open-vcdiff流式编码过程分析（三）
了解了基本原理，看代码就轻松了，把EncodeInternal的完整代码贴上来：
template<bool look_for_target_matches>
void VCDiffEngine::EncodeInternal(const char* target_data,
size_t target_size,
OutputStringInterface* diff,
CodeTableWriterInterface* coder) const {
if (!hashed_dictionary_) {
VCD_DFATAL << "Internal error: VCDiffEngine::Encode() "
"called before VCDiffEngine::Init()" << VCD_ENDL;
return;
}
if (target_size == 0) {
return; // Do nothing for empty target
}
// Special case for really small input
if (target_size < static_cast<size_t>(BlockHash::kBlockSize)) {
AddUnmatchedRemainder(target_data, target_size, coder);
FinishEncoding(target_size, diff, coder);
return;
}
RollingHash<BlockHash::kBlockSize> hasher;
BlockHash* target_hash = NULL;
if (look_for_target_matches) {
// Check matches against previously encoded target data
// in this same target window, as well as against the dictionary
target_hash = BlockHash::CreateTargetHash(target_data,
target_size,
dictionary_size());
if (!target_hash) {
VCD_DFATAL << "Instantiation of target hash failed" << VCD_ENDL;
return;
}
}
const char* const target_end = target_data + target_size;
const char* const start_of_last_block = target_end - BlockHash::kBlockSize;
// Offset of next bytes in string to ADD if NOT copied (i.e., not found in
// dictionary)
const char* next_encode = target_data;
// candidate_pos points to the start of the kBlockSize-byte block that may
// begin a match with the dictionary or previously encoded target data.
const char* candidate_pos = target_data;
uint32_t hash_value = hasher.Hash(candidate_pos);
while (1) {
const size_t bytes_encoded =
EncodeCopyForBestMatch<look_for_target_matches>(
hash_value,
candidate_pos,
next_encode,
(target_end - next_encode),
target_hash,
coder);
if (bytes_encoded > 0) {
next_encode += bytes_encoded; // Advance past COPYed data
candidate_pos = next_encode;
if (candidate_pos > start_of_last_block) {
break; // Reached end of target data
}
// candidate_pos has jumped ahead by bytes_encoded bytes, so UpdateHash
// can't be used to calculate the hash value at its new position.
hash_value = hasher.Hash(candidate_pos);
if (look_for_target_matches) {
// Update the target hash for the ADDed and COPYed data
target_hash->AddAllBlocksThroughIndex(
static_cast<int>(next_encode - target_data));
}
} else {
// No match, or match is too small to be worth a COPY instruction.
// Move to the next position in the target data.
if ((candidate_pos + 1) > start_of_last_block) {
break; // Reached end of target data
}
if (look_for_target_matches) {
target_hash->AddOneIndexHash(
static_cast<int>(candidate_pos - target_data),
hash_value);
}
hash_value = hasher.UpdateHash(hash_value,
candidate_pos[0],
candidate_pos[BlockHash::kBlockSize]);
++candidate_pos;
}
}
AddUnmatchedRemainder(next_encode, target_end - next_encode, coder);
FinishEncoding(target_size, diff, coder);
delete target_hash;
}
开头的几行，前文分析过了，从33行开始，是block匹配的逻辑。首先初始化几个变量：next_encode是为被编码的数据起点；candidate_pos是当前窗口的起点，之前介绍过，第一个16byte如果匹配不到block，窗口就会移动；target_end是本次函数调用传递进来的数据的重点，如果窗口的右端到了终点，本次编码就该结束了。
下面是一个循环，EncodeCopyForBestMatch函数，即之前介绍的，基于当前窗口的16个byte，找到一个匹配的block，然后尽量匹配其上下文，直到找到一个匹配最长的block，然后进行COPY编码，COPY前面可能会有一个ADD。EncodeCopyForBestMatch的返回值，是找到的最佳match数据的长度。
if (bytes_encoded>0)，这个分支，表示找到的match数据，并且成功进行了编码。这里的逻辑，就是next_encode，candidate_pos分别后移，后移的值即是math数据的长度，然后检查是否已经处理完毕了，如果没完，计算新窗口的哈希。下面的if(look_for_target_matchs)，表示如果需要在target数据自身里寻找match数据，那么已经编码的数据，也应该建立分段哈希，这个分段哈希称为target_hash_。
else分支，表示没有找到match（或者找到的match太短，不值得一次COPY编码）。这里的逻辑，candidate_pos_后移1，重新计算hash。然后如果需要自身匹配，加入target_hash，这里有点迷惑性，其实并不是每移动一个byte就需要增加一个block，看AddOneIndexHash的代码就清楚了，注意下面注释部分以及if判断：
// This function will be called to add blocks incrementally to the target hash
// as the encoding position advances through the target data. It will be
// called for every kBlockSize-byte block in the target data, regardless
// of whether the block is aligned evenly on a block boundary. The
// BlockHash will only store hash entries for the evenly-aligned blocks.
//
void AddOneIndexHash(int index, uint32_t hash_value) {
if (index == NextIndexToAdd()) {
AddBlock(hash_value);
}
}
循环结束后，需要对最后残留的不够一block的数据，通过函数AddUnmatchedRemainder直接编码为VCDiff的ADD，然后FinishEncoding写缓冲区。
然后再看一下之循环开头的EncodeCopyForBestMatch函数，代码：
template<bool look_for_target_matches>
inline size_t VCDiffEngine::EncodeCopyForBestMatch(
uint32_t hash_value,
const char* target_candidate_start,
const char* unencoded_target_start,
size_t unencoded_target_size,
const BlockHash* target_hash,
CodeTableWriterInterface* coder) const {
// When FindBestMatch() comes up with a match for a candidate block,
// it will populate best_match with the size, source offset,
// and target offset of the match.
BlockHash::Match best_match;
// First look for a match in the dictionary.
hashed_dictionary_->FindBestMatch(hash_value,
target_candidate_start,
unencoded_target_start,
unencoded_target_size,
&best_match);
// If target matching is enabled, then see if there is a better match
// within the target data that has been encoded so far.
if (look_for_target_matches) {
target_hash->FindBestMatch(hash_value,
target_candidate_start,
unencoded_target_start,
unencoded_target_size,
&best_match);
}
if (!ShouldGenerateCopyInstructionForMatchOfSize(best_match.size())) {
return 0;
}
if (best_match.target_offset() > 0) {
// Create an ADD instruction to encode all target bytes
// from the end of the last COPY match, if any, up to
// the beginning of this COPY match.
coder->Add(unencoded_target_start, best_match.target_offset());
}
coder->Copy(best_match.source_offset(), best_match.size());
return best_match.target_offset() // ADD size
+ best_match.size(); // + COPY size
}
这个函数本身没有太多信息量，首先调用FindBestMatch，找到最佳的match数据，然后就是ADD，COPY。重头戏都交给FindBestMatch了，我们来看一下这个函数的代码：
void BlockHash::FindBestMatch(uint32_t hash_value,
const char* target_candidate_start,
const char* target_start,
size_t target_size,
Match* best_match) const {
int match_counter = 0;
for (int block_number = FirstMatchingBlockInline(hash_value,
target_candidate_start);
(block_number >= 0) && !TooManyMatches(&match_counter);
block_number = NextMatchingBlock(block_number, target_candidate_start)) {
int source_match_offset = block_number * kBlockSize;
const int source_match_end = source_match_offset + kBlockSize;
int target_match_offset =
static_cast<int>(target_candidate_start - target_start);
const int target_match_end = target_match_offset + kBlockSize;
size_t match_size = kBlockSize;
{
// Extend match start towards beginning of unencoded data
const int limit_bytes_to_left = std::min(source_match_offset,
target_match_offset);
const int matching_bytes_to_left =
MatchingBytesToLeft(source_data_ + source_match_offset,
target_start + target_match_offset,
limit_bytes_to_left);
source_match_offset -= matching_bytes_to_left;
target_match_offset -= matching_bytes_to_left;
match_size += matching_bytes_to_left;
}
{
// Extend match end towards end of unencoded data
const size_t source_bytes_to_right = source_size_ - source_match_end;
const size_t target_bytes_to_right = target_size - target_match_end;
const size_t limit_bytes_to_right = std::min(source_bytes_to_right,
target_bytes_to_right);
match_size +=
MatchingBytesToRight(source_data_ + source_match_end,
target_start + target_match_end,
static_cast<int>(limit_bytes_to_right));
}
// Update in/out parameter if the best match found was better
// than any match already stored in *best_match.
best_match->ReplaceIfBetterMatch(match_size,
source_match_offset + starting_offset_,
target_match_offset);
}
}
未完，见下篇