blob: 3b9e5ca809e69152f5b8c2190308e8a60ff10f16 [file] [log] [blame]
// Copyright 2008 The open-vcdiff Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include <config.h>
#include <limits.h> // UCHAR_MAX
#include <string>
#include "addrcache.h"
#include "codetable.h"
#include "encodetable.h"
#include "instruction_map.h"
#include "logging.h"
#include "google/output_string.h"
#include "varint_bigendian.h"
#include "vcdiff_defs.h"
namespace open_vcdiff {
static const DeltaFileHeader kHeaderStandardFormat = {
0xD6, // Header1: "V" | 0x80
0xC3, // Header2: "C" | 0x80
0xC4, // Header3: "D" | 0x80
0x00, // Header4: Draft standard format
0x00 }; // Hdr_Indicator:
// No compression, no custom code table
static const DeltaFileHeader kHeaderExtendedFormat = {
0xD6, // Header1: "V" | 0x80
0xC3, // Header2: "C" | 0x80
0xC4, // Header3: "D" | 0x80
'S', // Header4: VCDIFF/SDCH, extensions used
0x00 }; // Hdr_Indicator:
// No compression, no custom code table
// VCDiffCodeTableWriter members and methods
// If interleaved is true, the encoder writes each delta file window
// by interleaving instructions and sizes with their corresponding
// addresses and data, rather than placing these elements into three
// separate sections. This facilitates providing partially
// decoded results when only a portion of a delta file window
// is received (e.g. when HTTP over TCP is used as the
// transmission protocol.) The interleaved format is
// not consistent with the VCDIFF draft standard.
//
VCDiffCodeTableWriter::VCDiffCodeTableWriter(bool interleaved)
: max_mode_(VCDiffAddressCache::DefaultLastMode()),
dictionary_size_(0),
target_length_(0),
code_table_data_(&VCDiffCodeTableData::kDefaultCodeTableData),
instruction_map_(NULL),
last_opcode_index_(-1),
add_checksum_(false),
checksum_(0) {
InitSectionPointers(interleaved);
}
VCDiffCodeTableWriter::VCDiffCodeTableWriter(
bool interleaved,
int near_cache_size,
int same_cache_size,
const VCDiffCodeTableData& code_table_data,
unsigned char max_mode)
: max_mode_(max_mode),
address_cache_(near_cache_size, same_cache_size),
dictionary_size_(0),
target_length_(0),
code_table_data_(&code_table_data),
instruction_map_(NULL),
last_opcode_index_(-1),
add_checksum_(false),
checksum_(0) {
InitSectionPointers(interleaved);
}
VCDiffCodeTableWriter::~VCDiffCodeTableWriter() {
if (code_table_data_ != &VCDiffCodeTableData::kDefaultCodeTableData) {
delete instruction_map_;
}
}
void VCDiffCodeTableWriter::InitSectionPointers(bool interleaved) {
if (interleaved) {
data_for_add_and_run_ = &instructions_and_sizes_;
addresses_for_copy_ = &instructions_and_sizes_;
} else {
data_for_add_and_run_ = &separate_data_for_add_and_run_;
addresses_for_copy_ = &separate_addresses_for_copy_;
}
}
bool VCDiffCodeTableWriter::Init(size_t dictionary_size) {
dictionary_size_ = dictionary_size;
if (!instruction_map_) {
if (code_table_data_ == &VCDiffCodeTableData::kDefaultCodeTableData) {
instruction_map_ = VCDiffInstructionMap::GetDefaultInstructionMap();
} else {
instruction_map_ = new VCDiffInstructionMap(*code_table_data_, max_mode_);
}
if (!instruction_map_) {
return false;
}
}
if (!address_cache_.Init()) {
return false;
}
target_length_ = 0;
last_opcode_index_ = -1;
return true;
}
void VCDiffCodeTableWriter::WriteHeader(
OutputStringInterface* out,
VCDiffFormatExtensionFlags format_extensions) {
if (format_extensions == VCD_STANDARD_FORMAT) {
out->append(reinterpret_cast<const char*>(&kHeaderStandardFormat),
sizeof(kHeaderStandardFormat));
} else {
out->append(reinterpret_cast<const char*>(&kHeaderExtendedFormat),
sizeof(kHeaderExtendedFormat));
}
// If custom cache table sizes or a custom code table were used
// for encoding, here is where they would be appended to *output.
// This implementation of the encoder does not use those features,
// although the decoder can understand and interpret them.
}
// The VCDiff format allows each opcode to represent either
// one or two delta instructions. This function will first
// examine the opcode generated by the last call to EncodeInstruction.
// If that opcode was a single-instruction opcode, this function checks
// whether there is a compound (double-instruction) opcode that can
// combine that single instruction with the instruction that is now
// being added, and so save a byte of space. In that case, the
// single-instruction opcode at position last_opcode_index_ will be
// overwritten with the new double-instruction opcode.
//
// In the majority of cases, no compound opcode will be possible,
// and a new single-instruction opcode will be appended to
// instructions_and_sizes_, followed by a representation of its size
// if the opcode does not implicitly give the instruction size.
//
// As an example, say instructions_and_sizes_ contains 10 bytes, the last
// of which contains the opcode 0x02 (ADD size 1). Because that was the
// most recently added opcode, last_opcode_index_ has the value 10.
// EncodeInstruction is then called with inst = VCD_COPY, size = 4, mode = 0.
// The function will replace the old opcode 0x02 with the double-instruction
// opcode 0xA3 (ADD size 1 + COPY size 4 mode 0).
//
// All of the double-instruction opcodes in the standard code table
// have implicit sizes, meaning that the size of the instruction will not
// need to be written to the instructions_and_sizes_ string separately
// from the opcode. If a custom code table were used that did not have
// this property, then instructions_and_sizes_ might contain a
// double-instruction opcode (say, COPY size 0 mode 0 + ADD size 0)
// followed by the size of the COPY, then by the size of the ADD.
// If using the SDCH interleaved format, the address of the COPY instruction
// would follow its size, so the ordering would be
// [Compound Opcode][Size of COPY][Address of COPY][Size of ADD]
//
void VCDiffCodeTableWriter::EncodeInstruction(VCDiffInstructionType inst,
size_t size,
unsigned char mode) {
if (!instruction_map_) {
VCD_DFATAL << "EncodeInstruction() called without calling Init()"
<< VCD_ENDL;
return;
}
if (last_opcode_index_ >= 0) {
const unsigned char last_opcode =
instructions_and_sizes_[last_opcode_index_];
// The encoding engine should not generate two ADD instructions in a row.
// This won't cause a failure, but it's inefficient encoding and probably
// represents a bug in the higher-level logic of the encoder.
if ((inst == VCD_ADD) &&
(code_table_data_->inst1[last_opcode] == VCD_ADD)) {
VCD_WARNING << "EncodeInstruction() called for two ADD instructions"
" in a row" << VCD_ENDL;
}
OpcodeOrNone compound_opcode = kNoOpcode;
if (size <= UCHAR_MAX) {
compound_opcode =
instruction_map_->LookupSecondOpcode(last_opcode,
inst,
static_cast<unsigned char>(size),
mode);
if (compound_opcode != kNoOpcode) {
instructions_and_sizes_[last_opcode_index_] =
static_cast<unsigned char>(compound_opcode);
last_opcode_index_ = -1;
return;
}
}
// Try finding a compound opcode with size 0.
compound_opcode = instruction_map_->LookupSecondOpcode(last_opcode,
inst,
0,
mode);
if (compound_opcode != kNoOpcode) {
instructions_and_sizes_[last_opcode_index_] =
static_cast<unsigned char>(compound_opcode);
last_opcode_index_ = -1;
AppendSizeToString(size, &instructions_and_sizes_);
return;
}
}
OpcodeOrNone opcode = kNoOpcode;
if (size <= UCHAR_MAX) {
opcode =
instruction_map_->LookupFirstOpcode(inst,
static_cast<unsigned char>(size),
mode);
if (opcode != kNoOpcode) {
instructions_and_sizes_.push_back(static_cast<char>(opcode));
last_opcode_index_ = static_cast<int>(instructions_and_sizes_.size() - 1);
return;
}
}
// There should always be an opcode with size 0.
opcode = instruction_map_->LookupFirstOpcode(inst, 0, mode);
if (opcode == kNoOpcode) {
VCD_DFATAL << "No matching opcode found for inst " << inst
<< ", mode " << mode << ", size 0" << VCD_ENDL;
return;
}
instructions_and_sizes_.push_back(static_cast<char>(opcode));
last_opcode_index_ = static_cast<int>(instructions_and_sizes_.size() - 1);
AppendSizeToString(size, &instructions_and_sizes_);
}
void VCDiffCodeTableWriter::Add(const char* data, size_t size) {
EncodeInstruction(VCD_ADD, size);
data_for_add_and_run_->append(data, size);
target_length_ += size;
}
void VCDiffCodeTableWriter::Copy(int32_t offset, size_t size) {
if (!instruction_map_) {
VCD_DFATAL << "VCDiffCodeTableWriter::Copy() called without calling Init()"
<< VCD_ENDL;
return;
}
// If a single interleaved stream of encoded values is used
// instead of separate sections for instructions, addresses, and data,
// then the string instructions_and_sizes_ may be the same as
// addresses_for_copy_. The address should therefore be encoded
// *after* the instruction and its size.
int32_t encoded_addr = 0;
const unsigned char mode = address_cache_.EncodeAddress(
offset,
static_cast<VCDAddress>(dictionary_size_ + target_length_),
&encoded_addr);
EncodeInstruction(VCD_COPY, size, mode);
if (address_cache_.WriteAddressAsVarintForMode(mode)) {
VarintBE<int32_t>::AppendToString(encoded_addr, addresses_for_copy_);
} else {
addresses_for_copy_->push_back(static_cast<unsigned char>(encoded_addr));
}
target_length_ += size;
}
void VCDiffCodeTableWriter::Run(size_t size, unsigned char byte) {
EncodeInstruction(VCD_RUN, size);
data_for_add_and_run_->push_back(byte);
target_length_ += size;
}
size_t VCDiffCodeTableWriter::CalculateLengthOfSizeAsVarint(size_t size) {
return VarintBE<int32_t>::Length(static_cast<int32_t>(size));
}
void VCDiffCodeTableWriter::AppendSizeToString(size_t size, string* out) {
VarintBE<int32_t>::AppendToString(static_cast<int32_t>(size), out);
}
void VCDiffCodeTableWriter::AppendSizeToOutputString(
size_t size,
OutputStringInterface* out) {
VarintBE<int32_t>::AppendToOutputString(static_cast<int32_t>(size), out);
}
// This calculation must match the items added between "Start of Delta Encoding"
// and "End of Delta Encoding" in Output(), below.
size_t VCDiffCodeTableWriter::CalculateLengthOfTheDeltaEncoding() const {
size_t length_of_the_delta_encoding =
CalculateLengthOfSizeAsVarint(target_length_) +
1 + // Delta_Indicator
CalculateLengthOfSizeAsVarint(separate_data_for_add_and_run_.size()) +
CalculateLengthOfSizeAsVarint(instructions_and_sizes_.size()) +
CalculateLengthOfSizeAsVarint(separate_addresses_for_copy_.size()) +
separate_data_for_add_and_run_.size() +
instructions_and_sizes_.size() +
separate_addresses_for_copy_.size();
if (add_checksum_) {
length_of_the_delta_encoding +=
VarintBE<int64_t>::Length(static_cast<int64_t>(checksum_));
}
return length_of_the_delta_encoding;
}
void VCDiffCodeTableWriter::Output(OutputStringInterface* out) {
if (instructions_and_sizes_.empty()) {
VCD_WARNING << "Empty input; no delta window produced" << VCD_ENDL;
} else {
const size_t length_of_the_delta_encoding =
CalculateLengthOfTheDeltaEncoding();
const size_t delta_window_size =
length_of_the_delta_encoding +
1 + // Win_Indicator
CalculateLengthOfSizeAsVarint(dictionary_size_) +
CalculateLengthOfSizeAsVarint(0) +
CalculateLengthOfSizeAsVarint(length_of_the_delta_encoding);
// append() will be called many times on the output string; make sure
// the output string is resized only once at most.
out->ReserveAdditionalBytes(delta_window_size);
// Add first element: Win_Indicator
if (add_checksum_) {
out->push_back(VCD_SOURCE | VCD_CHECKSUM);
} else {
out->push_back(VCD_SOURCE);
}
// Source segment size: dictionary size
AppendSizeToOutputString(dictionary_size_, out);
// Source segment position: 0 (start of dictionary)
AppendSizeToOutputString(0, out);
// [Here is where a secondary compressor would be used
// if the encoder and decoder supported that feature.]
AppendSizeToOutputString(length_of_the_delta_encoding, out);
// Start of Delta Encoding
const size_t size_before_delta_encoding = out->size();
AppendSizeToOutputString(target_length_, out);
out->push_back(0x00); // Delta_Indicator: no compression
AppendSizeToOutputString(separate_data_for_add_and_run_.size(), out);
AppendSizeToOutputString(instructions_and_sizes_.size(), out);
AppendSizeToOutputString(separate_addresses_for_copy_.size(), out);
if (add_checksum_) {
// The checksum is a 32-bit *unsigned* integer. VarintBE requires a
// signed type, so use a 64-bit signed integer to store the checksum.
VarintBE<int64_t>::AppendToOutputString(static_cast<int64_t>(checksum_),
out);
}
out->append(separate_data_for_add_and_run_.data(),
separate_data_for_add_and_run_.size());
out->append(instructions_and_sizes_.data(),
instructions_and_sizes_.size());
out->append(separate_addresses_for_copy_.data(),
separate_addresses_for_copy_.size());
// End of Delta Encoding
const size_t size_after_delta_encoding = out->size();
if (length_of_the_delta_encoding !=
(size_after_delta_encoding - size_before_delta_encoding)) {
VCD_DFATAL << "Internal error: calculated length of the delta encoding ("
<< length_of_the_delta_encoding
<< ") does not match actual length ("
<< (size_after_delta_encoding - size_before_delta_encoding)
<< VCD_ENDL;
}
separate_data_for_add_and_run_.clear();
instructions_and_sizes_.clear();
separate_addresses_for_copy_.clear();
if (target_length_ == 0) {
VCD_WARNING << "Empty target window" << VCD_ENDL;
}
}
// Reset state for next window; assume we are using same code table
// and dictionary. The caller will have to invoke Init() if a different
// dictionary is used.
//
// Notably, Init() calls address_cache_.Init(). This resets the address
// cache between delta windows, as required by RFC section 5.1.
if (!Init(dictionary_size_)) {
VCD_DFATAL << "Internal error: calling Init() to reset "
"VCDiffCodeTableWriter state failed" << VCD_ENDL;
}
}
// Verifies dictionary is compatible with writer.
bool VCDiffCodeTableWriter::VerifyDictionary(const char * /*dictionary*/,
size_t /*size*/) const {
// Arbitrary dictionaries are allowed.
return true;
}
// Verifies target chunk is compatible with writer.
bool VCDiffCodeTableWriter::VerifyChunk(const char * /*chunk*/,
size_t /*size*/) const {
// Arbitrary targets are allowed.
return true;
}
}; // namespace open_vcdiff