Add strictness to Schema::Validate()

components/policy/core/common/schema.cc

 // Copyright 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "components/policy/core/common/schema.h"
 
 #include <algorithm>
 #include <climits>
 #include <map>
 #include <utility>
@@ skipping 64 matching lines @@
   };
   for (size_t i = 0; i < ARRAYSIZE_UNSAFE(kSchemaToValueTypeMap); ++i) {
     if (kSchemaToValueTypeMap[i].schema_type == type_string) {
       *type = kSchemaToValueTypeMap[i].value_type;
       return true;
     }
   }
   return false;
 }
 
+bool StrategyAllowInvalidOnTopLevel(SchemaOnErrorStrategy strategy) {
+  return strategy == SCHEMA_ALLOW_INVALID ||
+         strategy == SCHEMA_ALLOW_INVALID_TOPLEVEL ||
+         strategy == SCHEMA_ALLOW_INVALID_TOPLEVEL_AND_ALLOW_UNKNOWN;
+}
+
+bool StrategyAllowUnknownOnTopLevel(SchemaOnErrorStrategy strategy) {
+  return strategy != SCHEMA_STRICT;
+}
+
+SchemaOnErrorStrategy StrategyForNextLevel(SchemaOnErrorStrategy strategy) {
+  static SchemaOnErrorStrategy next_level_strategy[] = {
+    SCHEMA_STRICT,         // SCHEMA_STRICT
+    SCHEMA_STRICT,         // SCHEMA_ALLOW_UNKNOWN_TOPLEVEL
+    SCHEMA_ALLOW_UNKNOWN,  // SCHEMA_ALLOW_UNKNOWN
+    SCHEMA_STRICT,         // SCHEMA_ALLOW_INVALID_TOPLEVEL
+    SCHEMA_ALLOW_UNKNOWN,  // SCHEMA_ALLOW_INVALID_TOPLEVEL_AND_ALLOW_UNKNOWN
+    SCHEMA_ALLOW_INVALID,  // SCHEMA_ALLOW_INVALID
+  };
+  return next_level_strategy[(int)strategy];
+}
+
 }  // namespace
 
 // Contains the internal data representation of a Schema. This can either wrap
 // a SchemaData owned elsewhere (currently used to wrap the Chrome schema, which
 // is generated at compile time), or it can own its own SchemaData.
 class Schema::InternalStorage
     : public base::RefCountedThreadSafe<InternalStorage> {
  public:
   static scoped_refptr<const InternalStorage> Wrap(const SchemaData* data);
 
@@ skipping 507 matching lines @@
   node_ = schema.node_;
   return *this;
 }
 
 // static
 Schema Schema::Wrap(const SchemaData* data) {
   scoped_refptr<const InternalStorage> storage = InternalStorage::Wrap(data);
   return Schema(storage, storage->root_node());
 }
 
-bool Schema::Validate(const base::Value& value) const {
+bool Schema::Validate(const base::Value& value,
+                      SchemaOnErrorStrategy strategy,
+                      std::string *error) const {

[Joao da Silva, 2014/01/23 15:52:36]

std::string* error

The style guide says that the "*" goes next to the type, not the variable

[binjin, 2014/01/23 17:20:13]

Done.

   if (!valid()) {
-    // Schema not found, invalid entry.
+    *error = "The schema is invalid.";
     return false;
   }
 
-  if (!value.IsType(type()))
+  if (!value.IsType(type())) {
+    *error = "The value type doesn't match the schema type.";
     return false;
+  }
 
   const base::DictionaryValue* dict = NULL;
   const base::ListValue* list = NULL;
   int int_value;
   std::string str_value;
   if (value.GetAsDictionary(&dict)) {
     for (base::DictionaryValue::Iterator it(*dict); !it.IsAtEnd();
          it.Advance()) {
-      if (!GetProperty(it.key()).Validate(it.value()))
-        return false;
+      Schema subschema = GetProperty(it.key());
+      if (!subschema.valid()) {
+        // Unknown property was detected.
+        if (!StrategyAllowUnknownOnTopLevel(strategy)) {
+          *error = "Unknown property: " + it.key();
+          return false;
+        }
+      } else {
+        std::string sub_error;
+        if (!subschema.Validate(it.value(),
+            StrategyForNextLevel(strategy), &sub_error)) {

[Joao da Silva, 2014/01/23 15:52:36]

The style guide says that arguments should be aligned. Try using clang-format on
this line.

[binjin, 2014/01/23 17:20:13]

Done.

+          // Invalid property was detected.
+          if (!StrategyAllowInvalidOnTopLevel(strategy)) {
+            *error = sub_error;
+            return false;
+          }
+        }
+      }
     }
   } else if (value.GetAsList(&list)) {
     for (base::ListValue::const_iterator it = list->begin();
          it != list->end(); ++it) {
-      if (!*it || !GetItems().Validate(**it))
-        return false;
+      if (!*it || !GetItems().Validate(**it,
+            StrategyForNextLevel(strategy), error)) {

[Joao da Silva, 2014/01/23 15:52:36]

Run clang-format on this line too

[binjin, 2014/01/23 17:20:13]

Done.

+        // Invalid list item was detected.
+        if (!StrategyAllowInvalidOnTopLevel(strategy))
+          return false;
+      }
     }
   } else if (value.GetAsInteger(&int_value)) {
-    return node_->extra == kInvalid ||
-        ValidateIntegerRestriction(node_->extra, int_value);
+    if (node_->extra != kInvalid &&
+        !ValidateIntegerRestriction(node_->extra, int_value)) {
+      *error = "Invalid value for integer";
+      return false;
+    }
   } else if (value.GetAsString(&str_value)) {
-    return node_->extra == kInvalid ||
-        ValidateStringRestriction(node_->extra, str_value.c_str());
+    if (node_->extra != kInvalid &&
+        !ValidateStringRestriction(node_->extra, str_value.c_str())) {
+      *error = "Invalid value for string";
+      return false;
+    }
   }
 
   return true;
 }
 
+bool Schema::Normalize(base::Value* value,
+                       SchemaOnErrorStrategy strategy,
+                       std::string *error) const {
+  if (!valid()) {
+    *error = "The schema is invalid.";
+    return false;
+  }
+
+  if (!value->IsType(type())) {

[Joao da Silva, 2014/01/23 15:52:36]

We could handle INTEGER to DOUBLE promotion here, what do you think?

[binjin, 2014/01/23 17:20:13]

I added special check to allow Integer to Double promotion, but I prefer not to
do the in-place promotion here, since it need dirty hack to replace the type
value of FundamentalValue and also FundamentalValue::GetAsDouble can take care
of this later.

+    *error = "The value type doesn't match the schema type.";
+    return false;
+  }
+
+  base::DictionaryValue* dict = NULL;
+  base::ListValue* list = NULL;
+  if (value->GetAsDictionary(&dict)) {
+    std::vector<std::string> drop_list;  // Contains the keys to drop.
+    for (base::DictionaryValue::Iterator it(*dict); !it.IsAtEnd();
+         it.Advance()) {
+      Schema subschema = GetProperty(it.key());
+      if (!subschema.valid()) {
+        // Unknown property was detected.
+        if (StrategyAllowUnknownOnTopLevel(strategy)) {
+          drop_list.push_back(it.key());
+        } else {
+          *error = "Unknown property: " + it.key();
+          return false;
+        }
+      } else {
+        base::Value* sub_value = NULL;
+        dict->GetWithoutPathExpansion(it.key(), &sub_value);
+        std::string sub_error;
+        if (!subschema.Normalize(sub_value,
+              StrategyForNextLevel(strategy), &sub_error)) {

[Joao da Silva, 2014/01/23 15:52:36]

Run clang-format on this line

[binjin, 2014/01/23 17:20:13]

Done.

+          // Invalid property was detected.
+          if (StrategyAllowInvalidOnTopLevel(strategy)) {
+            drop_list.push_back(it.key());
+          } else {
+            *error = sub_error;
+            return false;
+          }
+        }
+      }
+    }
+    for (std::vector<std::string>::const_iterator it = drop_list.begin();
+         it != drop_list.end(); ++it) {
+      dict->Remove(*it, NULL);

[Joao da Silva, 2014/01/23 15:52:36]

RemoveWithoutPathExpansion

[binjin, 2014/01/23 17:20:13]

Done.

+    }
+    return true;
+  } else if (value->GetAsList(&list)) {
+    std::vector<size_t> drop_list;  // Contains the indexes to drop.
+    for (size_t index = 0; index < list->GetSize(); index++) {
+      base::Value* sub_value = NULL;
+      std::string sub_error;
+      list->Get(index, &sub_value);
+      if (!sub_value || !GetItems().Normalize(sub_value,
+            StrategyForNextLevel(strategy), &sub_error)) {

[Joao da Silva, 2014/01/23 15:52:36]

Run clang-format on this line

[binjin, 2014/01/23 17:20:13]

Done.

+        // Invalid list item was detected.
+        if (StrategyAllowInvalidOnTopLevel(strategy)) {
+          drop_list.push_back(index);
+        } else {
+          *error = sub_error;
+          return false;
+        }
+      }
+    }
+    for (std::vector<size_t>::reverse_iterator it = drop_list.rbegin();
+         it != drop_list.rend(); ++it) {
+      list->Remove(*it, NULL);

[Joao da Silva, 2014/01/23 15:52:36]

This is expensive, because it will search for the value. Use list->Erase(it,
NULL);

[binjin, 2014/01/23 17:20:13]

The type of *it is size_t, and I actually called ListValue::Remove(size_t, ...).

[Joao da Silva, 2014/01/23 20:18:47]

I was confused by the drop_list of the dictionary, which is a vector<string>.
Nevermind, this is fine.

+    }
+    return true;
+  }
+
+  return Validate(*value, strategy, error);
+}
+
 // static
 Schema Schema::Parse(const std::string& content, std::string* error) {
   // Validate as a generic JSON schema, and ignore unknown attributes; they
   // may become used in a future version of the schema format.
   scoped_ptr<base::DictionaryValue> dict = JSONSchemaValidator::IsValidSchema(
       content, JSONSchemaValidator::OPTIONS_IGNORE_UNKNOWN_ATTRIBUTES, error);
   if (!dict)
     return Schema();
 
   // Validate the main type.
@@ skipping 93 matching lines @@
   const RestrictionNode* rnode = storage_->restriction(index);
   for (int i = rnode->enumeration_restriction.offset_begin;
        i < rnode->enumeration_restriction.offset_end; i++) {
     if (strcmp(*storage_->string_enums(i), str) == 0)
       return true;
   }
   return false;
 }
 
 }  // namespace policy