emkornfield commented on code in PR #14803: URL: https://github.com/apache/arrow/pull/14803#discussion_r1037829130
########## cpp/src/parquet/page_index.cc: ########## @@ -0,0 +1,254 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you 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 "parquet/page_index.h" +#include "parquet/encoding.h" +#include "parquet/metadata.h" +#include "parquet/thrift_internal.h" + +#include <map> + +namespace parquet { + +namespace { + +template <typename DType> +void Decode(std::unique_ptr<typename EncodingTraits<DType>::Decoder>& decoder, + const std::string& src, typename DType::c_type* dst) { + decoder->SetData(/*num_values=*/1, reinterpret_cast<const uint8_t*>(src.c_str()), + static_cast<int>(src.size())); + decoder->Decode(dst, /*max_values=*/1); +} + +template <> +void Decode<ByteArrayType>(std::unique_ptr<ByteArrayDecoder>&, const std::string& src, + ByteArray* dst) { + dst->len = static_cast<uint32_t>(src.size()); + dst->ptr = reinterpret_cast<const uint8_t*>(src.c_str()); +} + +template <typename DType> +class TypedColumnIndexImpl : public TypedColumnIndex<DType> { + public: + using T = typename DType::c_type; + + explicit TypedColumnIndexImpl(const ColumnDescriptor& descr, + const std::vector<bool>& null_pages, + const std::vector<std::string>& min_values, + const std::vector<std::string>& max_values, + const BoundaryOrder& boundary_order, + const bool has_null_count = false, + const std::vector<int64_t>& null_counts = {}) + : null_pages_(null_pages), + encoded_min_values_(min_values), + encoded_max_values_(max_values), + boundary_order_(boundary_order), + has_null_count_(has_null_count), + null_counts_(null_counts) { + // Decode min and max values into a compact form (i.e. w/o null page) + auto plain_decoder = MakeTypedDecoder<DType>(Encoding::PLAIN, &descr); + T value; + for (size_t i = 0; i < null_pages_.size(); ++i) { + if (!null_pages_[i]) { + // page index -> min/max slot index + page_indexes_.emplace(i, min_values_.size()); + Decode<DType>(plain_decoder, encoded_min_values_[i], &value); + min_values_.push_back(value); + Decode<DType>(plain_decoder, encoded_max_values_[i], &value); + max_values_.push_back(value); + } + } + } + + explicit TypedColumnIndexImpl(const ColumnDescriptor& descr, + const format::ColumnIndex& column_index) + : TypedColumnIndexImpl( + descr, column_index.null_pages, column_index.min_values, + column_index.max_values, + static_cast<BoundaryOrder>(static_cast<int>(column_index.boundary_order)), + column_index.__isset.null_counts, column_index.null_counts) {} + + int64_t num_pages() const override { return static_cast<int64_t>(null_pages_.size()); } + + bool is_null_page(int64_t page_id) const override { + if (page_id >= num_pages()) { + throw ParquetException("Page index is out of bound"); + } + return null_pages_[page_id]; + } + + BoundaryOrder boundary_order() const override { return boundary_order_; } + + bool has_null_counts() const override { return has_null_count_; } + + int64_t null_count(int64_t page_id) const override { + if (page_id >= num_pages()) { + throw ParquetException("Page index is out of bound"); + } + return null_counts_[page_id]; + } + + T min_value(int64_t page_id) const override { + return min_values_[GetMinMaxSlot(page_id)]; + } + + T max_value(int64_t page_id) const override { + return max_values_[GetMinMaxSlot(page_id)]; + } + + const std::string& encoded_min(int64_t page_id) const override { + if (page_id >= num_pages()) { + throw ParquetException("Page index is out of bound"); + } + return encoded_min_values_[page_id]; + } + + const std::string& encoded_max(int64_t page_id) const override { + if (page_id >= num_pages()) { + throw ParquetException("Page index is out of bound"); + } + return encoded_max_values_[page_id]; + } + + const std::vector<bool>& null_pages() const override { return null_pages_; } + + const std::vector<int64_t>& null_counts() const override { return null_counts_; } + + const std::vector<T>& min_values() const override { return min_values_; } + + const std::vector<T>& max_values() const override { return max_values_; } + + std::vector<int64_t> GetValidPageIndices() const override { + std::vector<int64_t> valid_page_indices; + std::for_each(page_indexes_.cbegin(), page_indexes_.cend(), + [&](const std::pair<size_t, size_t>& v) { + valid_page_indices.push_back(v.first); + }); + return valid_page_indices; + } + + private: + size_t GetMinMaxSlot(int64_t page_id) const { + if (page_id >= static_cast<int64_t>(null_pages_.size())) { + throw ParquetException("Page index is out of bound"); + } + if (null_pages_[page_id]) { + throw ParquetException("Cannot get min/max value of null page"); + } + auto iter = page_indexes_.find(page_id); + if (iter == page_indexes_.cend()) { + throw ParquetException("min/max value is unavailable"); + } + return iter->second; + } + + /// Values that are copied directly from the thrift message. + std::vector<bool> null_pages_; + std::vector<std::string> encoded_min_values_; + std::vector<std::string> encoded_max_values_; + BoundaryOrder boundary_order_; + bool has_null_count_; + std::vector<int64_t> null_counts_; + + /// page_id -> slot_id in the buffer of min_values_ & max_values_ + std::map<size_t, size_t> page_indexes_; Review Comment: I guess there is a trade-off but is there a reason this data-structure is used instead of a std::vector<size_t>, that stores -1 for slots that are all null and the corresponding slot value otherwise? Also any reason for map vs unordered_map? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: github-unsubscr...@arrow.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
