Method Do

catboost/libs/data/pairs_data_loaders.cpp:28–103

    }

    void NCB::TDsvFlatPairsLoader::Do(IDatasetVisitor* visitor) {
        THolder<ILineDataReader> reader = GetLineDataReader(Args.Path, NCB::TDsvFormatOptions(), /*keepLineOrder*/false);

        const auto approxmatePairsCount = reader->GetDataLineCount(/*estimate*/true);
        TVector<TPair> pairs;
        pairs.reserve(approxmatePairsCount);
        TString line;
        ui64 lineNumber;
        THPTimer progressTimer;
        ui64 progressIndex = 0;
        while (reader->ReadLine(&line, &lineNumber)) {
            if (progressTimer.Passed() > 60/*seconds*/) {
                if (progressIndex < approxmatePairsCount) {
                    CATBOOST_DEBUG_LOG << "Last minute status: " << progressIndex / CeilDiv<ui32>(approxmatePairsCount, 100) << "% pairs loaded" << Endl;
                } else {
                    CATBOOST_DEBUG_LOG << "Last minute status: " << progressIndex << " pairs loaded" << Endl;
                }
                progressTimer.Reset();
            }
            TVector<TString> tokens = StringSplitter(line).Split('\t');
            if (tokens.empty()) {
                continue;
            }
            try {
                CB_ENSURE(tokens.ysize() == 2 || tokens.ysize() == 3,
                    "Each line should have two or three columns. This line has " << tokens.size()
                );
                TPair pair;

                size_t tokenIdx = 0;
                auto parseIdFunc = [&](TStringBuf description, ui32* id) {
                    CB_ENSURE(
                        TryFromString(tokens[tokenIdx], *id),
                        "Invalid " << description << " index: cannot parse as nonnegative index ("
                        << tokens[tokenIdx] << ')'
                    );
                    *id -= Args.DatasetSubset.Range.Begin;
                    ++tokenIdx;
                };
                parseIdFunc(TStringBuf("Winner"), &pair.WinnerId);
                parseIdFunc(TStringBuf("Loser"), &pair.LoserId);

                if (tokens.ysize() == 3) {
                    CB_ENSURE(
                        TryFromString(tokens[2], pair.Weight),
                        "Invalid weight: cannot parse as float (" << tokens[2] << ')'
                    );
                } else {
                    pair.Weight = 1.0f;
                }
                if (pair.WinnerId < Args.DatasetSubset.GetSize() &&
                    pair.LoserId < Args.DatasetSubset.GetSize())
                {
                    pairs.push_back(std::move(pair));
                } else {
                    CB_ENSURE(
                        pair.WinnerId >= Args.DatasetSubset.GetSize() &&
                        pair.LoserId >= Args.DatasetSubset.GetSize(),