susumu.yata
null+****@clear*****
Tue Dec 16 10:41:25 JST 2014
susumu.yata 2014-11-10 17:05:05 +0900 (Mon, 10 Nov 2014) New Revision: 286400083efc103f43d6a034394d55b535eca3b8 https://github.com/groonga/grnxx/commit/286400083efc103f43d6a034394d55b535eca3b8 Message: Add a draft of Column<Text>. Modified files: lib/grnxx/impl/column/base.cpp lib/grnxx/impl/column/scalar.hpp lib/grnxx/impl/column/scalar/Makefile.am lib/grnxx/impl/column/scalar/int.cpp lib/grnxx/impl/column/scalar/text.cpp lib/grnxx/impl/column/scalar/text.hpp Modified: lib/grnxx/impl/column/base.cpp (+4 -4) =================================================================== --- lib/grnxx/impl/column/base.cpp 2014-11-10 15:36:00 +0900 (b5b1634) +++ lib/grnxx/impl/column/base.cpp 2014-11-10 17:05:05 +0900 (797c86f) @@ -195,10 +195,10 @@ std::unique_ptr<ColumnBase> ColumnBase::create( column.reset(new impl::Column<GeoPoint>(table, name, options)); break; } -// case TEXT_DATA: { -// column.reset(new impl::Column<Text>(table, name, options)); -// break; -// } + case TEXT_DATA: { + column.reset(new impl::Column<Text>(table, name, options)); + break; + } // case BOOL_VECTOR_DATA: { // column.reset(new impl::Column<Vector<Bool>>(table, name, options)); // break; Modified: lib/grnxx/impl/column/scalar.hpp (+1 -1) =================================================================== --- lib/grnxx/impl/column/scalar.hpp 2014-11-10 15:36:00 +0900 (d6d9c3a) +++ lib/grnxx/impl/column/scalar.hpp 2014-11-10 17:05:05 +0900 (f443aff) @@ -5,6 +5,6 @@ #include "grnxx/impl/column/scalar/float.hpp" #include "grnxx/impl/column/scalar/geo_point.hpp" #include "grnxx/impl/column/scalar/int.hpp" -//#include "grnxx/impl/column/scalar/text.hpp" +#include "grnxx/impl/column/scalar/text.hpp" #endif // GRNXX_IMPL_COLUMN_SCALAR_HPP Modified: lib/grnxx/impl/column/scalar/Makefile.am (+2 -3) =================================================================== --- lib/grnxx/impl/column/scalar/Makefile.am 2014-11-10 15:36:00 +0900 (1e31021) +++ lib/grnxx/impl/column/scalar/Makefile.am 2014-11-10 17:05:05 +0900 (e381afc) @@ -12,9 +12,8 @@ libgrnxx_impl_column_scalar_la_SOURCES = \ bool.cpp \ float.cpp \ geo_point.cpp \ - int.cpp - -# text.cpp + int.cpp \ + text.cpp libgrnxx_impl_column_scalar_includedir = ${includedir}/grnxx/impl/column/scalar libgrnxx_impl_column_scalar_include_HEADERS = \ Modified: lib/grnxx/impl/column/scalar/int.cpp (+1 -0) =================================================================== --- lib/grnxx/impl/column/scalar/int.cpp 2014-11-10 15:36:00 +0900 (784892b) +++ lib/grnxx/impl/column/scalar/int.cpp 2014-11-10 17:05:05 +0900 (4528629) @@ -209,6 +209,7 @@ void Column<Int>::set_key(Int row_id, const Datum &key) { if (value_id >= values_.size()) { values_.resize(value_id + 1, Int::na()); } + // TODO: N/A is not available. Int value = parse_datum(key); // TODO: Update indexes if exist. // for (size_t i = 0; i < num_indexes(); ++i) try { Modified: lib/grnxx/impl/column/scalar/text.cpp (+354 -214) =================================================================== --- lib/grnxx/impl/column/scalar/text.cpp 2014-11-10 15:36:00 +0900 (2db612a) +++ lib/grnxx/impl/column/scalar/text.cpp 2014-11-10 17:05:05 +0900 (b1b6b66) @@ -1,268 +1,408 @@ -#include "grnxx/impl/column/column_text.hpp" +#include "grnxx/impl/column/scalar/text.hpp" + +#include <cstring> +//#include <set> -#include "grnxx/cursor.hpp" #include "grnxx/impl/db.hpp" #include "grnxx/impl/table.hpp" -#include "grnxx/index.hpp" - -#include <set> +//#include "grnxx/impl/index.hpp" namespace grnxx { namespace impl { -bool Column<Text>::set(Error *error, Int row_id, const Datum &datum) { - if (datum.type() != TEXT_DATA) { - GRNXX_ERROR_SET(error, INVALID_ARGUMENT, "Wrong data type"); - return false; +Column<Text>::Column(Table *table, + const String &name, + const ColumnOptions &) + : ColumnBase(table, name, TEXT_DATA), + headers_(), + bodies_() {} + +Column<Text>::~Column() {} + +void Column<Text>::set(Int row_id, const Datum &datum) { + Text new_value = parse_datum(datum); + if (!table_->test_row(row_id)) { + throw "Invalid row ID"; // TODO + } + if (is_key_) { + if (new_value.is_na()) { + throw "N/A key"; // TODO + } } - if (!table_->test_row(error, row_id)) { - return false; + if (new_value.is_na()) { + unset(row_id); + return; } Text old_value = get(row_id); - Text new_value = datum.force_text(); - if (new_value != old_value) { - if (has_key_attribute_ && contains(datum)) { - GRNXX_ERROR_SET(error, ALREADY_EXISTS, "Key duplicate"); - return false; - } - for (Int i = 0; i < num_indexes(); ++i) { - if (!indexes_[i]->insert(error, row_id, datum)) { - for (Int j = 0; j < i; ++i) { - indexes_[j]->remove(nullptr, row_id, datum); - } - return false; - } - } - Int offset = bodies_.size(); - UInt new_header; - if (new_value.size() < 0xFFFF) { - if (!bodies_.resize(error, offset + new_value.size())) { - return false; - } - std::memcpy(&bodies_[offset], new_value.data(), new_value.size()); - new_header = (offset << 16) | new_value.size(); - } else { - // The size of a long text is stored in front of the body. - if ((offset % sizeof(Int)) != 0) { - offset += sizeof(Int) - (offset % sizeof(Int)); - } - if (!bodies_.resize(error, offset + sizeof(Int) + new_value.size())) { - return false; - } - *reinterpret_cast<Int *>(&bodies_[offset]) = new_value.size(); - std::memcpy(&bodies_[offset + sizeof(Int)], - new_value.data(), new_value.size()); - new_header = (offset << 16) | 0xFFFF; - } - for (Int i = 0; i < num_indexes(); ++i) { - indexes_[i]->remove(nullptr, row_id, old_value); - } - headers_[row_id] = new_header; + if (old_value == new_value) { + return; } - return true; -} - -bool Column<Text>::get(Error *error, Int row_id, Datum *datum) const { - if (!table_->test_row(error, row_id)) { - return false; + if (is_key_ && contains(datum)) { + throw "Key already exists"; // TODO } - *datum = get(row_id); - return true; -} - -unique_ptr<Column<Text>> Column<Text>::create( - Error *error, - Table *table, - const StringCRef &name, - const ColumnOptions &options) { - unique_ptr<Column> column(new (nothrow) Column); - if (!column) { - GRNXX_ERROR_SET(error, NO_MEMORY, "Memory allocation failed"); - return nullptr; + if (!old_value.is_na()) { + // TODO: Remove the old value from indexes. +// for (size_t i = 0; i < num_indexes(); ++i) { +// indexes_[i]->remove(row_id, old_value); +// } } - if (!column->initialize_base(error, table, name, TEXT_DATA, options)) { - return nullptr; + size_t value_id = row_id.value(); + if (value_id >= headers_.size()) { + headers_.resize(value_id + 1, na_header()); } - if (!column->headers_.resize(error, table->max_row_id() + 1, 0)) { - return nullptr; + // TODO: Insert the new value into indexes. +// for (size_t i = 0; i < num_indexes(); ++i) try { +// indexes_[i]->insert(row_id, datum)) { +// } catch (...) { +// for (size_t j = 0; j < i; ++i) { +// indexes_[j]->remove(row_id, datum); +// } +// throw; +// } + // TODO: Error handling. + size_t offset = bodies_.size(); + size_t size = new_value.size().value(); + uint64_t header; + if (size < 0xFFFF) { + bodies_.resize(offset + size); + std::memcpy(&bodies_[offset], new_value.data(), size); + header = (offset << 16) | size; + } else { + // The size of a long text is stored in front of the body. + if ((offset % sizeof(uint64_t)) != 0) { + offset += sizeof(uint64_t) - (offset % sizeof(uint64_t)); + } + bodies_.resize(offset + sizeof(uint64_t) + size); + *reinterpret_cast<uint64_t *>(&bodies_[offset]) = size; + std::memcpy(&bodies_[offset + sizeof(uint64_t)], new_value.data(), size); + header = (offset << 16) | 0xFFFF; } - return column; + headers_[value_id] = header; } -Column<Text>::~Column() {} +//bool Column<Text>::set(Error *error, Int row_id, const Datum &datum) { +// if (datum.type() != TEXT_DATA) { +// GRNXX_ERROR_SET(error, INVALID_ARGUMENT, "Wrong data type"); +// return false; +// } +// if (!table_->test_row(error, row_id)) { +// return false; +// } +// Text old_value = get(row_id); +// Text new_value = datum.force_text(); +// if (new_value != old_value) { +// if (has_key_attribute_ && contains(datum)) { +// GRNXX_ERROR_SET(error, ALREADY_EXISTS, "Key duplicate"); +// return false; +// } +// for (Int i = 0; i < num_indexes(); ++i) { +// if (!indexes_[i]->insert(error, row_id, datum)) { +// for (Int j = 0; j < i; ++i) { +// indexes_[j]->remove(nullptr, row_id, datum); +// } +// return false; +// } +// } +// Int offset = bodies_.size(); +// UInt new_header; +// if (new_value.size() < 0xFFFF) { +// if (!bodies_.resize(error, offset + new_value.size())) { +// return false; +// } +// std::memcpy(&bodies_[offset], new_value.data(), new_value.size()); +// new_header = (offset << 16) | new_value.size(); +// } else { +// // The size of a long text is stored in front of the body. +// if ((offset % sizeof(Int)) != 0) { +// offset += sizeof(Int) - (offset % sizeof(Int)); +// } +// if (!bodies_.resize(error, offset + sizeof(Int) + new_value.size())) { +// return false; +// } +// *reinterpret_cast<Int *>(&bodies_[offset]) = new_value.size(); +// std::memcpy(&bodies_[offset + sizeof(Int)], +// new_value.data(), new_value.size()); +// new_header = (offset << 16) | 0xFFFF; +// } +// for (Int i = 0; i < num_indexes(); ++i) { +// indexes_[i]->remove(nullptr, row_id, old_value); +// } +// headers_[row_id] = new_header; +// } +// return true; +//} -bool Column<Text>::set_key_attribute(Error *error) { - if (has_key_attribute_) { - GRNXX_ERROR_SET(error, INVALID_OPERATION, - "This column is a key column"); - return false; +void Column<Text>::get(Int row_id, Datum *datum) const { + size_t value_id = row_id.value(); + if (value_id >= headers_.size()) { + *datum = Text::na(); + } else { + // TODO + *datum = get(row_id); } - // TODO: An index should be used if possible. - try { - std::set<Text> set; - // TODO: Functor-based inline callback may be better in this case, - // because it does not require memory allocation. - auto cursor = table_->create_cursor(nullptr); - if (!cursor) { - return false; +} + +bool Column<Text>::contains(const Datum &datum) const { + // TODO: Use an index if exists. + Text value = parse_datum(datum); + if (value.is_na()) { + for (size_t i = 0; i < headers_.size(); ++i) { + if (headers_[i] == na_header()) { + return true; + } } - Array<Record> records; - for ( ; ; ) { - auto result = cursor->read(nullptr, 1024, &records); - if (!result.is_ok) { - return false; - } else { - break; + } else { + for (size_t i = 0; i < headers_.size(); ++i) { + // TODO: Improve this. + if (get(Int(i)) == value) { + return true; + } + } + } + return false; +} + +Int Column<Text>::find_one(const Datum &datum) const { + // TODO: Use an index if exists. + Text value = parse_datum(datum); + if (value.is_na()) { + for (size_t i = 0; i < headers_.size(); ++i) { + if (headers_[i] == na_header()) { + return Int(i); } - for (Int i = 0; i < result.count; ++i) { - if (!set.insert(get(records.get_row_id(i))).second) { - GRNXX_ERROR_SET(error, INVALID_OPERATION, "Key duplicate"); - return false; - } + } + } else { + for (size_t i = 0; i < headers_.size(); ++i) { + // TODO: Improve this. + if (get(Int(i)) == value) { + return Int(i); } - records.clear(); } - } catch (...) { - GRNXX_ERROR_SET(error, NO_MEMORY, "Memory allocation failed"); - return false; } - has_key_attribute_ = true; - return true; + return Int::na(); } -bool Column<Text>::unset_key_attribute(Error *error) { - if (!has_key_attribute_) { - GRNXX_ERROR_SET(error, INVALID_OPERATION, - "This column is not a key column"); - return false; +//Int Column<Text>::find_one(const Datum &datum) const { +// // TODO: Cursor should not be used because it takes time. +// // Also, cursor operations can fail due to memory allocation. +// Text value = datum.force_text(); +// if (indexes_.size() != 0) { +// auto cursor = indexes_[0]->find(nullptr, value); +// Array<Record> records; +// auto result = cursor->read(nullptr, 1, &records); +// if (!result.is_ok || (result.count == 0)) { +// return NULL_ROW_ID; +// } +// return true; +// } else { +// // TODO: A full scan takes time. +// // An index should be required for a key column. + +// // TODO: Functor-based inline callback may be better in this case, +// // because it does not require memory allocation. + +// // Scan the column to find "value". +// auto cursor = table_->create_cursor(nullptr); +// if (!cursor) { +// return NULL_ROW_ID; +// } +// Array<Record> records; +// for ( ; ; ) { +// auto result = cursor->read(nullptr, 1024, &records); +// if (!result.is_ok || result.count == 0) { +// return NULL_ROW_ID; +// } +// for (Int i = 0; i < result.count; ++i) { +// if (get(records.get_row_id(i)) == value) { +// return records.get_row_id(i); +// } +// } +// records.clear(); +// } +// } +// return NULL_ROW_ID; +//} + +void Column<Text>::set_key_attribute() { + if (is_key_) { + throw "Key column"; // TODO } - has_key_attribute_ = false; - return true; + throw "Not supported yet"; // TODO } -bool Column<Text>::set_initial_key(Error *error, - Int row_id, - const Datum &key) { - if (!has_key_attribute_) { - GRNXX_ERROR_SET(error, INVALID_OPERATION, - "This column is not a key column"); - return false; +//bool Column<Text>::set_key_attribute(Error *error) { +// if (has_key_attribute_) { +// GRNXX_ERROR_SET(error, INVALID_OPERATION, +// "This column is a key column"); +// return false; +// } +// // TODO: An index should be used if possible. +// try { +// std::set<Text> set; +// // TODO: Functor-based inline callback may be better in this case, +// // because it does not require memory allocation. +// auto cursor = table_->create_cursor(nullptr); +// if (!cursor) { +// return false; +// } +// Array<Record> records; +// for ( ; ; ) { +// auto result = cursor->read(nullptr, 1024, &records); +// if (!result.is_ok) { +// return false; +// } else { +// break; +// } +// for (Int i = 0; i < result.count; ++i) { +// if (!set.insert(get(records.get_row_id(i))).second) { +// GRNXX_ERROR_SET(error, INVALID_OPERATION, "Key duplicate"); +// return false; +// } +// } +// records.clear(); +// } +// } catch (...) { +// GRNXX_ERROR_SET(error, NO_MEMORY, "Memory allocation failed"); +// return false; +// } +// has_key_attribute_ = true; +// return true; +//} + +void Column<Text>::unset_key_attribute() { + if (!is_key_) { + throw "Not key column"; // TODO } - if (has_key_attribute_ && contains(key)) { - GRNXX_ERROR_SET(error, ALREADY_EXISTS, "Key duplicate"); - return false; + is_key_ = true; +} + +void Column<Text>::set_key(Int row_id, const Datum &key) { + if (!is_key_) { + throw "Not key column"; // TODO } - if (row_id >= headers_.size()) { - if (!headers_.resize(error, row_id + 1, 0)) { - return false; - } + if (contains(key)) { + throw "Key already exists"; // TODO } - Text value = key.force_text(); - for (Int i = 0; i < num_indexes(); ++i) { - if (!indexes_[i]->insert(error, row_id, value)) { - for (Int j = 0; j < i; ++j) { - indexes_[j]->remove(nullptr, row_id, value); - } - return false; - } + size_t value_id = row_id.value(); + if (value_id >= headers_.size()) { + headers_.resize(value_id + 1, na_header()); } - Int offset = bodies_.size(); - UInt header; - if (value.size() < 0xFFFF) { - if (!bodies_.resize(error, offset + value.size())) { - return false; - } - std::memcpy(&bodies_[offset], value.data(), value.size()); - header = (offset << 16) | value.size(); + // TODO: N/A is not available. + Text value = parse_datum(key); + // TODO: Update indexes if exist. +// for (size_t i = 0; i < num_indexes(); ++i) try { +// indexes_[i]->insert(row_id, value); +// } catch (...) { +// for (size_t j = 0; j < i; ++j) { +// indexes_[j]->remove(row_id, value); +// } +// throw; +// } + // TODO: Error handling. + size_t offset = bodies_.size(); + size_t size = value.size().value(); + uint64_t header; + if (size < 0xFFFF) { + bodies_.resize(offset + size); + std::memcpy(&bodies_[offset], value.data(), size); + header = (offset << 16) | size; } else { // The size of a long text is stored in front of the body. - if ((offset % sizeof(Int)) != 0) { - offset += sizeof(Int) - (offset % sizeof(Int)); - } - if (!bodies_.resize(error, offset + sizeof(Int) + value.size())) { - return false; + if ((offset % sizeof(uint64_t)) != 0) { + offset += sizeof(uint64_t) - (offset % sizeof(uint64_t)); } - *reinterpret_cast<Int *>(&bodies_[offset]) = value.size(); - std::memcpy(&bodies_[offset + sizeof(Int)], value.data(), value.size()); + bodies_.resize(offset + sizeof(uint64_t) + size); + *reinterpret_cast<uint64_t *>(&bodies_[offset]) = size; + std::memcpy(&bodies_[offset + sizeof(uint64_t)], value.data(), size); header = (offset << 16) | 0xFFFF; } - headers_[row_id] = header; - return true; + headers_[value_id] = header; } -bool Column<Text>::set_default_value(Error *error, Int row_id) { - if (has_key_attribute_) { - GRNXX_ERROR_SET(error, INVALID_OPERATION, - "This column is a key column"); - return false; - } - if (row_id >= headers_.size()) { - if (!headers_.resize(error, row_id + 1)) { - return false; - } - } - Text value = TypeTraits<Text>::default_value(); - for (Int i = 0; i < num_indexes(); ++i) { - if (!indexes_[i]->insert(error, row_id, value)) { - for (Int j = 0; j < i; ++j) { - indexes_[j]->remove(nullptr, row_id, value); - } - return false; - } +//bool Column<Text>::set_initial_key(Error *error, +// Int row_id, +// const Datum &key) { +// if (!has_key_attribute_) { +// GRNXX_ERROR_SET(error, INVALID_OPERATION, +// "This column is not a key column"); +// return false; +// } +// if (has_key_attribute_ && contains(key)) { +// GRNXX_ERROR_SET(error, ALREADY_EXISTS, "Key duplicate"); +// return false; +// } +// if (row_id >= headers_.size()) { +// if (!headers_.resize(error, row_id + 1, 0)) { +// return false; +// } +// } +// Text value = key.force_text(); +// for (Int i = 0; i < num_indexes(); ++i) { +// if (!indexes_[i]->insert(error, row_id, value)) { +// for (Int j = 0; j < i; ++j) { +// indexes_[j]->remove(nullptr, row_id, value); +// } +// return false; +// } +// } +// Int offset = bodies_.size(); +// UInt header; +// if (value.size() < 0xFFFF) { +// if (!bodies_.resize(error, offset + value.size())) { +// return false; +// } +// std::memcpy(&bodies_[offset], value.data(), value.size()); +// header = (offset << 16) | value.size(); +// } else { +// // The size of a long text is stored in front of the body. +// if ((offset % sizeof(Int)) != 0) { +// offset += sizeof(Int) - (offset % sizeof(Int)); +// } +// if (!bodies_.resize(error, offset + sizeof(Int) + value.size())) { +// return false; +// } +// *reinterpret_cast<Int *>(&bodies_[offset]) = value.size(); +// std::memcpy(&bodies_[offset + sizeof(Int)], value.data(), value.size()); +// header = (offset << 16) | 0xFFFF; +// } +// headers_[row_id] = header; +// return true; +//} + +void Column<Text>::unset(Int row_id) { + Text value = get(row_id); + if (!value.is_na()) { + // TODO: Update indexes if exist. +// for (size_t i = 0; i < num_indexes(); ++i) { +// indexes_[i]->remove(row_id, value); +// } + headers_[row_id.value()] = na_header(); } - headers_[row_id] = 0; - return true; } -void Column<Text>::unset(Int row_id) { - for (Int i = 0; i < num_indexes(); ++i) { - indexes_[i]->remove(nullptr, row_id, get(row_id)); +void Column<Text>::read(ArrayCRef<Record> records, + ArrayRef<Text> values) const { + if (records.size() != values.size()) { + throw "Data size conflict"; // TODO + } + for (size_t i = 0; i < records.size(); ++i) { + values.set(i, get(records[i].row_id)); } - headers_[row_id] = 0; } -Int Column<Text>::find_one(const Datum &datum) const { - // TODO: Cursor should not be used because it takes time. - // Also, cursor operations can fail due to memory allocation. - Text value = datum.force_text(); - if (indexes_.size() != 0) { - auto cursor = indexes_[0]->find(nullptr, value); - Array<Record> records; - auto result = cursor->read(nullptr, 1, &records); - if (!result.is_ok || (result.count == 0)) { - return NULL_ROW_ID; +Text Column<Text>::parse_datum(const Datum &datum) { + switch (datum.type()) { + case NA_DATA: { + return Text::na(); } - return true; - } else { - // TODO: A full scan takes time. - // An index should be required for a key column. - - // TODO: Functor-based inline callback may be better in this case, - // because it does not require memory allocation. - - // Scan the column to find "value". - auto cursor = table_->create_cursor(nullptr); - if (!cursor) { - return NULL_ROW_ID; + case TEXT_DATA: { + return datum.as_text(); } - Array<Record> records; - for ( ; ; ) { - auto result = cursor->read(nullptr, 1024, &records); - if (!result.is_ok || result.count == 0) { - return NULL_ROW_ID; - } - for (Int i = 0; i < result.count; ++i) { - if (get(records.get_row_id(i)) == value) { - return records.get_row_id(i); - } - } - records.clear(); + default: { + throw "Wrong data type"; // TODO } } - return NULL_ROW_ID; } -Column<Text>::Column() : ColumnBase(), headers_(), bodies_() {} - } // namespace impl } // namespace grnxx Modified: lib/grnxx/impl/column/scalar/text.hpp (+53 -36) =================================================================== --- lib/grnxx/impl/column/scalar/text.hpp 2014-11-10 15:36:00 +0900 (fb60e55) +++ lib/grnxx/impl/column/scalar/text.hpp 2014-11-10 17:05:05 +0900 (52559ff) @@ -1,71 +1,88 @@ #ifndef GRNXX_IMPL_COLUMN_SCALAR_TEXT_HPP #define GRNXX_IMPL_COLUMN_SCALAR_TEXT_HPP -#include "grnxx/impl/column/column.hpp" +#include <limits> +#include <cstdint> + +#include "grnxx/impl/column/base.hpp" namespace grnxx { namespace impl { +template <typename T> class Column; + template <> -class Column<Text> : public impl::ColumnBase { +class Column<Text> : public ColumnBase { public: - // -- Public API -- + // -- Public API (grnxx/column.hpp) -- - bool set(Error *error, Int row_id, const Datum &datum); - bool get(Error *error, Int row_id, Datum *datum) const; + Column(Table *table, const String &name, const ColumnOptions &options); + ~Column(); - // -- Internal API -- + void set(Int row_id, const Datum &datum); + void get(Int row_id, Datum *datum) const; - // Create a new column. - // - // Returns a pointer to the column on success. - // On failure, returns nullptr and stores error information into "*error" if - // "error" != nullptr. - static unique_ptr<Column> create(Error *error, - Table *table, - const StringCRef &name, - const ColumnOptions &options); + bool contains(const Datum &datum) const; + Int find_one(const Datum &datum) const; - ~Column(); + // -- Internal API (grnxx/impl/column/base.hpp) -- - bool set_key_attribute(Error *error); - bool unset_key_attribute(Error *error); + void set_key_attribute(); + void unset_key_attribute(); - bool set_initial_key(Error *error, Int row_id, const Datum &key); - bool set_default_value(Error *error, Int row_id); + void set_key(Int row_id, const Datum &key); void unset(Int row_id); - Int find_one(const Datum &datum) const; - // Return a value identified by "row_id". + // -- Internal API -- + + // Return a value. // - // Assumes that "row_id" is valid. Otherwise, the result is undefined. + // If "row_id" is valid, returns the stored value. + // If "row_id" is invalid, returns N/A. + // + // TODO: Text cannot reuse allocated memory because of this interface. Text get(Int row_id) const { - Int size = static_cast<Int>(headers_[row_id] & 0xFFFF); + size_t value_id = row_id.value(); + if (value_id >= headers_.size()) { + return Text::na(); + } + if (headers_[value_id] == na_header()) { + return Text::na(); + } + size_t size = headers_[value_id] & 0xFFFF; if (size == 0) { - return Text("", 0); + return Text(nullptr, 0); } - Int offset = static_cast<Int>(headers_[row_id] >> 16); + size_t offset = headers_[value_id] >> 16; if (size < 0xFFFF) { return Text(&bodies_[offset], size); } else { // The size of a long text is stored in front of the body. - size = *reinterpret_cast<const Int *>(&bodies_[offset]); - return StringCRef(&bodies_[offset + sizeof(Int)], size); + size = *reinterpret_cast<const uint64_t *>(&bodies_[offset]); + return Text(&bodies_[offset + sizeof(uint64_t)], size); } } - +// Text get(Int row_id) const { +// size_t value_id = row_id.value(); +// if (value_id >= values_.size()) { +// return Text::na(); +// } +// return values_[value_id]; +// } // Read values. - void read(ArrayCRef<Record> records, ArrayRef<Text> values) const { - for (Int i = 0; i < records.size(); ++i) { - values.set(i, get(records.get_row_id(i))); - } - } + // + // On failure, throws an exception. + void read(ArrayCRef<Record> records, ArrayRef<Text> values) const; private: - Array<UInt> headers_; + Array<uint64_t> headers_; Array<char> bodies_; - Column(); + static constexpr uint64_t na_header() { + return std::numeric_limits<uint64_t>::max(); + } + + static Text parse_datum(const Datum &datum); }; } // namespace impl -------------- next part -------------- HTML����������������������������...Descargar
