github-actions[bot] commented on code in PR #69032: URL: https://github.com/apache/doris/pull/69032#discussion_r4312354163
########## be/src/core/column/column_spatial.cpp: ########## @@ -0,0 +1,322 @@ +// 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 "core/column/column_spatial.h" + +#include <crc32c/crc32c.h> + +#include <algorithm> +#include <cstring> + +#include "common/exception.h" +#include "core/column/columns_common.h" +#include "exec/common/sip_hash.h" +#include "exec/sort/sort_block.h" +#include "util/hash_util.hpp" + +namespace doris { + +namespace { + +template <typename HashValue> +void for_each_spatial_value(const ColumnSpatial& column, size_t start, size_t end, + const uint8_t* null_map, HashValue&& hash_value) { + for (size_t row = start; row < end; ++row) { + if (null_map == nullptr || null_map[row] == 0) { + hash_value(row, column.get_data_at(row)); + } + } +} + +} // namespace + +void ColumnSpatial::update_hash_with_value(size_t n, SipHash& hash) const { + const auto value = get_data_at(n); + hash.update(reinterpret_cast<const char*>(&value.size), sizeof(value.size)); + hash.update(value.data, value.size); +} + +void ColumnSpatial::update_hashes_with_value(uint64_t* __restrict hashes, + const uint8_t* __restrict null_data) const { + for_each_spatial_value(*this, 0, size(), null_data, [&](size_t row, StringRef value) { + hashes[row] = HashUtil::xxHash64WithSeed(value.data, value.size, hashes[row]); + }); +} + +void ColumnSpatial::update_xxHash_with_value(size_t start, size_t end, uint64_t& hash, + const uint8_t* __restrict null_data) const { + for_each_spatial_value(*this, start, end, null_data, [&](size_t, StringRef value) { + hash = HashUtil::xxHash64WithSeed(value.data, value.size, hash); + }); +} + +void ColumnSpatial::update_crcs_with_value(uint32_t* __restrict hashes, PrimitiveType /* type */, + uint32_t rows, uint32_t /* offset */, + const uint8_t* __restrict null_data) const { + DCHECK_EQ(rows, size()); + for_each_spatial_value(*this, 0, rows, null_data, [&](size_t row, StringRef value) { + hashes[row] = + HashUtil::zlib_crc_hash(value.data, cast_set<uint32_t>(value.size), hashes[row]); + }); +} + +void ColumnSpatial::update_crc_with_value(size_t start, size_t end, uint32_t& hash, + const uint8_t* __restrict null_data) const { + for_each_spatial_value(*this, start, end, null_data, [&](size_t, StringRef value) { + hash = HashUtil::zlib_crc_hash(value.data, cast_set<uint32_t>(value.size), hash); + }); +} + +void ColumnSpatial::update_crc32c_batch(uint32_t* __restrict hashes, + const uint8_t* __restrict null_map) const { + for_each_spatial_value(*this, 0, size(), null_map, [&](size_t row, StringRef value) { + hashes[row] = crc32c_extend(hashes[row], reinterpret_cast<const uint8_t*>(value.data), + value.size); + }); +} + +void ColumnSpatial::update_crc32c_single(size_t start, size_t end, uint32_t& hash, + const uint8_t* __restrict null_map) const { + for_each_spatial_value(*this, start, end, null_map, [&](size_t, StringRef value) { + hash = crc32c_extend(hash, reinterpret_cast<const uint8_t*>(value.data), value.size); + }); +} + +void ColumnSpatial::insert_data(const char* pos, size_t length) { + if (length <= StringView::kInlineSize) { + _data.emplace_back(pos, cast_set<uint32_t>(length)); + } else { + _data.emplace_back(_arena->insert(pos, length), cast_set<uint32_t>(length)); + } +} + +int ColumnSpatial::compare_at(size_t n, size_t m, const IColumn& rhs, + int /* nan_direction_hint */) const { + const auto& spatial = assert_cast<const ColumnSpatial&>(rhs); + DCHECK_EQ(_primitive_type, spatial._primitive_type); + return _data[n].compare(spatial._data[m]); +} + +MutableColumnPtr ColumnSpatial::clone_resized(size_t size) const { + auto result = create(_primitive_type); + const size_t copied = std::min(this->size(), size); + for (size_t i = 0; i < copied; ++i) { + const auto value = get_data_at(i); + result->insert_data(value.data, value.size); + } + result->insert_many_defaults(size - copied); + return result; +} + +void ColumnSpatial::insert_range_from(const IColumn& src, size_t start, size_t length) { + const auto& spatial = assert_cast<const ColumnSpatial&>(src); + DCHECK_EQ(_primitive_type, spatial._primitive_type); + if (start + length > spatial.size()) { + throw Exception(ErrorCode::INTERNAL_ERROR, + "Spatial column range start = {}, length = {} is out of bounds for {} rows", + start, length, spatial.size()); + } + for (size_t i = 0; i < length; ++i) { + const auto value = spatial.get_data_at(start + i); + insert_data(value.data, value.size); + } +} + +void ColumnSpatial::insert_indices_from(const IColumn& src, const uint32_t* begin, + const uint32_t* end) { + const auto& spatial = assert_cast<const ColumnSpatial&>(src); + DCHECK_EQ(_primitive_type, spatial._primitive_type); + for (const auto* it = begin; it != end; ++it) { + const auto value = spatial.get_data_at(*it); + insert_data(value.data, value.size); + } +} + +bool ColumnSpatial::has_enough_capacity(const IColumn& src) const { + const auto& spatial = assert_cast<const ColumnSpatial&>(src); + return _data.capacity() - _data.size() > spatial.size(); +} + +ColumnPtr ColumnSpatial::filter(const IColumn::Filter& filter, ssize_t result_size_hint) const { + column_match_filter_size(size(), filter.size()); + auto result = create(_primitive_type); + if (result_size_hint > 0) { + result->_data.reserve(result_size_hint); + } + for (size_t i = 0; i < size(); ++i) { + if (filter[i]) { + const auto value = get_data_at(i); + result->insert_data(value.data, value.size); + } + } + return result; +} + +size_t ColumnSpatial::filter(const IColumn::Filter& filter) { + column_match_filter_size(size(), filter.size()); + // Moving views within this column keeps their arena alive and avoids copying retained WKB. + size_t retained = 0; + for (size_t i = 0; i < size(); ++i) { + if (filter[i]) { + _data[retained++] = _data[i]; + } + } + _data.resize(retained); + return retained; +} + +void ColumnSpatial::erase(size_t start, size_t length) { + // Move retained views first; reclaim their arena periodically below to bound window memory. + // Match ColumnVector: empty/out-of-range requests are no-ops and long ranges are clipped. + if (start >= size() || length == 0) { + return; + } + length = std::min(length, size() - start); + std::move(_data.begin() + start + length, _data.end(), _data.begin() + start); + _data.resize(size() - length); + size_t retained_bytes = 0; + for (const auto& value : _data) { + retained_bytes += value.isInline() ? 0 : value.size(); + } + // A bounded analytic frame can discard most WKB while continuing to append new blocks. + // Once more than half the payload is dead, copy survivors once and release the old arena. + // Keeping the old storage until both swaps also preserves ownership if copying throws. + if (retained_bytes < _arena->used_size() / 2) { + auto compacted = create(_primitive_type); + compacted->insert_range_from(*this, 0, size()); + _data.swap(compacted->_data); + _arena.swap(compacted->_arena); + } +} + +MutableColumnPtr ColumnSpatial::permute(const IColumn::Permutation& perm, size_t limit) const { + limit = limit ? std::min(size(), limit) : size(); + if (perm.size() < limit) { + throw Exception(ErrorCode::INTERNAL_ERROR, "Size of permutation is less than required"); + } + auto result = create(_primitive_type); + for (size_t i = 0; i < limit; ++i) { + const auto value = get_data_at(perm[i]); + result->insert_data(value.data, value.size); + } + return result; +} + +void ColumnSpatial::replace_column_data(const IColumn& rhs, size_t row, size_t self_row) { + DCHECK_LT(self_row, size()); + const auto& spatial = assert_cast<const ColumnSpatial&>(rhs); + DCHECK_EQ(_primitive_type, spatial._primitive_type); + const auto value = spatial.get_data_at(row); + if (value.size <= StringView::kInlineSize) { + _data[self_row] = StringView(value.data, cast_set<uint32_t>(value.size)); + } else { + _data[self_row] = + StringView(_arena->insert(value.data, value.size), cast_set<uint32_t>(value.size)); + } +} + +size_t ColumnSpatial::get_max_row_byte_size() const { + size_t maximum = 0; + for (const auto& value : _data) { + maximum = std::max(maximum, static_cast<size_t>(value.size())); + } + return maximum + sizeof(uint32_t); +} + +size_t ColumnSpatial::deserialize_impl(const char* pos) { + const auto value_size = unaligned_load<uint32_t>(pos); + pos += sizeof(value_size); + insert_data(pos, value_size); + return value_size + sizeof(value_size); +} + +size_t ColumnSpatial::serialize_impl(char* pos, size_t row) const { + const auto value = _data[row]; + const auto value_size = value.size(); + memcpy_fixed<uint32_t>(pos, reinterpret_cast<const char*>(&value_size)); + memcpy(pos + sizeof(uint32_t), value.data(), value_size); + return value_size + sizeof(uint32_t); +} + +size_t ColumnSpatial::serialize_size_at(size_t row) const { + return _data[row].size() + sizeof(uint32_t); +} + +StringRef ColumnSpatial::serialize_value_into_arena(size_t n, Arena& arena, + char const*& begin) const { + char* position = arena.alloc_continue(serialize_size_at(n), begin); + return {position, serialize_impl(position, n)}; +} + +const char* ColumnSpatial::deserialize_and_insert_from_arena(const char* pos) { + return pos + deserialize_impl(pos); +} + +void ColumnSpatial::serialize(StringRef* keys, size_t num_rows) const { + for (size_t i = 0; i < num_rows; ++i) { + keys[i].size += serialize_impl(const_cast<char*>(keys[i].data + keys[i].size), i); + } +} + +void ColumnSpatial::deserialize(StringRef* keys, size_t num_rows) { + for (size_t i = 0; i < num_rows; ++i) { + const auto size = deserialize_impl(keys[i].data); + keys[i].data += size; + keys[i].size -= size; + } +} + +template <bool positive> +struct ColumnSpatial::less { + const ColumnSpatial& parent; + bool operator()(size_t lhs, size_t rhs) const { + const int comparison = parent._data[lhs].compare(parent._data[rhs]); + return positive ? comparison < 0 : comparison > 0; + } +}; + +void ColumnSpatial::get_permutation(bool reverse, size_t /* limit */, int /* nan_direction_hint */, + HybridSorter& sorter, IColumn::Permutation& result) const { + result.resize(size()); + for (size_t i = 0; i < size(); ++i) { + result[i] = i; + } + if (reverse) { + sorter.sort(result.begin(), result.end(), less<false> {*this}); + } else { + sorter.sort(result.begin(), result.end(), less<true> {*this}); + } +} + +void ColumnSpatial::insert_many_strings(const StringRef* strings, size_t num) { + for (size_t i = 0; i < num; ++i) { + insert_data(strings[i].data, strings[i].size); + } +} + +void ColumnSpatial::insert_many_strings_overflow(const StringRef* strings, size_t num, + size_t /* max_length */) { + insert_many_strings(strings, num); +} + +void ColumnSpatial::sort_column(const ColumnSorter* sorter, EqualFlags& flags, + IColumn::Permutation& perms, EqualRange& range, + bool last_column) const { + sorter->sort_column(*this, flags, perms, range, last_column); Review Comment: [P2] Add nullable spatial dispatch to multi-column sorting For an OPTIONAL Iceberg spatial column, `SELECT id FROM t ORDER BY geom, id` reaches `ColumnSorter::sort_column(ColumnNullable)`, which forwards the nested `ColumnSpatial` while the sorter still stores the original Nullable column. There is no `ColumnSpatial` sorter overload, so this call selects the generic `IColumn` overload. Its equality comparator passes the Nullable RHS into `ColumnSpatial::compare_at`, which requires a spatial column and throws a bad cast when at least two non-null rows are compared. Geography is affected too, even when the optional column contains no NULLs. Add a sorting path that compares the nested columns consistently and cover multi-key sorting; single-key `get_permutation` does not exercise this dispatch. ########## fe/fe-core/src/main/java/org/apache/doris/nereids/rules/analysis/BindExpression.java: ########## @@ -304,6 +306,7 @@ private LogicalExternalRowLevelMergeSink<Plan> bindIcebergMergeSink( validateMergeNoLossyCoercion(sink.getCols(), sink.child(), ctx.cascadesContext.getCteContext()); } List<NamedExpression> castExprs = coerceMergeOutput(sink.getCols(), outputExprs); + validateMergeSpatialTypes(sink.getCols(), sink.child(), outputExprs); Review Comment: [P2] Skip spatial write validation for delete-only MERGE A readable, imported Iceberg v3 MOR table with `shapes ARRAY<GEOMETRY>` now rejects `MERGE INTO t USING s ON t.id=s.id WHEN MATCHED THEN DELETE` with the nested-spatial-write error. The relevant plan is: ```text MergeSink[writesDataFiles=false] Project(DELETE, row_id, id, shapes:ARRAY<GEOMETRY>) Join(t, s) ``` `buildDeleteProjection` forwards the target columns, so this unconditional call validates `shapes`. However, `VIcebergMergeSink` omits the data writer and sends only `row_id` to the position-delete writer; it never writes nested spatial values. Gate this validation with `sink.isWritesDataFiles()`, as the adjacent Variant check does, and cover delete-only MERGE on an imported nested-spatial table. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
