Pattern Validation Sometimes Fails When Encountering Empty Objects

Problem/Motivation

Sometimes during schema traversal, validation of data may fail when trying to validate empty object data against a schema. In some scenarios, provided pattern configuration containing an empty object may get cast as an empty array which then fails validation resulting in a type error containing output like the following:

Swaggest\JsonSchema\Exception\TypeException : Object expected, [] received

A hallmark of this issue is that the pattern configuration validates at the top level of the schema, but once it is being passed into the field processing logic in \Drupal\patternkit\PatternFieldProcessorPluginManager::processSchemaValues, the validation fails at a lower level when handling schema composition in \Drupal\patternkit\Schema\SchemaHelper::getCompositionSchema.

Steps to reproduce

Proposed resolution

This seems to be getting caused by inconsistent handling of encoding and decoding of JSON data which converts empty objects into empty arrays along the way as described in #3294740. Fixing that issue directly may help for all new data moving forward, but I'm concerned that existing site data may already be impacted and wouldn't be resolved easily with an automated update of the data.

As an alternative, we could approach this problem by adding more flexibility to the validation process where array values being checked against an object schema could be cast during validation. The Swaggest library offers a mechanic for this through a DataPreProcessor implementation.

User interface changes

None

API changes

New service method: SchemaFactory::getDefaultContext()

The SchemaFactory service (patternkit.schema.schema_factory), introduced in #3293860, exposes a new getDefaultContext() method that provides a globally configured default context for use in schema operations beyond the initial import completed by the existing createInstance() method on the schema factory. Without this context, any schema reference lookup operations will fail, and array-to-object conversions this issue seeks to fix will fail. Below is an example usage of this context with an existing schema:

      // Get a globally configured execution context for use in schema
      // operations. This will ensure expected options as well as configured
      // remote reference providers and data preprocessors are available.
      $context = \Drupal::service('patternkit.schema.schema_factory')->getDefaultContext();
      $validationResult = $schema->in($value, $context);