protobuf/objectivec/GPBDescriptor.m

1154 lines
36 KiB
Mathematica
Raw Normal View History

// Protocol Buffers - Google's data interchange format
// Copyright 2008 Google Inc. All rights reserved.
// https://developers.google.com/protocol-buffers/
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#import "GPBDescriptor_PackagePrivate.h"
#import <objc/runtime.h>
#import "GPBUtilities_PackagePrivate.h"
#import "GPBWireFormat.h"
#import "GPBMessage_PackagePrivate.h"
// Direct access is use for speed, to avoid even internally declaring things
// read/write, etc. The warning is enabled in the project to ensure code calling
// protos can turn on -Wdirect-ivar-access without issues.
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wdirect-ivar-access"
// The addresses of these variables are used as keys for objc_getAssociatedObject.
static const char kTextFormatExtraValueKey = 0;
static const char kParentClassValueKey = 0;
static const char kClassNameSuffixKey = 0;
// Utility function to generate selectors on the fly.
static SEL SelFromStrings(const char *prefix, const char *middle,
const char *suffix, BOOL takesArg) {
if (prefix == NULL && suffix == NULL && !takesArg) {
return sel_getUid(middle);
}
const size_t prefixLen = prefix != NULL ? strlen(prefix) : 0;
const size_t middleLen = strlen(middle);
const size_t suffixLen = suffix != NULL ? strlen(suffix) : 0;
size_t totalLen =
prefixLen + middleLen + suffixLen + 1; // include space for null on end.
if (takesArg) {
totalLen += 1;
}
char buffer[totalLen];
if (prefix != NULL) {
memcpy(buffer, prefix, prefixLen);
memcpy(buffer + prefixLen, middle, middleLen);
buffer[prefixLen] = (char)toupper(buffer[prefixLen]);
} else {
memcpy(buffer, middle, middleLen);
}
if (suffix != NULL) {
memcpy(buffer + prefixLen + middleLen, suffix, suffixLen);
}
if (takesArg) {
buffer[totalLen - 2] = ':';
}
// Always null terminate it.
buffer[totalLen - 1] = 0;
SEL result = sel_getUid(buffer);
return result;
}
static NSArray *NewFieldsArrayForHasIndex(int hasIndex,
NSArray *allMessageFields)
__attribute__((ns_returns_retained));
static NSArray *NewFieldsArrayForHasIndex(int hasIndex,
NSArray *allMessageFields) {
NSMutableArray *result = [[NSMutableArray alloc] init];
for (GPBFieldDescriptor *fieldDesc in allMessageFields) {
if (fieldDesc->description_->hasIndex == hasIndex) {
[result addObject:fieldDesc];
}
}
return result;
}
@implementation GPBDescriptor {
Class messageClass_;
GPBFileDescriptor *file_;
BOOL wireFormat_;
}
@synthesize messageClass = messageClass_;
@synthesize fields = fields_;
@synthesize oneofs = oneofs_;
@synthesize extensionRanges = extensionRanges_;
@synthesize extensionRangesCount = extensionRangesCount_;
@synthesize file = file_;
@synthesize wireFormat = wireFormat_;
+ (instancetype)
allocDescriptorForClass:(Class)messageClass
rootClass:(Class)rootClass
file:(GPBFileDescriptor *)file
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
fields:(void *)fieldDescriptions
fieldCount:(uint32_t)fieldCount
storageSize:(uint32_t)storageSize
flags:(GPBDescriptorInitializationFlags)flags {
// The rootClass is no longer used, but it is passed in to ensure it
// was started up during initialization also.
(void)rootClass;
NSMutableArray *fields = nil;
GPBFileSyntax syntax = file.syntax;
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
BOOL fieldsIncludeDefault =
(flags & GPBDescriptorInitializationFlag_FieldsWithDefault) != 0;
BOOL usesClassRefs =
(flags & GPBDescriptorInitializationFlag_UsesClassRefs) != 0;
BOOL proto3OptionalKnown =
(flags & GPBDescriptorInitializationFlag_Proto3OptionalKnown) != 0;
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
void *desc;
for (uint32_t i = 0; i < fieldCount; ++i) {
if (fields == nil) {
fields = [[NSMutableArray alloc] initWithCapacity:fieldCount];
}
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
// Need correctly typed pointer for array indexing below to work.
if (fieldsIncludeDefault) {
GPBMessageFieldDescriptionWithDefault *fieldDescWithDefault = fieldDescriptions;
desc = &(fieldDescWithDefault[i]);
} else {
GPBMessageFieldDescription *fieldDesc = fieldDescriptions;
desc = &(fieldDesc[i]);
}
GPBFieldDescriptor *fieldDescriptor =
[[GPBFieldDescriptor alloc] initWithFieldDescription:desc
includesDefault:fieldsIncludeDefault
usesClassRefs:usesClassRefs
proto3OptionalKnown:proto3OptionalKnown
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
syntax:syntax];
[fields addObject:fieldDescriptor];
[fieldDescriptor release];
}
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
BOOL wireFormat = (flags & GPBDescriptorInitializationFlag_WireFormat) != 0;
GPBDescriptor *descriptor = [[self alloc] initWithClass:messageClass
file:file
fields:fields
storageSize:storageSize
wireFormat:wireFormat];
[fields release];
return descriptor;
}
- (instancetype)initWithClass:(Class)messageClass
file:(GPBFileDescriptor *)file
fields:(NSArray *)fields
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
storageSize:(uint32_t)storageSize
wireFormat:(BOOL)wireFormat {
if ((self = [super init])) {
messageClass_ = messageClass;
file_ = file;
fields_ = [fields retain];
storageSize_ = storageSize;
wireFormat_ = wireFormat;
}
return self;
}
- (void)dealloc {
[fields_ release];
[oneofs_ release];
[super dealloc];
}
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
- (void)setupOneofs:(const char **)oneofNames
count:(uint32_t)count
firstHasIndex:(int32_t)firstHasIndex {
NSCAssert(firstHasIndex < 0, @"Should always be <0");
NSMutableArray *oneofs = [[NSMutableArray alloc] initWithCapacity:count];
for (uint32_t i = 0, hasIndex = firstHasIndex; i < count; ++i, --hasIndex) {
const char *name = oneofNames[i];
NSArray *fieldsForOneof = NewFieldsArrayForHasIndex(hasIndex, fields_);
NSCAssert(fieldsForOneof.count > 0,
@"No fields for this oneof? (%s:%d)", name, hasIndex);
GPBOneofDescriptor *oneofDescriptor =
[[GPBOneofDescriptor alloc] initWithName:name fields:fieldsForOneof];
[oneofs addObject:oneofDescriptor];
[oneofDescriptor release];
[fieldsForOneof release];
}
oneofs_ = oneofs;
}
- (void)setupExtraTextInfo:(const char *)extraTextFormatInfo {
// Extra info is a compile time option, so skip the work if not needed.
if (extraTextFormatInfo) {
NSValue *extraInfoValue = [NSValue valueWithPointer:extraTextFormatInfo];
for (GPBFieldDescriptor *fieldDescriptor in fields_) {
if (fieldDescriptor->description_->flags & GPBFieldTextFormatNameCustom) {
objc_setAssociatedObject(fieldDescriptor, &kTextFormatExtraValueKey,
extraInfoValue,
OBJC_ASSOCIATION_RETAIN_NONATOMIC);
}
}
}
}
- (void)setupExtensionRanges:(const GPBExtensionRange *)ranges count:(int32_t)count {
extensionRanges_ = ranges;
extensionRangesCount_ = count;
}
- (void)setupContainingMessageClass:(Class)messageClass {
objc_setAssociatedObject(self, &kParentClassValueKey,
messageClass,
OBJC_ASSOCIATION_ASSIGN);
}
- (void)setupContainingMessageClassName:(const char *)msgClassName {
// Note: Only fetch the class here, can't send messages to it because
// that could cause cycles back to this class within +initialize if
// two messages have each other in fields (i.e. - they build a graph).
Class clazz = objc_getClass(msgClassName);
NSAssert(clazz, @"Class %s not defined", msgClassName);
[self setupContainingMessageClass:clazz];
}
- (void)setupMessageClassNameSuffix:(NSString *)suffix {
if (suffix.length) {
objc_setAssociatedObject(self, &kClassNameSuffixKey,
suffix,
OBJC_ASSOCIATION_RETAIN_NONATOMIC);
}
}
- (NSString *)name {
return NSStringFromClass(messageClass_);
}
- (GPBDescriptor *)containingType {
Class parentClass = objc_getAssociatedObject(self, &kParentClassValueKey);
return [parentClass descriptor];
}
- (NSString *)fullName {
NSString *className = NSStringFromClass(self.messageClass);
GPBFileDescriptor *file = self.file;
NSString *objcPrefix = file.objcPrefix;
if (objcPrefix && ![className hasPrefix:objcPrefix]) {
NSAssert(0,
@"Class didn't have correct prefix? (%@ - %@)",
className, objcPrefix);
return nil;
}
GPBDescriptor *parent = self.containingType;
NSString *name = nil;
if (parent) {
NSString *parentClassName = NSStringFromClass(parent.messageClass);
// The generator will add _Class to avoid reserved words, drop it.
NSString *suffix = objc_getAssociatedObject(parent, &kClassNameSuffixKey);
if (suffix) {
if (![parentClassName hasSuffix:suffix]) {
NSAssert(0,
@"ParentMessage class didn't have correct suffix? (%@ - %@)",
className, suffix);
return nil;
}
parentClassName =
[parentClassName substringToIndex:(parentClassName.length - suffix.length)];
}
NSString *parentPrefix = [parentClassName stringByAppendingString:@"_"];
if (![className hasPrefix:parentPrefix]) {
NSAssert(0,
@"Class didn't have the correct parent name prefix? (%@ - %@)",
parentPrefix, className);
return nil;
}
name = [className substringFromIndex:parentPrefix.length];
} else {
name = [className substringFromIndex:objcPrefix.length];
}
// The generator will add _Class to avoid reserved words, drop it.
NSString *suffix = objc_getAssociatedObject(self, &kClassNameSuffixKey);
if (suffix) {
if (![name hasSuffix:suffix]) {
NSAssert(0,
@"Message class didn't have correct suffix? (%@ - %@)",
name, suffix);
return nil;
}
name = [name substringToIndex:(name.length - suffix.length)];
}
NSString *prefix = (parent != nil ? parent.fullName : file.package);
NSString *result;
if (prefix.length > 0) {
result = [NSString stringWithFormat:@"%@.%@", prefix, name];
} else {
result = name;
}
return result;
}
- (id)copyWithZone:(NSZone *)zone {
#pragma unused(zone)
return [self retain];
}
- (GPBFieldDescriptor *)fieldWithNumber:(uint32_t)fieldNumber {
for (GPBFieldDescriptor *descriptor in fields_) {
if (GPBFieldNumber(descriptor) == fieldNumber) {
return descriptor;
}
}
return nil;
}
- (GPBFieldDescriptor *)fieldWithName:(NSString *)name {
for (GPBFieldDescriptor *descriptor in fields_) {
if ([descriptor.name isEqual:name]) {
return descriptor;
}
}
return nil;
}
- (GPBOneofDescriptor *)oneofWithName:(NSString *)name {
for (GPBOneofDescriptor *descriptor in oneofs_) {
if ([descriptor.name isEqual:name]) {
return descriptor;
}
}
return nil;
}
@end
@implementation GPBFileDescriptor {
NSString *package_;
NSString *objcPrefix_;
GPBFileSyntax syntax_;
}
@synthesize package = package_;
@synthesize objcPrefix = objcPrefix_;
@synthesize syntax = syntax_;
- (instancetype)initWithPackage:(NSString *)package
objcPrefix:(NSString *)objcPrefix
syntax:(GPBFileSyntax)syntax {
self = [super init];
if (self) {
package_ = [package copy];
objcPrefix_ = [objcPrefix copy];
syntax_ = syntax;
}
return self;
}
- (instancetype)initWithPackage:(NSString *)package
syntax:(GPBFileSyntax)syntax {
self = [super init];
if (self) {
package_ = [package copy];
syntax_ = syntax;
}
return self;
}
- (void)dealloc {
[package_ release];
[objcPrefix_ release];
[super dealloc];
}
@end
@implementation GPBOneofDescriptor
@synthesize fields = fields_;
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
- (instancetype)initWithName:(const char *)name fields:(NSArray *)fields {
self = [super init];
if (self) {
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
name_ = name;
fields_ = [fields retain];
for (GPBFieldDescriptor *fieldDesc in fields) {
fieldDesc->containingOneof_ = self;
}
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
caseSel_ = SelFromStrings(NULL, name, "OneOfCase", NO);
}
return self;
}
- (void)dealloc {
[fields_ release];
[super dealloc];
}
- (NSString *)name {
return (NSString * _Nonnull)@(name_);
}
- (GPBFieldDescriptor *)fieldWithNumber:(uint32_t)fieldNumber {
for (GPBFieldDescriptor *descriptor in fields_) {
if (GPBFieldNumber(descriptor) == fieldNumber) {
return descriptor;
}
}
return nil;
}
- (GPBFieldDescriptor *)fieldWithName:(NSString *)name {
for (GPBFieldDescriptor *descriptor in fields_) {
if ([descriptor.name isEqual:name]) {
return descriptor;
}
}
return nil;
}
@end
uint32_t GPBFieldTag(GPBFieldDescriptor *self) {
GPBMessageFieldDescription *description = self->description_;
GPBWireFormat format;
if ((description->flags & GPBFieldMapKeyMask) != 0) {
// Maps are repeated messages on the wire.
format = GPBWireFormatForType(GPBDataTypeMessage, NO);
} else {
format = GPBWireFormatForType(description->dataType,
((description->flags & GPBFieldPacked) != 0));
}
return GPBWireFormatMakeTag(description->number, format);
}
uint32_t GPBFieldAlternateTag(GPBFieldDescriptor *self) {
GPBMessageFieldDescription *description = self->description_;
NSCAssert((description->flags & GPBFieldRepeated) != 0,
@"Only valid on repeated fields");
GPBWireFormat format =
GPBWireFormatForType(description->dataType,
((description->flags & GPBFieldPacked) == 0));
return GPBWireFormatMakeTag(description->number, format);
}
@implementation GPBFieldDescriptor {
GPBGenericValue defaultValue_;
// Message ivars
Class msgClass_;
// Enum ivars.
// If protos are generated with GenerateEnumDescriptors on then it will
// be a enumDescriptor, otherwise it will be a enumVerifier.
union {
GPBEnumDescriptor *enumDescriptor_;
GPBEnumValidationFunc enumVerifier_;
} enumHandling_;
}
@synthesize msgClass = msgClass_;
@synthesize containingOneof = containingOneof_;
- (instancetype)init {
// Throw an exception if people attempt to not use the designated initializer.
self = [super init];
if (self != nil) {
[self doesNotRecognizeSelector:_cmd];
self = nil;
}
return self;
}
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
- (instancetype)initWithFieldDescription:(void *)description
includesDefault:(BOOL)includesDefault
usesClassRefs:(BOOL)usesClassRefs
proto3OptionalKnown:(BOOL)proto3OptionalKnown
syntax:(GPBFileSyntax)syntax {
if ((self = [super init])) {
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
GPBMessageFieldDescription *coreDesc;
if (includesDefault) {
coreDesc = &(((GPBMessageFieldDescriptionWithDefault *)description)->core);
} else {
coreDesc = description;
}
description_ = coreDesc;
getSel_ = sel_getUid(coreDesc->name);
setSel_ = SelFromStrings("set", coreDesc->name, NULL, YES);
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
GPBDataType dataType = coreDesc->dataType;
BOOL isMessage = GPBDataTypeIsMessage(dataType);
BOOL isMapOrArray = GPBFieldIsMapOrArray(self);
// If proto3 optionals weren't know, compute the flag for the rest of the runtime.
if (!proto3OptionalKnown) {
// If it was...
// - proto3 syntax
// - not repeated/map
// - not in a oneof (negative has index)
// - not a message (the flag doesn't make sense for messages)
BOOL clearOnZero = ((syntax == GPBFileSyntaxProto3) &&
!isMapOrArray &&
(coreDesc->hasIndex >= 0) &&
!isMessage);
if (clearOnZero) {
coreDesc->flags |= GPBFieldClearHasIvarOnZero;
}
}
if (isMapOrArray) {
// map<>/repeated fields get a *Count property (inplace of a has*) to
// support checking if there are any entries without triggering
// autocreation.
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
hasOrCountSel_ = SelFromStrings(NULL, coreDesc->name, "_Count", NO);
} else {
// Know it is a single field; it gets has/setHas selectors if...
// - not in a oneof (negative has index)
// - not clearing on zero
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
if ((coreDesc->hasIndex >= 0) &&
((coreDesc->flags & GPBFieldClearHasIvarOnZero) == 0)) {
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
hasOrCountSel_ = SelFromStrings("has", coreDesc->name, NULL, NO);
setHasSel_ = SelFromStrings("setHas", coreDesc->name, NULL, YES);
}
}
// Extra type specific data.
if (isMessage) {
// Note: Only fetch the class here, can't send messages to it because
// that could cause cycles back to this class within +initialize if
// two messages have each other in fields (i.e. - they build a graph).
if (usesClassRefs) {
msgClass_ = coreDesc->dataTypeSpecific.clazz;
} else {
// Backwards compatibility for sources generated with older protoc.
const char *className = coreDesc->dataTypeSpecific.className;
msgClass_ = objc_getClass(className);
NSAssert(msgClass_, @"Class %s not defined", className);
}
} else if (dataType == GPBDataTypeEnum) {
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
if ((coreDesc->flags & GPBFieldHasEnumDescriptor) != 0) {
enumHandling_.enumDescriptor_ =
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
coreDesc->dataTypeSpecific.enumDescFunc();
} else {
enumHandling_.enumVerifier_ =
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
coreDesc->dataTypeSpecific.enumVerifier;
}
}
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
// Non map<>/repeated fields can have defaults in proto2 syntax.
if (!isMapOrArray && includesDefault) {
defaultValue_ = ((GPBMessageFieldDescriptionWithDefault *)description)->defaultValue;
if (dataType == GPBDataTypeBytes) {
// Data stored as a length prefixed (network byte order) c-string in
// descriptor structure.
const uint8_t *bytes = (const uint8_t *)defaultValue_.valueData;
if (bytes) {
uint32_t length;
memcpy(&length, bytes, sizeof(length));
length = ntohl(length);
bytes += sizeof(length);
defaultValue_.valueData =
[[NSData alloc] initWithBytes:bytes length:length];
}
}
}
}
return self;
}
- (void)dealloc {
if (description_->dataType == GPBDataTypeBytes &&
!(description_->flags & GPBFieldRepeated)) {
[defaultValue_.valueData release];
}
[super dealloc];
}
- (GPBDataType)dataType {
return description_->dataType;
}
- (BOOL)hasDefaultValue {
return (description_->flags & GPBFieldHasDefaultValue) != 0;
}
- (uint32_t)number {
return description_->number;
}
- (NSString *)name {
return (NSString * _Nonnull)@(description_->name);
}
- (BOOL)isRequired {
return (description_->flags & GPBFieldRequired) != 0;
}
- (BOOL)isOptional {
return (description_->flags & GPBFieldOptional) != 0;
}
- (GPBFieldType)fieldType {
GPBFieldFlags flags = description_->flags;
if ((flags & GPBFieldRepeated) != 0) {
return GPBFieldTypeRepeated;
} else if ((flags & GPBFieldMapKeyMask) != 0) {
return GPBFieldTypeMap;
} else {
return GPBFieldTypeSingle;
}
}
- (GPBDataType)mapKeyDataType {
switch (description_->flags & GPBFieldMapKeyMask) {
case GPBFieldMapKeyInt32:
return GPBDataTypeInt32;
case GPBFieldMapKeyInt64:
return GPBDataTypeInt64;
case GPBFieldMapKeyUInt32:
return GPBDataTypeUInt32;
case GPBFieldMapKeyUInt64:
return GPBDataTypeUInt64;
case GPBFieldMapKeySInt32:
return GPBDataTypeSInt32;
case GPBFieldMapKeySInt64:
return GPBDataTypeSInt64;
case GPBFieldMapKeyFixed32:
return GPBDataTypeFixed32;
case GPBFieldMapKeyFixed64:
return GPBDataTypeFixed64;
case GPBFieldMapKeySFixed32:
return GPBDataTypeSFixed32;
case GPBFieldMapKeySFixed64:
return GPBDataTypeSFixed64;
case GPBFieldMapKeyBool:
return GPBDataTypeBool;
case GPBFieldMapKeyString:
return GPBDataTypeString;
default:
NSAssert(0, @"Not a map type");
return GPBDataTypeInt32; // For lack of anything better.
}
}
- (BOOL)isPackable {
return (description_->flags & GPBFieldPacked) != 0;
}
- (BOOL)isValidEnumValue:(int32_t)value {
NSAssert(description_->dataType == GPBDataTypeEnum,
@"Field Must be of type GPBDataTypeEnum");
if (description_->flags & GPBFieldHasEnumDescriptor) {
return enumHandling_.enumDescriptor_.enumVerifier(value);
} else {
return enumHandling_.enumVerifier_(value);
}
}
- (GPBEnumDescriptor *)enumDescriptor {
if (description_->flags & GPBFieldHasEnumDescriptor) {
return enumHandling_.enumDescriptor_;
} else {
return nil;
}
}
- (GPBGenericValue)defaultValue {
// Depends on the fact that defaultValue_ is initialized either to "0/nil" or
// to an actual defaultValue in our initializer.
GPBGenericValue value = defaultValue_;
if (!(description_->flags & GPBFieldRepeated)) {
// We special handle data and strings. If they are nil, we replace them
// with empty string/empty data.
GPBDataType type = description_->dataType;
if (type == GPBDataTypeBytes && value.valueData == nil) {
value.valueData = GPBEmptyNSData();
} else if (type == GPBDataTypeString && value.valueString == nil) {
value.valueString = @"";
}
}
return value;
}
- (NSString *)textFormatName {
if ((description_->flags & GPBFieldTextFormatNameCustom) != 0) {
NSValue *extraInfoValue =
objc_getAssociatedObject(self, &kTextFormatExtraValueKey);
// Support can be left out at generation time.
if (!extraInfoValue) {
return nil;
}
const uint8_t *extraTextFormatInfo = [extraInfoValue pointerValue];
return GPBDecodeTextFormatName(extraTextFormatInfo, GPBFieldNumber(self),
self.name);
}
// The logic here has to match SetCommonFieldVariables() from
// objectivec_field.cc in the proto compiler.
NSString *name = self.name;
NSUInteger len = [name length];
// Remove the "_p" added to reserved names.
if ([name hasSuffix:@"_p"]) {
name = [name substringToIndex:(len - 2)];
len = [name length];
}
// Remove "Array" from the end for repeated fields.
if (((description_->flags & GPBFieldRepeated) != 0) &&
[name hasSuffix:@"Array"]) {
name = [name substringToIndex:(len - 5)];
len = [name length];
}
// Groups vs. other fields.
if (description_->dataType == GPBDataTypeGroup) {
// Just capitalize the first letter.
unichar firstChar = [name characterAtIndex:0];
if (firstChar >= 'a' && firstChar <= 'z') {
NSString *firstCharString =
[NSString stringWithFormat:@"%C", (unichar)(firstChar - 'a' + 'A')];
NSString *result =
[name stringByReplacingCharactersInRange:NSMakeRange(0, 1)
withString:firstCharString];
return result;
}
return name;
} else {
// Undo the CamelCase.
NSMutableString *result = [NSMutableString stringWithCapacity:len];
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
for (uint32_t i = 0; i < len; i++) {
unichar c = [name characterAtIndex:i];
if (c >= 'A' && c <= 'Z') {
if (i > 0) {
[result appendFormat:@"_%C", (unichar)(c - 'A' + 'a')];
} else {
[result appendFormat:@"%C", c];
}
} else {
[result appendFormat:@"%C", c];
}
}
return result;
}
}
@end
@implementation GPBEnumDescriptor {
NSString *name_;
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
// valueNames_ is a single c string with all of the value names appended
// together, each null terminated. -calcValueNameOffsets fills in
// nameOffsets_ with the offsets to allow quicker access to the individual
// names.
const char *valueNames_;
const int32_t *values_;
GPBEnumValidationFunc enumVerifier_;
const uint8_t *extraTextFormatInfo_;
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
uint32_t *nameOffsets_;
uint32_t valueCount_;
}
@synthesize name = name_;
@synthesize enumVerifier = enumVerifier_;
+ (instancetype)
allocDescriptorForName:(NSString *)name
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
valueNames:(const char *)valueNames
values:(const int32_t *)values
count:(uint32_t)valueCount
enumVerifier:(GPBEnumValidationFunc)enumVerifier {
GPBEnumDescriptor *descriptor = [[self alloc] initWithName:name
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
valueNames:valueNames
values:values
count:valueCount
enumVerifier:enumVerifier];
return descriptor;
}
+ (instancetype)
allocDescriptorForName:(NSString *)name
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
valueNames:(const char *)valueNames
values:(const int32_t *)values
count:(uint32_t)valueCount
enumVerifier:(GPBEnumValidationFunc)enumVerifier
extraTextFormatInfo:(const char *)extraTextFormatInfo {
// Call the common case.
GPBEnumDescriptor *descriptor = [self allocDescriptorForName:name
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
valueNames:valueNames
values:values
count:valueCount
enumVerifier:enumVerifier];
// Set the extra info.
descriptor->extraTextFormatInfo_ = (const uint8_t *)extraTextFormatInfo;
return descriptor;
}
- (instancetype)initWithName:(NSString *)name
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
valueNames:(const char *)valueNames
values:(const int32_t *)values
count:(uint32_t)valueCount
enumVerifier:(GPBEnumValidationFunc)enumVerifier {
if ((self = [super init])) {
name_ = [name copy];
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
valueNames_ = valueNames;
values_ = values;
valueCount_ = valueCount;
enumVerifier_ = enumVerifier;
}
return self;
}
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
- (void)dealloc {
[name_ release];
if (nameOffsets_) free(nameOffsets_);
[super dealloc];
}
- (void)calcValueNameOffsets {
@synchronized(self) {
if (nameOffsets_ != NULL) {
return;
}
uint32_t *offsets = malloc(valueCount_ * sizeof(uint32_t));
if (!offsets) return;
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
const char *scan = valueNames_;
for (uint32_t i = 0; i < valueCount_; ++i) {
offsets[i] = (uint32_t)(scan - valueNames_);
while (*scan != '\0') ++scan;
++scan; // Step over the null.
}
nameOffsets_ = offsets;
}
}
- (NSString *)enumNameForValue:(int32_t)number {
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
for (uint32_t i = 0; i < valueCount_; ++i) {
if (values_[i] == number) {
return [self getEnumNameForIndex:i];
}
}
return nil;
}
- (BOOL)getValue:(int32_t *)outValue forEnumName:(NSString *)name {
// Must have the prefix.
NSUInteger prefixLen = name_.length + 1;
if ((name.length <= prefixLen) || ![name hasPrefix:name_] ||
([name characterAtIndex:prefixLen - 1] != '_')) {
return NO;
}
// Skip over the prefix.
const char *nameAsCStr = [name UTF8String];
nameAsCStr += prefixLen;
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
if (nameOffsets_ == NULL) [self calcValueNameOffsets];
if (nameOffsets_ == NULL) return NO;
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
// Find it.
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
for (uint32_t i = 0; i < valueCount_; ++i) {
const char *valueName = valueNames_ + nameOffsets_[i];
if (strcmp(nameAsCStr, valueName) == 0) {
if (outValue) {
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
*outValue = values_[i];
}
return YES;
}
}
return NO;
}
- (BOOL)getValue:(int32_t *)outValue forEnumTextFormatName:(NSString *)textFormatName {
if (nameOffsets_ == NULL) [self calcValueNameOffsets];
if (nameOffsets_ == NULL) return NO;
for (uint32_t i = 0; i < valueCount_; ++i) {
NSString *valueTextFormatName = [self getEnumTextFormatNameForIndex:i];
if ([valueTextFormatName isEqual:textFormatName]) {
if (outValue) {
*outValue = values_[i];
}
return YES;
}
}
return NO;
}
- (NSString *)textFormatNameForValue:(int32_t)number {
// Find the EnumValue descriptor and its index.
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
BOOL foundIt = NO;
uint32_t valueDescriptorIndex;
for (valueDescriptorIndex = 0; valueDescriptorIndex < valueCount_;
++valueDescriptorIndex) {
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
if (values_[valueDescriptorIndex] == number) {
foundIt = YES;
break;
}
}
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
if (!foundIt) {
return nil;
}
return [self getEnumTextFormatNameForIndex:valueDescriptorIndex];
}
- (uint32_t)enumNameCount {
return valueCount_;
}
- (NSString *)getEnumNameForIndex:(uint32_t)index {
if (nameOffsets_ == NULL) [self calcValueNameOffsets];
if (nameOffsets_ == NULL) return nil;
if (index >= valueCount_) {
return nil;
}
const char *valueName = valueNames_ + nameOffsets_[index];
NSString *fullName = [NSString stringWithFormat:@"%@_%s", name_, valueName];
return fullName;
}
- (NSString *)getEnumTextFormatNameForIndex:(uint32_t)index {
if (nameOffsets_ == NULL) [self calcValueNameOffsets];
if (nameOffsets_ == NULL) return nil;
if (index >= valueCount_) {
return nil;
}
NSString *result = nil;
// Naming adds an underscore between enum name and value name, skip that also.
const char *valueName = valueNames_ + nameOffsets_[index];
Shrink ObjC overhead (generated size and some runtime sizes) NOTE: This is a binary breaking change as structure sizes have changed size and/or order. - Drop capturing field options, no other options were captured and other mobile targeted languages don't try to capture this sort information (saved 8 bytes for every field defined (in static data and again in field descriptor instance size data). - No longer generate/compile in the messages/enums in descriptor.proto. If developers need it, they should generate it and compile it in. Reduced the overhead of the core library. - Compute the number of has_bits actually needs to avoid over reserving. - Let the boolean single fields store via a has_bit to avoid storage, makes the common cases of the instance size smaller. - Reorder some flags and down size the enums to contain the bits needed. - Reorder the items in the structures to manually ensure they are are packed better (especially when generating 64bit code - 8 bytes for every field, 16 bytes for every extension, instance sizes 8 bytes also). - Split off the structure initialization so when the default is zero, the generated static storage doesn't need to reserve the space. This is batched at the message level, so all the fields for the message have to have zero defaults to get the saves. By definition all proto3 syntax files fall into this case but it also saves space for the proto2 that use the standard defaults. (saves 8 bytes of static data for every field that had a zero default) - Don't track the enums defined by a message. Nothing in the runtime needs it and it was just generation and runtime overhead. (saves 8 bytes per enum) - Ensure EnumDescriptors are started up threadsafe in all cases. - Split some of the Descriptor initialization into multiple methods so the generated code isn't padded with lots of zero/nil args. - Change how oneof info is feed to the runtime enabling us to generate less static data (8 bytes saved per oneof for 64bit). - Change how enum value informat is capture to pack the data and only decode it if it ends up being needed. Avoids padding issues causing bloat of 64bit, and removes the needs for extra pointers in addition to the data (just the data and one pointer now).
2016-03-17 14:04:21 +00:00
NSString *shortName = @(valueName);
// See if it is in the map of special format handling.
if (extraTextFormatInfo_) {
result = GPBDecodeTextFormatName(extraTextFormatInfo_,
(int32_t)index, shortName);
}
// Logic here needs to match what objectivec_enum.cc does in the proto
// compiler.
if (result == nil) {
NSUInteger len = [shortName length];
NSMutableString *worker = [NSMutableString stringWithCapacity:len];
for (NSUInteger i = 0; i < len; i++) {
unichar c = [shortName characterAtIndex:i];
if (i > 0 && c >= 'A' && c <= 'Z') {
[worker appendString:@"_"];
}
[worker appendFormat:@"%c", toupper((char)c)];
}
result = worker;
}
return result;
}
@end
@implementation GPBExtensionDescriptor {
GPBGenericValue defaultValue_;
}
- (instancetype)initWithExtensionDescription:(GPBExtensionDescription *)desc
usesClassRefs:(BOOL)usesClassRefs {
if ((self = [super init])) {
description_ = desc;
if (!usesClassRefs) {
// Legacy without class ref support.
const char *className = description_->messageOrGroupClass.name;
if (className) {
Class clazz = objc_lookUpClass(className);
NSAssert(clazz != Nil, @"Class %s not defined", className);
description_->messageOrGroupClass.clazz = clazz;
}
const char *extendedClassName = description_->extendedClass.name;
if (extendedClassName) {
Class clazz = objc_lookUpClass(extendedClassName);
NSAssert(clazz, @"Class %s not defined", extendedClassName);
description_->extendedClass.clazz = clazz;
}
}
GPBDataType type = description_->dataType;
if (type == GPBDataTypeBytes) {
// Data stored as a length prefixed c-string in descriptor records.
const uint8_t *bytes =
(const uint8_t *)description_->defaultValue.valueData;
if (bytes) {
uint32_t length;
memcpy(&length, bytes, sizeof(length));
// The length is stored in network byte order.
length = ntohl(length);
bytes += sizeof(length);
defaultValue_.valueData =
[[NSData alloc] initWithBytes:bytes length:length];
}
} else if (type == GPBDataTypeMessage || type == GPBDataTypeGroup) {
// The default is looked up in -defaultValue instead since extensions
// aren't common, we avoid the hit startup hit and it avoid initialization
// order issues.
} else {
defaultValue_ = description_->defaultValue;
}
}
return self;
}
- (instancetype)initWithExtensionDescription:(GPBExtensionDescription *)desc {
return [self initWithExtensionDescription:desc usesClassRefs:NO];
}
- (void)dealloc {
if ((description_->dataType == GPBDataTypeBytes) &&
!GPBExtensionIsRepeated(description_)) {
[defaultValue_.valueData release];
}
[super dealloc];
}
- (instancetype)copyWithZone:(NSZone *)zone {
#pragma unused(zone)
// Immutable.
return [self retain];
}
- (NSString *)singletonName {
return (NSString * _Nonnull)@(description_->singletonName);
}
- (const char *)singletonNameC {
return description_->singletonName;
}
- (uint32_t)fieldNumber {
return description_->fieldNumber;
}
- (GPBDataType)dataType {
return description_->dataType;
}
- (GPBWireFormat)wireType {
return GPBWireFormatForType(description_->dataType,
GPBExtensionIsPacked(description_));
}
- (GPBWireFormat)alternateWireType {
NSAssert(GPBExtensionIsRepeated(description_),
@"Only valid on repeated extensions");
return GPBWireFormatForType(description_->dataType,
!GPBExtensionIsPacked(description_));
}
- (BOOL)isRepeated {
return GPBExtensionIsRepeated(description_);
}
- (BOOL)isPackable {
return GPBExtensionIsPacked(description_);
}
- (Class)msgClass {
return description_->messageOrGroupClass.clazz;
}
- (Class)containingMessageClass {
return description_->extendedClass.clazz;
}
- (GPBEnumDescriptor *)enumDescriptor {
if (description_->dataType == GPBDataTypeEnum) {
GPBEnumDescriptor *enumDescriptor = description_->enumDescriptorFunc();
return enumDescriptor;
}
return nil;
}
- (id)defaultValue {
if (GPBExtensionIsRepeated(description_)) {
return nil;
}
switch (description_->dataType) {
case GPBDataTypeBool:
return @(defaultValue_.valueBool);
case GPBDataTypeFloat:
return @(defaultValue_.valueFloat);
case GPBDataTypeDouble:
return @(defaultValue_.valueDouble);
case GPBDataTypeInt32:
case GPBDataTypeSInt32:
case GPBDataTypeEnum:
case GPBDataTypeSFixed32:
return @(defaultValue_.valueInt32);
case GPBDataTypeInt64:
case GPBDataTypeSInt64:
case GPBDataTypeSFixed64:
return @(defaultValue_.valueInt64);
case GPBDataTypeUInt32:
case GPBDataTypeFixed32:
return @(defaultValue_.valueUInt32);
case GPBDataTypeUInt64:
case GPBDataTypeFixed64:
return @(defaultValue_.valueUInt64);
case GPBDataTypeBytes:
// Like message fields, the default is zero length data.
return (defaultValue_.valueData ? defaultValue_.valueData
: GPBEmptyNSData());
case GPBDataTypeString:
// Like message fields, the default is zero length string.
return (defaultValue_.valueString ? defaultValue_.valueString : @"");
case GPBDataTypeGroup:
case GPBDataTypeMessage:
return nil;
}
}
- (NSComparisonResult)compareByFieldNumber:(GPBExtensionDescriptor *)other {
int32_t selfNumber = description_->fieldNumber;
int32_t otherNumber = other->description_->fieldNumber;
if (selfNumber < otherNumber) {
return NSOrderedAscending;
} else if (selfNumber == otherNumber) {
return NSOrderedSame;
} else {
return NSOrderedDescending;
}
}
@end
#pragma clang diagnostic pop