
+    errors = list(all_metrics)
+    assert len(errors) == 1
+    assert "description" in errors[0]
+
+    all_metrics = parser.parse_objects(contents, {"skip_schema_validation": True})
+    errors = list(all_metrics)
+    assert len(errors) == 0
+    assert all_metrics.value["category"]["metric"].description == 42
+
+
 def test_interesting_disables_others():
     # Test that if we get an "interesting" config param, only the metrics in
     # the provided config files are enabled.