Implements a C++ version of JSONSchemaValidator.

chrome/common/json_schema_validator.cc

+#include "chrome/common/json_schema_validator.h"
+
+#include <cmath>
+#include <limits>
+
+#include "app/l10n_util.h"
+#include "base/string_number_conversions.h"
+#include "base/string_util.h"
+#include "base/values.h"
+#include "grit/generated_resources.h"
+
+namespace {
+
+double GetNumberValue(Value* value) {
+  double result = 0;
+  if (value->GetAsReal(&result))
+    return result;
+
+  int int_result = 0;
+  if (value->GetAsInteger(&int_result)) {
+    return int_result;
+  }
+
+  NOTREACHED() << "Unexpected value type: " << value->GetType();
+  return 0;
+}
+
+bool GetNumberFromDictionary(DictionaryValue* value, const std::string& key,
+                             double* number) {
+  if (value->GetReal(key, number))
+    return true;
+
+  int int_value = 0;
+  if (value->GetInteger(key, &int_value)) {
+    *number = int_value;
+    return true;
+  }
+
+  return false;
+}
+
+// Classifies a Value as one of the JSON schema primitive types.
+std::string GetJSONSchemaType(Value* value) {
+  switch (value->GetType()) {
+    case Value::TYPE_NULL:
+      return "null";
+    case Value::TYPE_BOOLEAN:
+      return "boolean";
+    case Value::TYPE_INTEGER:
+      return "integer";
+    case Value::TYPE_REAL: {
+      double double_value = 0;
+      value->GetAsReal(&double_value);
+      return double_value > floor(double_value) ? "number" : "integer";

[asargent_no_longer_on_chrome, 2010/11/08 23:47:18]

Should integral doubles that are > INT_MAX have type "number" instead of
"integer"?

[Aaron Boodman, 2010/11/10 19:03:35]

I don't think so. "integer" here just means "integral". I changed it to instead
check the max precisely representable integer.

+    }
+    case Value::TYPE_STRING:
+      return "string";
+    case Value::TYPE_DICTIONARY:
+      return "object";
+    case Value::TYPE_LIST:
+      return "array";
+    default:
+      NOTREACHED() << "Unexpected value type: " << value->GetType();
+      return "";
+  }
+}
+
+}  // namespace
+
+
+JSONSchemaValidator::Error::Error() {
+}
+
+JSONSchemaValidator::Error::Error(const std::string& message)
+    : path(message) {
+}
+
+JSONSchemaValidator::Error::Error(const std::string& path,
+                                  const std::string& message)
+    : path(path), message(message) {
+}
+
+
+const char JSONSchemaValidator::kUnknownTypeReference[] =
+    "Unknown schema reference: *.";
+const char JSONSchemaValidator::kInvalidChoice[] =
+    "Value does not match any valid type choices.";
+const char JSONSchemaValidator::kInvalidEnum[] =
+    "Value does not match any valid enum choices.";
+const char JSONSchemaValidator::kObjectPropertyIsRequired[] =
+    "Property is required.";
+const char JSONSchemaValidator::kUnexpectedProperty[] =
+    "Unexpected property.";
+const char JSONSchemaValidator::kArrayMinItems[] =
+    "Array must have at least * items.";
+const char JSONSchemaValidator::kArrayMaxItems[] =
+    "Array must not have more than * items.";
+const char JSONSchemaValidator::kArrayItemRequired[] =
+    "Item is required.";
+const char JSONSchemaValidator::kStringMinLength[] =
+    "String must be at least * characters long.";
+const char JSONSchemaValidator::kStringMaxLength[] =
+    "String must not be more than * characters long.";
+const char JSONSchemaValidator::kStringPattern[] =
+    "String must match the pattern: *.";
+const char JSONSchemaValidator::kInfinityNaNNotSupported[] =
+    "Value must not be infinity or NaN.";
+const char JSONSchemaValidator::kNumberMinimum[] =
+    "Value must not be less than *.";
+const char JSONSchemaValidator::kNumberMaximum[] =
+    "Value must not be greater than *.";
+const char JSONSchemaValidator::kNumberDecimalPlaces[] =
+    "Value must not have more than * decimal places.";
+const char JSONSchemaValidator::kInvalidType[] =
+    "Expected '*' but got '*'.";
+
+
+// static
+std::string JSONSchemaValidator::FormatErrorMessage(const std::string& format,
+                                                    const std::string& s1) {
+  std::string ret_val = format;
+  ReplaceFirstSubstringAfterOffset(&ret_val, 0, "*", s1);
+  return ret_val;
+}
+
+// static
+std::string JSONSchemaValidator::FormatErrorMessage(const std::string& format,
+                                                    const std::string& s1,
+                                                    const std::string& s2) {
+  std::string ret_val = format;
+  ReplaceFirstSubstringAfterOffset(&ret_val, 0, "*", s1);
+  ReplaceFirstSubstringAfterOffset(&ret_val, 0, "*", s2);
+  return ret_val;
+}
+
+JSONSchemaValidator::JSONSchemaValidator(DictionaryValue* schema)
+    : schema_root_(schema), default_allow_additional_properties_(false) {
+}
+
+JSONSchemaValidator::JSONSchemaValidator(DictionaryValue* schema,
+                                         ListValue* types)
+    : schema_root_(schema), default_allow_additional_properties_(false) {
+  if (!types)
+    return;
+
+  for (size_t i = 0; i < types->GetSize(); ++i) {
+    DictionaryValue* type = NULL;
+    CHECK(types->GetDictionary(i, &type));
+
+    std::string id;
+    CHECK(type->GetString("id", &id));
+
+    types_[id] = type;
+  }
+}
+
+
+bool JSONSchemaValidator::Validate(Value* instance) {
+  errors_.clear();
+  Validate(instance, schema_root_, "");
+  return errors_.empty();
+}
+
+void JSONSchemaValidator::Validate(Value* instance,
+                                   DictionaryValue* schema,
+                                   const std::string& path) {
+  // If this schema defines itself as reference type, save it in this.types.
+  std::string id;
+  if (schema->GetString("id", &id))
+    types_[id] = schema;

[asargent_no_longer_on_chrome, 2010/11/08 23:47:18]

What if the schema had 2 different type definitions with the same id? Shouldn't
that be an error? If so then check that types_ doesn't already have an entry for
id here. 

[Aaron Boodman, 2010/11/10 19:03:35]

Good idea. Done.

+
+  // If the schema has a $ref property, the instance must validate against
+  // that schema. It must be present in types_ to be referenced.
+  std::string ref;
+  if (schema->GetString("$ref", &ref)) {
+    TypeMap::iterator type = types_.find(ref);
+    if (type == types_.end()) {
+      errors_.push_back(
+          Error(path, FormatErrorMessage(kUnknownTypeReference, ref)));
+    } else {
+      Validate(instance, type->second, path);
+    }
+    return;
+  }
+
+  // If the schema has a choices property, the instance must validate against at
+  // least one of the items in that array.
+  ListValue* choices = NULL;
+  if (schema->GetList("choices", &choices)) {
+    ValidateChoices(instance, choices, path);
+    return;
+  }
+
+  // If the schema has an enum property, the instance must be one of those
+  // values.
+  ListValue* enumeration = NULL;
+  if (schema->GetList("enum", &enumeration)) {
+    ValidateEnum(instance, enumeration, path);
+    return;
+  }
+
+  std::string type;
+  schema->GetString("type", &type);
+  CHECK(!type.empty());
+  if (type != "any") {
+    if (!ValidateType(instance, type, path))
+      return;
+
+    // These casts are safe because of checks in ValidateType().
+    if (type == "object")
+      ValidateObject(static_cast<DictionaryValue*>(instance), schema, path);
+    else if (type == "array")
+      ValidateArray(static_cast<ListValue*>(instance), schema, path);
+    else if (type == "string")
+      ValidateString(static_cast<StringValue*>(instance), schema, path);
+    else if (type == "number" || type == "integer")
+      ValidateNumber(instance, schema, path);
+    else if (type != "boolean" && type != "null")
+      CHECK(false) << "Unexpected type: " << type;
+  }
+}
+
+void JSONSchemaValidator::ValidateChoices(Value* instance,
+                                          ListValue* choices,
+                                          const std::string& path) {
+  size_t original_num_errors = errors_.size();
+
+  for (size_t i = 0; i < choices->GetSize(); ++i) {
+    DictionaryValue* choice = NULL;
+    CHECK(choices->GetDictionary(i, &choice));
+
+    errors_.resize(original_num_errors);

[asargent_no_longer_on_chrome, 2010/11/08 23:47:18]

It took me a moment to grok what you're doing with the resize of errors_ here -
to basically try validating against the schema of each choice in choices and
return if you get a no-error result. Consider adding a brief comment here,
and/or moving the resize of errors to the bottom of the body of the for loop
(that would also save the extra resize statement just outside the loop). 

[Aaron Boodman, 2010/11/10 19:03:35]

Good ideas. Done.

+    Validate(instance, choice, path);
+    if (errors_.size() == original_num_errors)
+      return;
+  }
+
+  errors_.resize(original_num_errors);
+  errors_.push_back(Error(path, kInvalidChoice));
+  return;
+}
+
+void JSONSchemaValidator::ValidateEnum(Value* instance,
+                                       ListValue* choices,
+                                       const std::string& path) {
+  for (size_t i = 0; i < choices->GetSize(); ++i) {
+    Value* choice = NULL;
+    choices->Get(i, &choice);

[asargent_no_longer_on_chrome, 2010/11/08 23:47:18]

should you DCHECK that choice is not null after this call to Get?

[Aaron Boodman, 2010/11/10 19:03:35]

Done.

+    switch (choice->GetType()) {
+      case Value::TYPE_NULL:
+      case Value::TYPE_BOOLEAN:
+      case Value::TYPE_STRING:
+        if (instance->Equals(choice))
+          return;
+
+      case Value::TYPE_INTEGER:
+      case Value::TYPE_REAL:
+        if (instance->IsType(Value::TYPE_INTEGER) ||
+            instance->IsType(Value::TYPE_REAL)) {
+          if (GetNumberValue(choice) == GetNumberValue(instance))
+            return;
+        }
+
+      default:
+        NOTREACHED() << "Unexpected type in enum: " << choice->GetType();
+    }
+  }
+
+  errors_.push_back(Error(path, kInvalidEnum));
+}
+
+void JSONSchemaValidator::ValidateObject(DictionaryValue* instance,
+                                         DictionaryValue* schema,
+                                         const std::string& path) {
+  DictionaryValue* properties = NULL;
+  CHECK(schema->GetDictionary("properties", &properties));
+  for (DictionaryValue::key_iterator key = properties->begin_keys();
+       key != properties->end_keys(); ++key) {
+    std::string prop_path = path.empty() ? *key : (path + "." + *key);
+    DictionaryValue* prop_schema = NULL;
+    CHECK(properties->GetDictionary(*key, &prop_schema));
+
+    Value* prop_value = NULL;
+    if (instance->Get(*key, &prop_value)) {
+      Validate(prop_value, prop_schema, prop_path);
+    } else {
+      // Properties are required unless there is an optional field set to
+      // 'true'.
+      bool is_optional = false;
+      prop_schema->GetBoolean("optional", &is_optional);
+      if (!is_optional) {
+        errors_.push_back(Error(prop_path, kObjectPropertyIsRequired));
+      }
+    }
+  }
+
+  DictionaryValue* additional_properties_schema = NULL;
+  if (SchemaAllowsAnyAdditionalItems(schema, &additional_properties_schema))
+    return;
+
+  // Validate additional properties.
+  for (DictionaryValue::key_iterator key = instance->begin_keys();
+       key != instance->end_keys(); ++key) {
+    if (properties->HasKey(*key))
+      continue;
+
+    std::string prop_path = path.empty() ? *key : path + "." + *key;
+    if (!additional_properties_schema) {
+      errors_.push_back(Error(prop_path, kUnexpectedProperty));
+    } else {
+      Value* prop_value = NULL;
+      CHECK(instance->Get(*key, &prop_value));
+      Validate(prop_value, additional_properties_schema, prop_path);
+    }
+  }
+}
+
+void JSONSchemaValidator::ValidateArray(ListValue* instance,
+                                        DictionaryValue* schema,
+                                        const std::string& path) {
+  DictionaryValue* single_type = NULL;
+  size_t instance_size = instance->GetSize();
+  if (schema->GetDictionary("items", &single_type)) {
+    int min_items = 0;
+    if (single_type->GetInteger("minItems", &min_items)) {
+      CHECK(min_items >= 0);
+      if (instance_size < static_cast<size_t>(min_items)) {
+        errors_.push_back(Error(path, FormatErrorMessage(
+            kArrayMinItems, base::IntToString(min_items))));
+      }
+    }
+
+    int max_items = 0;
+    if (single_type->GetInteger("maxItems", &max_items)) {
+      CHECK(max_items >= 0);
+      if (instance_size > static_cast<size_t>(max_items)) {
+        errors_.push_back(Error(path, FormatErrorMessage(
+            kArrayMaxItems, base::IntToString(max_items))));
+      }
+    }
+
+    // If the items property is a single schema, each item in the array must
+    // validate against that schema.
+    for (size_t i = 0; i < instance_size; ++i) {
+      Value* item = NULL;
+      CHECK(instance->Get(i, &item));
+      std::string i_str = base::UintToString(i);
+      std::string item_path = path.empty() ? i_str : (path + "." + i_str);
+      Validate(item, single_type, item_path);
+    }
+
+    return;
+  }
+
+  // Otherwise, the list must be a tuple type, where each item in the list has a
+  // particular schema.
+  ListValue* tuple_type = NULL;

[asargent_no_longer_on_chrome, 2010/11/08 23:47:18]

Consider splitting the rest of the code from here down into a separate method. 

[Aaron Boodman, 2010/11/10 19:03:35]

good idea. done.

+  CHECK(schema->GetList("items", &tuple_type));
+  size_t tuple_size = tuple_type->GetSize();
+  for (size_t i = 0; i < tuple_size; ++i) {
+    std::string i_str = base::UintToString(i);
+    std::string item_path = path.empty() ? i_str : (path + "." + i_str);
+    DictionaryValue* item_schema = NULL;
+    CHECK(tuple_type->GetDictionary(i, &item_schema));
+    Value* item_value = NULL;
+    instance->Get(i, &item_value);
+    if (item_value && item_value->GetType() != Value::TYPE_NULL) {
+      Validate(item_value, item_schema, item_path);
+    } else {
+      bool is_optional = false;
+      item_schema->GetBoolean("optional", &is_optional);
+      if (!is_optional) {
+        errors_.push_back(Error(item_path, kArrayItemRequired));
+        return;
+      }
+    }
+  }
+
+  DictionaryValue* additional_properties_schema = NULL;
+  if (SchemaAllowsAnyAdditionalItems(schema, &additional_properties_schema))
+    return;
+
+  if (additional_properties_schema) {
+    // Any additional properties must validate against the additionalProperties
+    // schema.
+    for (size_t i = tuple_size; i < instance_size; ++i) {
+      std::string i_str = base::UintToString(i);
+      std::string item_path = path.empty() ? i_str : (path + "." + i_str);
+      Value* item_value = NULL;
+      CHECK(instance->Get(i, &item_value));
+      Validate(item_value, additional_properties_schema, item_path);
+    }
+  } else if (instance_size > tuple_size) {
+    errors_.push_back(Error(path, FormatErrorMessage(
+        kArrayMaxItems, base::UintToString(tuple_size))));
+  }
+}
+
+void JSONSchemaValidator::ValidateString(StringValue* instance,
+                                         DictionaryValue* schema,
+                                         const std::string& path) {
+  std::string value;
+  CHECK(instance->GetAsString(&value));
+
+  int min_length = 0;
+  if (schema->GetInteger("minLength", &min_length)) {
+    CHECK(min_length >= 0);
+    if (value.size() < static_cast<size_t>(min_length)) {
+      errors_.push_back(Error(path, FormatErrorMessage(
+          kStringMinLength, base::IntToString(min_length))));
+    }
+  }
+
+  int max_length = 0;
+  if (schema->GetInteger("maxLength", &max_length)) {
+    CHECK(max_length >= 0);
+    if (value.size() > static_cast<size_t>(max_length)) {
+      errors_.push_back(Error(path, FormatErrorMessage(
+          kStringMaxLength, base::IntToString(max_length))));
+    }
+  }
+
+  CHECK(!schema->HasKey("pattern")) << "Pattern is not supported.";
+}
+
+void JSONSchemaValidator::ValidateNumber(Value* instance,
+                                         DictionaryValue* schema,
+                                         const std::string& path) {
+  double value = GetNumberValue(instance);
+
+  if (value == std::numeric_limits<double>::infinity() ||
+      value == std::numeric_limits<double>::quiet_NaN() ||
+      value == std::numeric_limits<double>::signaling_NaN()) {
+    errors_.push_back(Error(path, kInfinityNaNNotSupported));
+    return;
+  }
+
+  double minimum = 0;
+  if (GetNumberFromDictionary(schema, "minimum", &minimum)) {
+    if (value < minimum)
+      errors_.push_back(Error(path, FormatErrorMessage(
+          kNumberMinimum, base::DoubleToString(minimum))));
+  }
+
+  double maximum = 0;
+  if (GetNumberFromDictionary(schema, "maximum", &maximum)) {
+    if (value > maximum)
+      errors_.push_back(Error(path, FormatErrorMessage(
+          kNumberMaximum, base::DoubleToString(maximum))));
+  }
+
+  int max_decimal = 0;
+  if (schema->GetInteger("maxDecimal", &max_decimal)) {
+    double int_part = 0;
+    double fraction_part = modf(value * max_decimal, &int_part);
+    if (fraction_part > 0)
+      errors_.push_back(Error(path, FormatErrorMessage(
+          kNumberDecimalPlaces, base::IntToString(max_decimal))));
+  }
+}
+
+bool JSONSchemaValidator::ValidateType(Value* instance,
+                                       const std::string& expected_type,
+                                       const std::string& path) {
+  std::string actual_type = GetJSONSchemaType(instance);
+  if (expected_type == actual_type ||
+      (expected_type == "number" && actual_type == "integer")) {
+    return true;
+  } else {
+    errors_.push_back(Error(path, FormatErrorMessage(
+        kInvalidType, expected_type, actual_type)));
+    return false;
+  }
+}
+
+bool JSONSchemaValidator::SchemaAllowsAnyAdditionalItems(
+    DictionaryValue* schema, DictionaryValue** additional_properties_schema) {
+  // If the validator allows additional properties globally, and this schema
+  // doesn't override, then we can exit early.
+  schema->GetDictionary("additionalProperties", additional_properties_schema);
+
+  if (*additional_properties_schema) {
+    std::string additional_properties_type("any");
+    CHECK((*additional_properties_schema)->GetString(
+        "type", &additional_properties_type));
+    return additional_properties_type == "any";
+  } else {
+    return default_allow_additional_properties_;
+  }
+}