Skip to content

Fix UTF-8 partial-match ranking #101

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Uzer007-admin wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Fix UTF-8 partial-match ranking #101

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b2a611f
Fix UTF-8 partial match ranking
Uzer007-admin Jul 18, 2026
c33995f
Refine UTF-8 partial-match ranking implementation
Uzer007-admin Jul 23, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions cpp_src/core/ft/ft_fast/dataholder.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,13 +94,15 @@ size_t DataHolder<IdCont>::GetMemStat() {
for (auto& w : words_) {
res += sizeof(w) + w.vids.heap_size();
}
res += wordsCharsLen_.capacity() * sizeof(wordsCharsLen_[0]);

@MadSchemas MadSchemas Jul 21, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

wordsCharsLen_[0] is unsafe from ASAN/Debug standpoint. Size of value_type would be better

return res;
}

template <typename IdCont>
void DataHolder<IdCont>::Clear() {
IDataHolder::Clear();
words_.resize(0);
wordsCharsLen_.resize(0);
}

template <typename IdCont>
Expand All @@ -110,10 +112,12 @@ void DataHolder<IdCont>::StartCommit(bool complete_updated) {

Clear();
words_.clear();
wordsCharsLen_.clear();
lastStepWords_.clear();
} else if (NeedRecommitLast()) {
status_ = RecommitLast;
words_.erase(words_.begin() + steps.back().wordOffset_, words_.end());
wordsCharsLen_.erase(wordsCharsLen_.begin() + steps.back().wordOffset_, wordsCharsLen_.end());

for (auto& word : words_) {
word.RestoreState();
Expand Down
16 changes: 16 additions & 0 deletions cpp_src/core/ft/ft_fast/dataholder.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -185,6 +185,8 @@ class [[nodiscard]] IDataHolder {
template <typename IdCont>
class [[nodiscard]] DataHolder : public IDataHolder {
public:
using WordCharsLenType = uint16_t;

explicit DataHolder(FTConfig* c);
void Process(VDocsTexts& vdocsTexts, const std::vector<uint32_t>& vdocsIds, size_t numDocsTotal, size_t fieldSize, bool multithread,
std::vector<h_vector<float, 3>>& wordsCounts) final;
Expand All @@ -193,6 +195,19 @@ class [[nodiscard]] DataHolder : public IDataHolder {
void Clear() override final;
std::vector<PackedWordEntry<IdCont>>& GetWords() noexcept { return words_; }
const std::vector<PackedWordEntry<IdCont>>& GetWords() const noexcept { return words_; }
void ReserveWords(size_t capacity) {
words_.reserve(capacity);
wordsCharsLen_.reserve(capacity);
}
void AddWordCharsLen(WordCharsLenType charsLen) {
assertrx_dbg(wordsCharsLen_.size() < words_.size());
wordsCharsLen_.emplace_back(charsLen);
}
WordCharsLenType GetWordCharsLen(WordIdType id) const noexcept {
assertrx(!id.IsEmpty());
assertrx(id.b.id < wordsCharsLen_.size());
return wordsCharsLen_[id.b.id];
}
PackedWordEntry<IdCont>& GetWordEntry(WordIdType id) noexcept {
assertrx(!id.IsEmpty());
assertrx(id.b.id < words_.size());
Expand All @@ -204,6 +219,7 @@ class [[nodiscard]] DataHolder : public IDataHolder {
return words_[id.b.id];
}
std::vector<PackedWordEntry<IdCont>> words_;
std::vector<WordCharsLenType> wordsCharsLen_;
};

} // namespace reindexer
6 changes: 3 additions & 3 deletions cpp_src/core/ft/ft_fast/dataprocessor.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,9 +80,7 @@ typename DataProcessor<IdCont>::WordsVector DataProcessor<IdCont>::insertIntoSuf
word_hash wh;

suffix.reserve(words_um.size() * 20, words_um.size());
if (words.empty()) {
words.reserve(words_um.size());
}
holder.ReserveWords(words.size() + words_um.size());

@MadSchemas MadSchemas Jul 21, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

words.size() + words_um.size() does not looks correct for the cases, when words already contain something. Actually, current logic may overreserve here up to 2 times


WordsVector found;
found.reserve(words_um.size());
Expand All @@ -99,7 +97,9 @@ typename DataProcessor<IdCont>::WordsVector DataProcessor<IdCont>::insertIntoSuf
}
found.emplace_back(keyIt.first);

const auto charsLen = static_cast<typename DataHolder<IdCont>::WordCharsLenType>(getUTF8StringCharactersCount(keyIt.first));
words.emplace_back();
holder.AddWordCharsLen(charsLen);
pos = holder.BuildWordId(id);
suffix.insert(keyIt.first, pos);
holder.lastStepWords_[wh(keyIt.first)].emplace_back(pos);
Expand Down
10 changes: 6 additions & 4 deletions cpp_src/core/ft/ft_fast/selecterimpl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -293,6 +293,7 @@ ft::TermResults<IdCont> Selector<IdCont>::buildTermResults(const FtDSLEntry& ter
size_t matched = 0, vids = 0, excludedCnt = 0;
const std::string& patternUtf8 = variant.PatternUtf8();
const size_t patternBytes = patternUtf8.length();
const auto patternChars = static_cast<uint32_t>(getUTF8StringCharactersCount(patternUtf8));
for (const auto& step : holder_.steps) {
if (variant.lowRelevance && totalVids >= lowRelevanceLimit) {
break;
Expand Down Expand Up @@ -329,11 +330,12 @@ ft::TermResults<IdCont> Selector<IdCont>::buildTermResults(const FtDSLEntry& ter
break;
}

// ToDo fix it (broken for russian utf8 symbols)
const int matchDif = std::abs(long(word.length() - variant.PatternUtf8().length() + wordLengthBeforePattern));
const uint32_t wordChars = holder_.GetWordCharsLen(wordId);
const uint32_t unmatchedChars = wordChars > patternChars ? wordChars - patternChars : 0;
const float boost = std::max(getTermBoost(std::string(word)), variant.boost);
const float decreasePenalty = static_cast<float>(holder_.cfg_->partialMatchDecrease * matchDif) /
std::max<float>(variant.PatternUtf8().length(), kMinPartialMatchDenominator);
const float decreasePenalty =
(static_cast<float>(holder_.cfg_->partialMatchDecrease) * static_cast<float>(unmatchedChars)) /
static_cast<float>(std::max(patternChars, kMinPartialMatchDenominator));
float proc = std::max<float>(variant.proc - decreasePenalty, isPrefix ? rankingCfg.PrefixMin() : rankingCfg.SuffixMin());
proc = std::min<float>(proc, variant.proc);
if (boost > 0.0f) {
Expand Down