feat: track logical paths in recursive validation functions

Signed-off-by: Dan Hudlow <[email protected]>

docs/openapi-ruleset-utilities.md

@@ -371,6 +371,18 @@ composed by those schemas.
 Composed schemas **do not** include nested schemas (`property`, `additionalProperties`,
 `patternProperties`, and `items` schemas).
 
+The provided validate function is called with two arguments:
+- `schema` — the composed schema
+- `path` — the array of path segments to locate the composed schema within the resolved document
+
+The provided `validate()` function is guaranteed to be called:
+- for a schema before any of its composed schemas
+- more recently for the schema that composes it (its "composition parent") than for that schema's
+   siblings (or their descendants) in the composition tree
+
+However, it is not guaranteed that the `validate()` function is called in any particular order
+for a schema's directly composed schemas.
+
 WARNING: It is only safe to use this function for a "resolved" schema — it cannot traverse `$ref`
 references.
 
@@ -404,14 +416,30 @@ schemas.
 Nested schemas included via `allOf`, `oneOf`, and `anyOf` are validated, but composed schemas
 are not themselves validated. By default, nested schemas included via `not` are not validated.
 
+The provided validate function is called with three arguments:
+- `nestedSchema`: the nested schema
+- `path`: the array of path segments to locate the nested schema within the resolved document
+- `logicalPath`: the array of path segments to locate an instance of `nestedSchema` within an
+   instance of `schema` (the schema for which `validateNestedSchemas()` was originally called.)
+   Within the logical path, an arbitrary array item is represented by `[]` and an arbitrary
+   dictionary property is represented by `*`.
+
+The provided `validate()` function is guaranteed to be called:
+- for a schema before any of its nested schemas
+- more recently for the schema that nests it (its "nesting parent") than for that schema's
+   siblings (or their descendants) in the nesting tree
+
+However, it is not guaranteed that the `validate()` function is called in any particular order
+for a schema's directly nested schemas.
+
 WARNING: It is only safe to use this function for a "resolved" schema — it cannot traverse `$ref`
 references.
 
 #### Parameters
 
 - **`schema`** `<object>`: simple or composite OpenAPI 3.x schema object
 - **`path`** `<Array>`: path array for the provided schema
-- **`validate`** `<function>`: a `(schema, path) => errors` function to validate a simple schema
+- **`validate`** `<function>`: a `(schema, path, logicalPath) => errors` function to validate a simple schema
 - **`includeSelf`** `<boolean>`: validate the provided schema in addition to its nested schemas (defaults to `true`)
 - **`includeNot`** `<boolean>`: validate schemas composed with `not` (defaults to `false`)
 
@@ -433,14 +461,35 @@ Subschemas include property schemas, 'additionalProperties', and 'patternPropert
 (such as those in an 'allOf', 'anyOf' or 'oneOf' property), plus all subschemas
 of those schemas.
 
+The provided `validate()` function is called with three arguments:
+- `subschema`: the composed or nested schema
+- `path`: the array of path segments to locate the subschema within the resolved document
+- `logicalPath`: the array of path segments to locate an instance of `subschema` within an
+   instance of `schema` (the schema for which `validateSubschemas()` was originally called.)
+   Within the logical path, an arbitrary array item is represented by `[]` and an arbitrary
+   dictionary property is represented by `*`.
+
+The provided `validate()` function is guaranteed to be called:
+- for a schema before any of its composed schemas
+- for a schema before any of its nested schemas
+- more recently for the schema that composes it (its "composition parent") than for that schema's
+   siblings (or their descendants) in the portion of composition tree local to the composition
+   parent's branch of the nesting tree
+- more recently for the schema that nests it (its "nesting parent") than for that schema's
+   siblings (or their descendants) in the portion of nesting tree local to the nesting parent's
+   branch of the composition tree
+
+However, it is not guaranteed that the `validate()` function is called in any particular order
+for a schema's directly composed or directly nested schemas.
+
 WARNING: It is only safe to use this function for a "resolved" schema — it cannot traverse `$ref`
 references.
 
 #### Parameters
 
 - **`schema`** `<object>`: simple or composite OpenAPI 3.x schema object
 - **`path`** `<Array>`: path array for the provided schema
-- **`validate`** `<function>`: a `(schema, path) => errors` function to validate a simple schema
+- **`validate`** `<function>`: a `(schema, path, logicalPath) => errors` function to validate a simple schema
 - **`includeSelf`** `<boolean>`: validate the provided schema in addition to its subschemas (defaults to `true`)
 - **`includeNot`** `<boolean>`: validate schemas composed with `not` (defaults to `true`)
 

packages/utilities/src/utils/schema-path.js

@@ -0,0 +1,53 @@
+/**
+ * Copyright 2024 IBM Corporation.
+ * SPDX-License-Identifier: Apache2.0
+ */
+
+// Necessary to get exceptions thrown for attempts to modify frozen objects
+'use strict';
+
+/**
+ * @private
+ */
+class SchemaPath extends Array {
+  constructor(physical, logical = []) {
+    super(...physical);
+    this.logical = Object.freeze([...logical]);
+    Object.freeze(this);
+  }
+
+  withProperty(name) {
+    return new SchemaPath(
+      [...this, 'properties', name],
+      [...this.logical, name]
+    );
+  }
+
+  withAdditionalProperty() {
+    return new SchemaPath(
+      [...this, 'additionalProperties'],
+      [...this.logical, '*']
+    );
+  }
+
+  withPatternProperty(pattern) {
+    return new SchemaPath(
+      [...this, 'patternProperties', pattern],
+      [...this.logical, '*']
+    );
+  }
+
+  withArrayItem() {
+    return new SchemaPath([...this, 'items'], [...this.logical, '[]']);
+  }
+
+  withApplicator(applicator, index) {
+    return new SchemaPath([...this, applicator, String(index)], this.logical);
+  }
+
+  withNot() {
+    return new SchemaPath([...this, 'not'], this.logical);
+  }
+}
+
+module.exports = SchemaPath;

packages/utilities/src/utils/validate-composed-schemas.js

@@ -3,6 +3,11 @@
  * SPDX-License-Identifier: Apache2.0
  */
 
+/**
+ * @private
+ */
+const SchemaPath = require('./schema-path');
+
 /**
  * Performs validation on a schema and all of its composed schemas.
  *
@@ -19,6 +24,18 @@
  * Composed schemas **do not** include nested schemas (`property`, `additionalProperties`,
  * `patternProperties`, and `items` schemas).
  *
+ * The provided validate function is called with two arguments:
+ * - `schema` — the composed schema
+ * - `path` — the array of path segments to locate the composed schema within the resolved document
+ *
+ * The provided `validate()` function is guaranteed to be called:
+ * - for a schema before any of its composed schemas
+ * - more recently for the schema that composes it (its "composition parent") than for that schema's
+ *    siblings (or their descendants) in the composition tree
+ *
+ * However, it is not guaranteed that the `validate()` function is called in any particular order
+ * for a schema's directly composed schemas.
+ *
  * WARNING: It is only safe to use this function for a "resolved" schema — it cannot traverse `$ref`
  * references.
  * @param {object} schema simple or composite OpenAPI 3.x schema object
@@ -30,7 +47,7 @@
  */
 function validateComposedSchemas(
   schema,
-  path,
+  p, // internally, we pass a SchemaPath to this, but the external contract is just an array
   validate,
   includeSelf = true,
   includeNot = true
@@ -42,22 +59,31 @@ function validateComposedSchemas(
   }
 
   const errors = [];
+  const path = p instanceof SchemaPath ? p : new SchemaPath(p);
 
   if (includeSelf) {
-    errors.push(...validate(schema, path));
+    // We intentionally do not use `path.logical` here because the documented external contract
+    // for `validateComposedSchemas()` is that the validate function is called with two arguments.
+    // Tracking the logical path is only useful when `validateComposedSchemas()` is embedded in
+    // another utility that can pass it an instance of `SchemaPath` (which is not exported).
+    errors.push(...validate(schema, [...path], p?.logical));
   }
 
   if (includeNot && schema.not) {
     errors.push(
-      ...validateComposedSchemas(schema.not, [...path, 'not'], validate)
+      ...validateComposedSchemas(schema.not, path.withNot(), validate)
     );
   }
 
   for (const applicatorType of ['allOf', 'oneOf', 'anyOf']) {
     if (Array.isArray(schema[applicatorType])) {
       schema[applicatorType].forEach((s, i) => {
         errors.push(
-          ...validateComposedSchemas(s, [...path, applicatorType, i], validate)
+          ...validateComposedSchemas(
+            s,
+            path.withApplicator(applicatorType, i),
+            validate
+          )
         );
       });
     }

packages/utilities/src/utils/validate-nested-schemas.js

@@ -7,6 +7,10 @@
  * @private
  */
 const isObject = require('./is-object');
+/**
+ * @private
+ */
+const SchemaPath = require('./schema-path');
 
 /**
  * Performs validation on a schema and all of its nested schemas.
@@ -27,26 +31,42 @@ const isObject = require('./is-object');
  * Nested schemas included via `allOf`, `oneOf`, and `anyOf` are validated, but composed schemas
  * are not themselves validated. By default, nested schemas included via `not` are not validated.
  *
+ * The provided validate function is called with three arguments:
+ * - `nestedSchema`: the nested schema
+ * - `path`: the array of path segments to locate the nested schema within the resolved document
+ * - `logicalPath`: the array of path segments to locate an instance of `nestedSchema` within an
+ *    instance of `schema` (the schema for which `validateNestedSchemas()` was originally called.)
+ *    Within the logical path, an arbitrary array item is represented by `[]` and an arbitrary
+ *    dictionary property is represented by `*`.
+ *
+ * The provided `validate()` function is guaranteed to be called:
+ * - for a schema before any of its nested schemas
+ * - more recently for the schema that nests it (its "nesting parent") than for that schema's
+ *    siblings (or their descendants) in the nesting tree
+ *
+ * However, it is not guaranteed that the `validate()` function is called in any particular order
+ * for a schema's directly nested schemas.
+ *
  * WARNING: It is only safe to use this function for a "resolved" schema — it cannot traverse `$ref`
  * references.
  * @param {object} schema simple or composite OpenAPI 3.x schema object
  * @param {Array} path path array for the provided schema
- * @param {Function} validate a `(schema, path) => errors` function to validate a simple schema
+ * @param {Function} validate a `(schema, path, logicalPath) => errors` function to validate a simple schema
  * @param {boolean} includeSelf validate the provided schema in addition to its nested schemas (defaults to `true`)
  * @param {boolean} includeNot validate schemas composed with `not` (defaults to `false`)
  * @returns {Array} validation errors
  */
 function validateNestedSchemas(
   schema,
-  path,
+  p, // internally, we pass a SchemaPath to this, but the external contract is just an array
   validate,
   includeSelf = true,
   includeNot = false
 ) {
   // Make sure 'schema' is an object.
   if (!isObject(schema)) {
     throw new Error(
-      `the entity at location ${path.join('.')} must be a schema object`
+      `the entity at location ${p.join('.')} must be a schema object`
     );
   }
 
@@ -57,17 +77,18 @@ function validateNestedSchemas(
   }
 
   const errors = [];
+  const path = p instanceof SchemaPath ? p : new SchemaPath(p);
 
   if (includeSelf) {
-    errors.push(...validate(schema, path));
+    errors.push(...validate(schema, [...path], [...path.logical]));
   }
 
   if (schema.properties) {
     for (const property of Object.entries(schema.properties)) {
       errors.push(
         ...validateNestedSchemas(
           property[1],
-          [...path, 'properties', property[0]],
+          path.withProperty(property[0]),
           validate,
           true,
           includeNot
@@ -80,7 +101,7 @@ function validateNestedSchemas(
     errors.push(
       ...validateNestedSchemas(
         schema.items,
-        [...path, 'items'],
+        path.withArrayItem(),
         validate,
         true,
         includeNot
@@ -95,19 +116,36 @@ function validateNestedSchemas(
     errors.push(
       ...validateNestedSchemas(
         schema.additionalProperties,
-        [...path, 'additionalProperties'],
+        path.withAdditionalProperty(),
         validate,
         true,
         includeNot
       )
     );
   }
 
+  if (
+    schema.patternProperties &&
+    typeof schema.patternProperties === 'object'
+  ) {
+    for (const entry of Object.entries(schema.patternProperties)) {
+      errors.push(
+        ...validateNestedSchemas(
+          entry[1],
+          path.withPatternProperty(entry[0]),
+          validate,
+          true,
+          includeNot
+        )
+      );
+    }
+  }
+
   if (includeNot && schema.not) {
     errors.push(
       ...validateNestedSchemas(
         schema.not,
-        [...path, 'not'],
+        path.withNot(),
         validate,
         false,
         includeNot
@@ -121,7 +159,7 @@ function validateNestedSchemas(
         errors.push(
           ...validateNestedSchemas(
             s,
-            [...path, applicatorType, i],
+            path.withApplicator(applicatorType, i),
             validate,
             false,
             includeNot
@@ -131,23 +169,6 @@ function validateNestedSchemas(
     }
   }
 
-  if (
-    schema.patternProperties &&
-    typeof schema.patternProperties === 'object'
-  ) {
-    for (const entry of Object.entries(schema.patternProperties)) {
-      errors.push(
-        ...validateNestedSchemas(
-          entry[1],
-          [...path, 'patternProperties', entry[0]],
-          validate,
-          true,
-          includeNot
-        )
-      );
-    }
-  }
-
   return errors;
 }