Merge

dbms/include/DB/Columns/ColumnAggregateFunction.h

@@ -113,21 +113,21 @@ public:
 		return res;
 	}
 
- 	std::string getName() const { return "ColumnAggregateFunction"; }
+	std::string getName() const override { return "ColumnAggregateFunction"; }
 
-	size_t sizeOfField() const { return sizeof(getData()[0]); }
+	size_t sizeOfField() const override { return sizeof(getData()[0]); }
 
-	size_t size() const
+	size_t size() const override
 	{
 		return getData().size();
 	}
 
- 	ColumnPtr cloneEmpty() const
- 	{
+	ColumnPtr cloneEmpty() const override
+	{
 		return new ColumnAggregateFunction(holder->func, Arenas(1, new Arena));
 	};
 
-	Field operator[](size_t n) const
+	Field operator[](size_t n) const override
 	{
 		Field field = String();
 		{
@@ -137,7 +137,7 @@ public:
 		return field;
 	}
 
-	void get(size_t n, Field & res) const
+	void get(size_t n, Field & res) const override
 	{
 		res = String();
 		{
@@ -146,17 +146,17 @@ public:
 		}
 	}
 
-	StringRef getDataAt(size_t n) const
+	StringRef getDataAt(size_t n) const override
 	{
 		return StringRef(reinterpret_cast<const char *>(&getData()[n]), sizeof(getData()[n]));
 	}
 
-	void insertData(const char * pos, size_t length)
+	void insertData(const char * pos, size_t length) override
 	{
 		getData().push_back(*reinterpret_cast<const AggregateDataPtr *>(pos));
 	}
 
-	void insertFrom(const IColumn & src, size_t n)
+	void insertFrom(const IColumn & src, size_t n) override
 	{
 		getData().push_back(static_cast<const ColumnAggregateFunction &>(src).getData()[n]);
 	}
@@ -167,7 +167,7 @@ public:
 		holder.get()->func.get()->merge(getData().back(), static_cast<const ColumnAggregateFunction &>(src).getData()[n]);
 	}
 
-	void insert(const Field & x)
+	void insert(const Field & x) override
 	{
 		IAggregateFunction * function = holder.get()->func;
 
@@ -177,17 +177,17 @@ public:
 		function->deserializeMerge(getData().back(), read_buffer);
 	}
 
-	void insertDefault()
+	void insertDefault() override
 	{
 		throw Exception("Method insertDefault is not supported for ColumnAggregateFunction.", ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	size_t byteSize() const
+	size_t byteSize() const override
 	{
 		return getData().size() * sizeof(getData()[0]);
 	}
 
-	ColumnPtr cut(size_t start, size_t length) const
+	ColumnPtr cut(size_t start, size_t length) const override
 	{
 		if (start + length > getData().size())
 			throw Exception("Parameters start = "
@@ -205,7 +205,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr filter(const Filter & filter) const
+	ColumnPtr filter(const Filter & filter) const override
 	{
 		size_t size = getData().size();
 		if (size != filter.size())
@@ -225,7 +225,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr permute(const Permutation & perm, size_t limit) const
+	ColumnPtr permute(const Permutation & perm, size_t limit) const override
 	{
 		size_t size = getData().size();
 
@@ -247,22 +247,22 @@ public:
 		return res;
 	}
 
-	ColumnPtr replicate(const Offsets_t & offsets) const
+	ColumnPtr replicate(const Offsets_t & offsets) const override
 	{
 		throw Exception("Method replicate is not supported for ColumnAggregateFunction.", ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	void getExtremes(Field & min, Field & max) const
+	void getExtremes(Field & min, Field & max) const override
 	{
 		throw Exception("Method getExtremes is not supported for ColumnAggregateFunction.", ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const
+	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const override
 	{
 		return 0;
 	}
 
-	void getPermutation(bool reverse, size_t limit, Permutation & res) const
+	void getPermutation(bool reverse, size_t limit, Permutation & res) const override
 	{
 		size_t s = getData().size();
 		res.resize(s);

dbms/include/DB/Columns/ColumnArray.h

@@ -42,19 +42,19 @@ public:
 		}
 	}
 
-	std::string getName() const { return "ColumnArray(" + data->getName() + ")"; }
+	std::string getName() const override { return "ColumnArray(" + data->getName() + ")"; }
 
-	ColumnPtr cloneEmpty() const
+	ColumnPtr cloneEmpty() const override
 	{
 		return new ColumnArray(data->cloneEmpty());
 	}
 
-	size_t size() const
+	size_t size() const override
 	{
 		return getOffsets().size();
 	}
 
-	Field operator[](size_t n) const
+	Field operator[](size_t n) const override
 	{
 		size_t offset = offsetAt(n);
 		size_t size = sizeAt(n);
@@ -66,7 +66,7 @@ public:
 		return res;
 	}
 
-	void get(size_t n, Field & res) const
+	void get(size_t n, Field & res) const override
 	{
 		size_t offset = offsetAt(n);
 		size_t size = sizeAt(n);
@@ -77,7 +77,7 @@ public:
 			data->get(offset + i, res_arr[i]);
 	}
 
-	StringRef getDataAt(size_t n) const
+	StringRef getDataAt(size_t n) const override
 	{
 		/** Работает для массивов значений фиксированной длины.
 		  * Для массивов строк и массивов массивов полученный кусок памяти может не взаимно-однозначно соответствовать элементам.
@@ -87,7 +87,7 @@ public:
 		return StringRef(begin.data, end.data - begin.data);
 	}
 
-	void insertData(const char * pos, size_t length)
+	void insertData(const char * pos, size_t length) override
 	{
 		/** Аналогично - только для массивов значений фиксированной длины.
 		  */
@@ -98,14 +98,17 @@ public:
 		size_t field_size = data_->sizeOfField();
 
 		const char * end = pos + length;
-		for (; pos + field_size <= end; pos += field_size)
+		size_t elems = 0;
+		for (; pos + field_size <= end; pos += field_size, ++elems)
 			data_->insertData(pos, field_size);
 
 		if (pos != end)
 			throw Exception("Incorrect length argument for method ColumnArray::insertData", ErrorCodes::BAD_ARGUMENTS);
+
+		getOffsets().push_back((getOffsets().size() == 0 ? 0 : getOffsets().back()) + elems);
 	}
 
-	ColumnPtr cut(size_t start, size_t length) const
+	ColumnPtr cut(size_t start, size_t length) const override
 	{
 		if (length == 0)
 			return new ColumnArray(data);
@@ -138,7 +141,7 @@ public:
 		return res;
 	}
 
-	void insert(const Field & x)
+	void insert(const Field & x) override
 	{
 		const Array & array = DB::get<const Array &>(x);
 		size_t size = array.size();
@@ -147,7 +150,7 @@ public:
 		getOffsets().push_back((getOffsets().size() == 0 ? 0 : getOffsets().back()) + size);
 	}
 
-	void insertFrom(const IColumn & src_, size_t n)
+	void insertFrom(const IColumn & src_, size_t n) override
 	{
 		const ColumnArray & src = static_cast<const ColumnArray &>(src_);
 		size_t size = src.sizeAt(n);
@@ -159,12 +162,12 @@ public:
 		getOffsets().push_back((getOffsets().size() == 0 ? 0 : getOffsets().back()) + size);
 	}
 
-	void insertDefault()
+	void insertDefault() override
 	{
 		getOffsets().push_back(getOffsets().size() == 0 ? 0 : getOffsets().back());
 	}
 
-	ColumnPtr filter(const Filter & filt) const
+	ColumnPtr filter(const Filter & filt) const override
 	{
 		size_t size = getOffsets().size();
 		if (size != filt.size())
@@ -203,7 +206,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr permute(const Permutation & perm, size_t limit) const
+	ColumnPtr permute(const Permutation & perm, size_t limit) const override
 	{
 		size_t size = getOffsets().size();
 
@@ -241,7 +244,7 @@ public:
 		return res;
 	}
 
-	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const final
+	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const override
 	{
 		const ColumnArray & rhs = static_cast<const ColumnArray &>(rhs_);
 
@@ -276,7 +279,7 @@ public:
 		}
 	};
 
-	void getPermutation(bool reverse, size_t limit, Permutation & res) const
+	void getPermutation(bool reverse, size_t limit, Permutation & res) const override
 	{
 		size_t s = size();
 		if (limit >= s)
@@ -302,18 +305,18 @@ public:
 		}
 	}
 
-	void reserve(size_t n)
+	void reserve(size_t n) override
 	{
 		getOffsets().reserve(n);
 		getData().reserve(n);		/// Средний размер массивов тут никак не учитывается. Или считается, что он не больше единицы.
 	}
 
-	size_t byteSize() const
+	size_t byteSize() const override
 	{
 		return data->byteSize() + getOffsets().size() * sizeof(getOffsets()[0]);
 	}
 
-	void getExtremes(Field & min, Field & max) const
+	void getExtremes(Field & min, Field & max) const override
 	{
 		min = Array();
 		max = Array();
@@ -350,7 +353,7 @@ public:
 	const ColumnPtr & getOffsetsColumn() const { return offsets; }
 
 
-	ColumnPtr replicate(const Offsets_t & replicate_offsets) const
+	ColumnPtr replicate(const Offsets_t & replicate_offsets) const override
 	{
 		/// Не получается реализовать в общем случае.
 

dbms/include/DB/Columns/ColumnConst.h

@@ -19,7 +19,7 @@ using Poco::SharedPtr;
 class IColumnConst : public IColumn
 {
 public:
-	bool isConst() const { return true; }
+	bool isConst() const override { return true; }
 	virtual ColumnPtr convertToFullColumn() const = 0;
 };
 
@@ -37,21 +37,21 @@ public:
 	/// Для ColumnConst<String> data_type_ должен быть ненулевым, если тип данных FixedString.
 	ColumnConst(size_t s_, const T & data_, DataTypePtr data_type_ = DataTypePtr()) : s(s_), data(data_), data_type(data_type_) {}
 
-	std::string getName() const { return "ColumnConst<" + TypeName<T>::get() + ">"; }
-	bool isNumeric() const { return IsNumber<T>::value; }
-	bool isFixed() const { return IsNumber<T>::value; }
-	size_t sizeOfField() const { return sizeof(T); }
-	ColumnPtr cloneResized(size_t s_) const { return new ColumnConst(s_, data); }
-	size_t size() const { return s; }
-	Field operator[](size_t n) const { return FieldType(data); }
-	void get(size_t n, Field & res) const { res = FieldType(data); }
+	std::string getName() const override { return "ColumnConst<" + TypeName<T>::get() + ">"; }
+	bool isNumeric() const override { return IsNumber<T>::value; }
+	bool isFixed() const override { return IsNumber<T>::value; }
+	size_t sizeOfField() const override { return sizeof(T); }
+	ColumnPtr cloneResized(size_t s_) const override { return new ColumnConst(s_, data); }
+	size_t size() const override { return s; }
+	Field operator[](size_t n) const override { return FieldType(data); }
+	void get(size_t n, Field & res) const override { res = FieldType(data); }
 
-	ColumnPtr cut(size_t start, size_t length) const
+	ColumnPtr cut(size_t start, size_t length) const override
 	{
 		return new ColumnConst<T>(length, data, data_type);
 	}
 
-	void insert(const Field & x)
+	void insert(const Field & x) override
 	{
 		if (x.get<FieldType>() != FieldType(data))
 			throw Exception("Cannot insert different element into constant column " + getName(),
@@ -59,12 +59,12 @@ public:
 		++s;
 	}
 
-	void insertData(const char * pos, size_t length)
+	void insertData(const char * pos, size_t length) override
 	{
 		throw Exception("Cannot insert element into constant column " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	void insertFrom(const IColumn & src, size_t n)
+	void insertFrom(const IColumn & src, size_t n) override
 	{
 		if (data != static_cast<const ColumnConst<T> &>(src).data)
 			throw Exception("Cannot insert different element into constant column " + getName(),
@@ -72,9 +72,9 @@ public:
 		++s;
 	}
 
-	void insertDefault() { ++s; }
+	void insertDefault() override { ++s; }
 
-	ColumnPtr filter(const Filter & filt) const
+	ColumnPtr filter(const Filter & filt) const override
 	{
 		if (s != filt.size())
 			throw Exception("Size of filter doesn't match size of column.", ErrorCodes::SIZES_OF_COLUMNS_DOESNT_MATCH);
@@ -82,7 +82,7 @@ public:
 		return new ColumnConst<T>(countBytesInFilter(filt), data, data_type);
 	}
 
-	ColumnPtr replicate(const Offsets_t & offsets) const
+	ColumnPtr replicate(const Offsets_t & offsets) const override
 	{
 		if (s != offsets.size())
 			throw Exception("Size of offsets doesn't match size of column.", ErrorCodes::SIZES_OF_COLUMNS_DOESNT_MATCH);
@@ -91,9 +91,9 @@ public:
 		return new ColumnConst<T>(replicated_size, data, data_type);
 	}
 
-	size_t byteSize() const { return sizeof(data) + sizeof(s); }
+	size_t byteSize() const override { return sizeof(data) + sizeof(s); }
 
-	ColumnPtr permute(const Permutation & perm, size_t limit) const
+	ColumnPtr permute(const Permutation & perm, size_t limit) const override
 	{
 		if (limit == 0)
 			limit = s;
@@ -106,7 +106,7 @@ public:
 		return new ColumnConst<T>(limit, data, data_type);
 	}
 
-	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const
+	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const override
 	{
 		const ColumnConst<T> & rhs = static_cast<const ColumnConst<T> &>(rhs_);
 		return data < rhs.data	/// TODO: правильное сравнение NaN-ов в константных столбцах.
@@ -116,30 +116,33 @@ public:
 				: 1);
 	}
 
-	void getPermutation(bool reverse, size_t limit, Permutation & res) const
+	void getPermutation(bool reverse, size_t limit, Permutation & res) const override
 	{
 		res.resize(s);
 		for (size_t i = 0; i < s; ++i)
 			res[i] = i;
 	}
 
-	StringRef getDataAt(size_t n) const;
-	StringRef getDataAtWithTerminatingZero(size_t n) const;
-	UInt64 get64(size_t n) const;
+	StringRef getDataAt(size_t n) const override;
+	StringRef getDataAtWithTerminatingZero(size_t n) const override;
+	UInt64 get64(size_t n) const override;
 
 	/** Более эффективные методы манипуляции */
 	T & getData() { return data; }
 	const T & getData() const { return data; }
 
 	/** Преобразование из константы в полноценный столбец */
-	ColumnPtr convertToFullColumn() const;
+	ColumnPtr convertToFullColumn() const override;
 
-	void getExtremes(Field & min, Field & max) const
+	void getExtremes(Field & min, Field & max) const override
 	{
 		min = FieldType(data);
 		max = FieldType(data);
 	}
 
+	DataTypePtr & getDataType() { return data_type; }
+	const DataTypePtr & getDataType() const { return data_type; }
+
 private:
 	size_t s;
 	T data;

dbms/include/DB/Columns/ColumnExpression.h

@@ -6,7 +6,7 @@
 
 namespace DB
 {
-	
+
 /** Столбец, содержащий лямбда-выражение.
   * Ведёт себя как столбец-константа. Содержит выражение, но не входные или выходные данные.
   */
@@ -15,19 +15,19 @@ class ColumnExpression final : public IColumnDummy
 public:
 	ColumnExpression(size_t s_, ExpressionActionsPtr expression_, const NamesAndTypes & arguments_, DataTypePtr return_type_, std::string return_name_)
 		: IColumnDummy(s_), expression(expression_), arguments(arguments_), return_type(return_type_), return_name(return_name_) {}
-	
+
 	ColumnExpression(size_t s_, ExpressionActionsPtr expression_, const NamesAndTypesList & arguments_, DataTypePtr return_type_, std::string return_name_)
 		: IColumnDummy(s_), expression(expression_), arguments(arguments_.begin(), arguments_.end()), return_type(return_type_), return_name(return_name_) {}
-	
-	std::string getName() const { return "ColumnExpression"; }
-	ColumnPtr cloneDummy(size_t s_) const { return new ColumnExpression(s_, expression, arguments, return_type, return_name); }
-	
+
+	std::string getName() const override { return "ColumnExpression"; }
+	ColumnPtr cloneDummy(size_t s_) const override { return new ColumnExpression(s_, expression, arguments, return_type, return_name); }
+
 	const ExpressionActionsPtr & getExpression() const { return expression; }
 	const DataTypePtr & getReturnType() const { return return_type; }
 	const std::string & getReturnName() const { return return_name; }
-	
+
 	const NamesAndTypes & getArguments() const { return arguments; }
-	
+
 	Names getArgumentNames() const
 	{
 		Names res(arguments.size());
@@ -35,12 +35,12 @@ public:
 			res[i] = arguments[i].name;
 		return res;
 	}
-	
+
 private:
 	ExpressionActionsPtr expression;
 	NamesAndTypes arguments;
 	DataTypePtr return_type;
 	std::string return_name;
 };
-	
+
 }

dbms/include/DB/Columns/ColumnFixedString.h

@@ -27,49 +27,49 @@ public:
 	/** Создать пустой столбец строк фиксированной длины n */
 	ColumnFixedString(size_t n_) : n(n_) {}
 
-	std::string getName() const { return "ColumnFixedString"; }
+	std::string getName() const override { return "ColumnFixedString"; }
 
-	ColumnPtr cloneEmpty() const
+	ColumnPtr cloneEmpty() const override
 	{
 		return new ColumnFixedString(n);
 	}
 
-	size_t size() const
+	size_t size() const override
 	{
 		return chars.size() / n;
 	}
 
-	size_t sizeOfField() const
+	size_t sizeOfField() const override
 	{
 		return n;
 	}
 
-	bool isFixed() const
+	bool isFixed() const override
 	{
 		return true;
 	}
 
-	size_t byteSize() const
+	size_t byteSize() const override
 	{
 		return chars.size() + sizeof(n);
 	}
 
-	Field operator[](size_t index) const
+	Field operator[](size_t index) const override
 	{
 		return String(reinterpret_cast<const char *>(&chars[n * index]), n);
 	}
 
-	void get(size_t index, Field & res) const
+	void get(size_t index, Field & res) const override
 	{
 		res.assignString(reinterpret_cast<const char *>(&chars[n * index]), n);
 	}
 
-	StringRef getDataAt(size_t index) const
+	StringRef getDataAt(size_t index) const override
 	{
 		return StringRef(&chars[n * index], n);
 	}
 
-	void insert(const Field & x)
+	void insert(const Field & x) override
 	{
 		const String & s = DB::get<const String &>(x);
 
@@ -81,7 +81,7 @@ public:
 		memcpy(&chars[old_size], s.data(), s.size());
 	}
 
-	void insertFrom(const IColumn & src_, size_t index)
+	void insertFrom(const IColumn & src_, size_t index) override
 	{
 		const ColumnFixedString & src = static_cast<const ColumnFixedString &>(src_);
 
@@ -93,7 +93,7 @@ public:
 		memcpy(&chars[old_size], &src.chars[n * index], n);
 	}
 
-	void insertData(const char * pos, size_t length)
+	void insertData(const char * pos, size_t length) override
 	{
 		if (length > n)
 			throw Exception("Too large string for FixedString column", ErrorCodes::TOO_LARGE_STRING_SIZE);
@@ -103,12 +103,12 @@ public:
 		memcpy(&chars[old_size], pos, length);
 	}
 
-	void insertDefault()
+	void insertDefault() override
 	{
 		chars.resize_fill(chars.size() + n);
 	}
 
-	int compareAt(size_t p1, size_t p2, const IColumn & rhs_, int nan_direction_hint) const
+	int compareAt(size_t p1, size_t p2, const IColumn & rhs_, int nan_direction_hint) const override
 	{
 		const ColumnFixedString & rhs = static_cast<const ColumnFixedString &>(rhs_);
 		return memcmp(&chars[p1 * n], &rhs.chars[p2 * n], n);
@@ -127,7 +127,7 @@ public:
 		}
 	};
 
-	void getPermutation(bool reverse, size_t limit, Permutation & res) const
+	void getPermutation(bool reverse, size_t limit, Permutation & res) const override
 	{
 		size_t s = size();
 		res.resize(s);
@@ -153,7 +153,7 @@ public:
 		}
 	}
 
-	ColumnPtr cut(size_t start, size_t length) const
+	ColumnPtr cut(size_t start, size_t length) const override
 	{
 		ColumnFixedString * res_ = new ColumnFixedString(n);
 		ColumnPtr res = res_;
@@ -162,7 +162,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr filter(const IColumn::Filter & filt) const
+	ColumnPtr filter(const IColumn::Filter & filt) const override
 	{
 		size_t col_size = size();
 		if (col_size != filt.size())
@@ -185,7 +185,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr permute(const Permutation & perm, size_t limit) const
+	ColumnPtr permute(const Permutation & perm, size_t limit) const override
 	{
 		size_t col_size = size();
 
@@ -214,7 +214,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr replicate(const Offsets_t & offsets) const
+	ColumnPtr replicate(const Offsets_t & offsets) const override
 	{
 		size_t col_size = size();
 		if (col_size != offsets.size())
@@ -243,7 +243,7 @@ public:
 		return res;
 	}
 
-	void getExtremes(Field & min, Field & max) const
+	void getExtremes(Field & min, Field & max) const override
 	{
 		min = String();
 		max = String();

dbms/include/DB/Columns/ColumnNested.h

@@ -47,7 +47,7 @@ public:
 		}
 	}
 
-	std::string getName() const
+	std::string getName() const override
 	{
 		std::string res;
 		{
@@ -63,7 +63,7 @@ public:
 		return "ColumnNested(" + res + ")";
 	}
 
-	ColumnPtr cloneEmpty() const
+	ColumnPtr cloneEmpty() const override
 	{
 		Columns res(data.size());
 		for (size_t i = 0; i < data.size(); ++i)
@@ -71,32 +71,32 @@ public:
 		return new ColumnNested(res);
 	}
 
-	size_t size() const
+	size_t size() const override
 	{
 		return getOffsets().size();
 	}
 
-	Field operator[](size_t n) const
+	Field operator[](size_t n) const override
 	{
 		throw Exception("Method operator[] is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	void get(size_t n, Field & res) const
+	void get(size_t n, Field & res) const override
 	{
 		throw Exception("Method get is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	StringRef getDataAt(size_t n) const
+	StringRef getDataAt(size_t n) const override
 	{
 		throw Exception("Method getDataAt is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	void insertData(const char * pos, size_t length)
+	void insertData(const char * pos, size_t length) override
 	{
 		throw Exception("Method insertData is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	ColumnPtr cut(size_t start, size_t length) const
+	ColumnPtr cut(size_t start, size_t length) const override
 	{
 		if (length == 0)
 			return new ColumnNested(data);
@@ -131,12 +131,12 @@ public:
 		return res;
 	}
 
-	void insert(const Field & x)
+	void insert(const Field & x) override
 	{
 		throw Exception("Method insert is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	void insertFrom(const IColumn & src_, size_t n)
+	void insertFrom(const IColumn & src_, size_t n) override
 	{
 		const ColumnNested & src = static_cast<const ColumnNested &>(src_);
 
@@ -158,14 +158,14 @@ public:
 		getOffsets().push_back((getOffsets().size() == 0 ? 0 : getOffsets().back()) + size);
 	}
 
-	void insertDefault()
+	void insertDefault() override
 	{
 		for (size_t i = 0; i < data.size(); ++i)
 			data[i]->insertDefault();
 		getOffsets().push_back(getOffsets().size() == 0 ? 1 : (getOffsets().back() + 1));
 	}
 
-	ColumnPtr filter(const Filter & filt) const
+	ColumnPtr filter(const Filter & filt) const override
 	{
 		size_t size = getOffsets().size();
 		if (size != filt.size())
@@ -205,12 +205,12 @@ public:
 		return res;
 	}
 
-	ColumnPtr replicate(const Offsets_t & offsets) const
+	ColumnPtr replicate(const Offsets_t & offsets) const override
 	{
 		throw Exception("Replication of ColumnNested is not implemented.", ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	ColumnPtr permute(const Permutation & perm, size_t limit) const
+	ColumnPtr permute(const Permutation & perm, size_t limit) const override
 	{
 		size_t size = getOffsets().size();
 		if (size != perm.size())
@@ -255,24 +255,24 @@ public:
 		return res;
 	}
 
-	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const
+	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const override
 	{
 		throw Exception("Method compareAt is not supported for ColumnNested.", ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	void getPermutation(bool reverse, size_t limit, Permutation & res) const
+	void getPermutation(bool reverse, size_t limit, Permutation & res) const override
 	{
 		throw Exception("Method getPermutation is not supported for ColumnNested.", ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	void reserve(size_t n)
+	void reserve(size_t n) override
 	{
 		getOffsets().reserve(n);
 		for (Columns::iterator it = data.begin(); it != data.end(); ++it)
 			(*it)->reserve(n);
 	}
 
-	size_t byteSize() const
+	size_t byteSize() const override
 	{
 		size_t size = getOffsets().size() * sizeof(getOffsets()[0]);
 		for (Columns::const_iterator it = data.begin(); it != data.end(); ++it)
@@ -280,7 +280,7 @@ public:
 		return size;
 	}
 
-	void getExtremes(Field & min, Field & max) const
+	void getExtremes(Field & min, Field & max) const override
 	{
 		throw Exception("Method getExtremes is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
@@ -289,9 +289,6 @@ public:
 	Columns & getData() { return data; }
 	const Columns & getData() const { return data; }
 
-// 	ColumnPtr & getDataPtr() { return data; }
-// 	const ColumnPtr & getDataPtr() const { return data; }
-
 	Offsets_t & ALWAYS_INLINE getOffsets()
 	{
 		return static_cast<ColumnOffsets_t &>(*offsets.get()).getData();

dbms/include/DB/Columns/ColumnReplicated.h

@@ -5,7 +5,7 @@
 
 namespace DB
 {
-	
+
 /** Содержит промежуточные данные для вычисления выражений в функциях высшего порядка.
   * Это - вложенный столбец произвольного размера.
   * Сам ColumnReplicated притворяется, как столбец указанного в конструкторе размера.
@@ -14,8 +14,8 @@ class ColumnReplicated final : public IColumnDummy
 {
 public:
 	ColumnReplicated(size_t s_, ColumnPtr nested_) : IColumnDummy(s_), nested(nested_) {}
-	std::string getName() const { return "ColumnReplicated"; }
-	ColumnPtr cloneDummy(size_t s_) const { return new ColumnReplicated(s_, nested); }
+	std::string getName() const override { return "ColumnReplicated"; }
+	ColumnPtr cloneDummy(size_t s_) const override { return new ColumnReplicated(s_, nested); }
 
 	ColumnPtr & getData() { return nested; }
 private:

dbms/include/DB/Columns/ColumnSet.h

@@ -17,11 +17,11 @@ public:
 	ColumnSet(size_t s_, SetPtr data_) : IColumnDummy(s_), data(data_) {}
 
 	/// Столбец не константный. Иначе столбец будет использоваться в вычислениях в ExpressionActions::prepare, когда множество из подзапроса ещё не готово.
-	bool isConst() const { return false; }
+	bool isConst() const override { return false; }
+
+	std::string getName() const override { return "ColumnSet"; }
+	ColumnPtr cloneDummy(size_t s_) const override { return new ColumnSet(s_, data); }
 
-	std::string getName() const { return "ColumnSet"; }
-	ColumnPtr cloneDummy(size_t s_) const { return new ColumnSet(s_, data); }
-	
 	SetPtr & getData() { return data; }
 	const SetPtr & getData() const { return data; }
 

dbms/include/DB/Columns/ColumnString.h

@@ -36,44 +36,44 @@ public:
 	/** Создать пустой столбец строк */
 	ColumnString() {}
 
-	std::string getName() const { return "ColumnString"; }
+	std::string getName() const override { return "ColumnString"; }
 
-	size_t size() const
+	size_t size() const override
 	{
 		return offsets.size();
 	}
 
-	size_t byteSize() const
+	size_t byteSize() const override
 	{
 		return chars.size() + offsets.size() * sizeof(offsets[0]);
 	}
 
-	ColumnPtr cloneEmpty() const
+	ColumnPtr cloneEmpty() const override
 	{
 		return new ColumnString;
 	}
 
-	Field operator[](size_t n) const
+	Field operator[](size_t n) const override
 	{
 		return Field(&chars[offsetAt(n)], sizeAt(n) - 1);
 	}
 
-	void get(size_t n, Field & res) const
+	void get(size_t n, Field & res) const override
 	{
 		res.assignString(&chars[offsetAt(n)], sizeAt(n) - 1);
 	}
 
-	StringRef getDataAt(size_t n) const
+	StringRef getDataAt(size_t n) const override
 	{
 		return StringRef(&chars[offsetAt(n)], sizeAt(n) - 1);
 	}
 
-	StringRef getDataAtWithTerminatingZero(size_t n) const
+	StringRef getDataAtWithTerminatingZero(size_t n) const override
 	{
 		return StringRef(&chars[offsetAt(n)], sizeAt(n));
 	}
 
-	void insert(const Field & x)
+	void insert(const Field & x) override
 	{
 		const String & s = DB::get<const String &>(x);
 		size_t old_size = chars.size();
@@ -84,7 +84,7 @@ public:
 		offsets.push_back((offsets.size() == 0 ? 0 : offsets.back()) + size_to_append);
 	}
 
-	void insertFrom(const IColumn & src_, size_t n)
+	void insertFrom(const IColumn & src_, size_t n) override
 	{
 		const ColumnString & src = static_cast<const ColumnString &>(src_);
 		size_t old_size = chars.size();
@@ -96,7 +96,7 @@ public:
 		offsets.push_back((offsets.size() == 0 ? 0 : offsets.back()) + size_to_append);
 	}
 
-	void insertData(const char * pos, size_t length)
+	void insertData(const char * pos, size_t length) override
 	{
 		size_t old_size = chars.size();
 
@@ -106,7 +106,7 @@ public:
 		offsets.push_back((offsets.size() == 0 ? 0 : offsets.back()) + length + 1);
 	}
 
-	void insertDataWithTerminatingZero(const char * pos, size_t length)
+	void insertDataWithTerminatingZero(const char * pos, size_t length) override
 	{
 		size_t old_size = chars.size();
 
@@ -115,7 +115,7 @@ public:
 		offsets.push_back((offsets.size() == 0 ? 0 : offsets.back()) + length);
 	}
 
-	ColumnPtr cut(size_t start, size_t length) const
+	ColumnPtr cut(size_t start, size_t length) const override
 	{
 		if (length == 0)
 			return new ColumnString;
@@ -150,7 +150,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr filter(const Filter & filt) const
+	ColumnPtr filter(const Filter & filt) const override
 	{
 		const size_t size = offsets.size();
 		if (size != filt.size())
@@ -257,7 +257,7 @@ public:
 		return res_;
 	}
 
-	ColumnPtr permute(const Permutation & perm, size_t limit) const
+	ColumnPtr permute(const Permutation & perm, size_t limit) const override
 	{
 		size_t size = offsets.size();
 
@@ -300,13 +300,13 @@ public:
 		return res;
 	}
 
-	void insertDefault()
+	void insertDefault() override
 	{
 		chars.push_back(0);
 		offsets.push_back(offsets.size() == 0 ? 1 : (offsets.back() + 1));
 	}
 
-	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const
+	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const override
 	{
 		const ColumnString & rhs = static_cast<const ColumnString &>(rhs_);
 
@@ -344,7 +344,7 @@ public:
 		}
 	};
 
-	void getPermutation(bool reverse, size_t limit, Permutation & res) const
+	void getPermutation(bool reverse, size_t limit, Permutation & res) const override
 	{
 		size_t s = offsets.size();
 		res.resize(s);
@@ -415,7 +415,7 @@ public:
 		}
 	}
 
-	ColumnPtr replicate(const Offsets_t & replicate_offsets) const
+	ColumnPtr replicate(const Offsets_t & replicate_offsets) const override
 	{
 		size_t col_size = size();
 		if (col_size != replicate_offsets.size())
@@ -457,13 +457,13 @@ public:
 		return res;
 	}
 
-	void reserve(size_t n)
+	void reserve(size_t n) override
 	{
 		offsets.reserve(n);
 		chars.reserve(n * DBMS_APPROX_STRING_SIZE);
 	}
 
-	void getExtremes(Field & min, Field & max) const
+	void getExtremes(Field & min, Field & max) const override
 	{
 		min = String();
 		max = String();

dbms/include/DB/Columns/ColumnTuple.h

@@ -25,22 +25,20 @@ public:
 		for (size_t i = 0; i < size; ++i)
 			columns[i] = data.getByPosition(i).column;
 	}
-	
-	std::string getName() const { return "Tuple"; }
-	
-	SharedPtr<IColumn> cloneEmpty() const
+
+	std::string getName() const override { return "Tuple"; }
+
+	SharedPtr<IColumn> cloneEmpty() const override
 	{
 		return new ColumnTuple(data.cloneEmpty());
 	}
 
-	size_t size() const
+	size_t size() const override
 	{
 		return data.rows();
 	}
 
-	bool empty() const { return size() == 0; }
-
-	Field operator[](size_t n) const
+	Field operator[](size_t n) const override
 	{
 		Array res;
 
@@ -50,7 +48,7 @@ public:
 		return res;
 	}
 
-	void get(size_t n, Field & res) const
+	void get(size_t n, Field & res) const override
 	{
 		size_t size = columns.size();
 		res = Array(size);
@@ -59,17 +57,17 @@ public:
 			columns[i]->get(n, res_arr[i]);
 	}
 
-	StringRef getDataAt(size_t n) const
+	StringRef getDataAt(size_t n) const override
 	{
 		throw Exception("Method getDataAt is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	void insertData(const char * pos, size_t length)
+	void insertData(const char * pos, size_t length) override
 	{
 		throw Exception("Method insertData is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	void insert(const Field & x)
+	void insert(const Field & x) override
 	{
 		const Array & arr = DB::get<const Array &>(x);
 
@@ -81,36 +79,36 @@ public:
 			columns[i]->insert(arr[i]);
 	}
 
-	void insertFrom(const IColumn & src_, size_t n)
+	void insertFrom(const IColumn & src_, size_t n) override
 	{
 		const ColumnTuple & src = static_cast<const ColumnTuple &>(src_);
-		
+
 		size_t size = columns.size();
 		if (src.columns.size() != size)
 			throw Exception("Cannot insert value of different size into tuple", ErrorCodes::CANNOT_INSERT_VALUE_OF_DIFFERENT_SIZE_INTO_TUPLE);
-			
+
 		for (size_t i = 0; i < size; ++i)
 			columns[i]->insertFrom(*src.columns[i], n);
 	}
 
-	void insertDefault()
+	void insertDefault() override
 	{
 		for (Columns::iterator it = columns.begin(); it != columns.end(); ++it)
 			(*it)->insertDefault();
 	}
 
 
-	ColumnPtr cut(size_t start, size_t length) const
+	ColumnPtr cut(size_t start, size_t length) const override
 	{
 		Block res_block = data.cloneEmpty();
-		
+
 		for (size_t i = 0; i < columns.size(); ++i)
 			res_block.getByPosition(i).column = data.getByPosition(i).column->cut(start, length);
 
 		return new ColumnTuple(res_block);
 	}
 
-	ColumnPtr filter(const Filter & filt) const
+	ColumnPtr filter(const Filter & filt) const override
 	{
 		Block res_block = data.cloneEmpty();
 
@@ -120,7 +118,7 @@ public:
 		return new ColumnTuple(res_block);
 	}
 
-	ColumnPtr permute(const Permutation & perm, size_t limit) const
+	ColumnPtr permute(const Permutation & perm, size_t limit) const override
 	{
 		Block res_block = data.cloneEmpty();
 
@@ -129,8 +127,8 @@ public:
 
 		return new ColumnTuple(res_block);
 	}
-	
-	ColumnPtr replicate(const Offsets_t & offsets) const
+
+	ColumnPtr replicate(const Offsets_t & offsets) const override
 	{
 		Block res_block = data.cloneEmpty();
 
@@ -140,13 +138,13 @@ public:
 		return new ColumnTuple(res_block);
 	}
 
-	int compareAt(size_t n, size_t m, const IColumn & rhs, int nan_direction_hint) const
+	int compareAt(size_t n, size_t m, const IColumn & rhs, int nan_direction_hint) const override
 	{
 		size_t size = columns.size();
 		for (size_t i = 0; i < size; ++i)
 			if (int res = columns[i]->compareAt(n, m, *static_cast<const ColumnTuple &>(rhs).columns[i], nan_direction_hint))
 				return res;
-		
+
 		return 0;
 	}
 
@@ -175,7 +173,7 @@ public:
 		}
 	};
 
-	void getPermutation(bool reverse, size_t limit, Permutation & res) const
+	void getPermutation(bool reverse, size_t limit, Permutation & res) const override
 	{
 		size_t rows = size();
 		res.resize(rows);
@@ -201,13 +199,13 @@ public:
 		}
 	}
 
-	void reserve(size_t n)
+	void reserve(size_t n) override
 	{
 		for (Columns::iterator it = columns.begin(); it != columns.end(); ++it)
 			(*it)->reserve(n);
 	}
 
-	size_t byteSize() const
+	size_t byteSize() const override
 	{
 		size_t res = 0;
 		for (Columns::const_iterator it = columns.begin(); it != columns.end(); ++it)
@@ -215,7 +213,7 @@ public:
 		return res;
 	}
 
-	void getExtremes(Field & min, Field & max) const
+	void getExtremes(Field & min, Field & max) const override
 	{
 		throw Exception("Method getExtremes is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}

dbms/include/DB/Columns/ColumnVector.h

@@ -77,7 +77,7 @@ template <> struct CompareHelper<Float64> : public FloatCompareHelper<Float64> {
 /** Шаблон столбцов, которые используют для хранения простой массив.
   */
 template <typename T>
-class ColumnVector : public IColumn
+class ColumnVector final : public IColumn
 {
 private:
 	typedef ColumnVector<T> Self;
@@ -86,44 +86,45 @@ public:
 	typedef PODArray<value_type> Container_t;
 
 	ColumnVector() {}
-	ColumnVector(size_t n) : data(n) {}
+	ColumnVector(const size_t n) : data{n} {}
+	ColumnVector(const size_t n, const value_type x) : data{n, x} {}
 
-	bool isNumeric() const { return IsNumber<T>::value; }
-	bool isFixed() const { return IsNumber<T>::value; }
+	bool isNumeric() const override { return IsNumber<T>::value; }
+	bool isFixed() const override { return IsNumber<T>::value; }
 
-	size_t sizeOfField() const { return sizeof(T); }
+	size_t sizeOfField() const override { return sizeof(T); }
 
-	size_t size() const
+	size_t size() const override
 	{
 		return data.size();
 	}
 
-	StringRef getDataAt(size_t n) const
+	StringRef getDataAt(size_t n) const override
 	{
 		return StringRef(reinterpret_cast<const char *>(&data[n]), sizeof(data[n]));
 	}
 
-	void insertFrom(const IColumn & src, size_t n)
+	void insertFrom(const IColumn & src, size_t n) override
 	{
 		data.push_back(static_cast<const Self &>(src).getData()[n]);
 	}
 
-	void insertData(const char * pos, size_t length)
+	void insertData(const char * pos, size_t length) override
 	{
 		data.push_back(*reinterpret_cast<const T *>(pos));
 	}
 
-	void insertDefault()
+	void insertDefault() override
 	{
 		data.push_back(T());
 	}
 
-	size_t byteSize() const
+	size_t byteSize() const override
 	{
 		return data.size() * sizeof(data[0]);
 	}
 
-	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const
+	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const override
 	{
 		return CompareHelper<T>::compare(data[n], static_cast<const Self &>(rhs_).data[m], nan_direction_hint);
 	}
@@ -142,7 +143,7 @@ public:
 		bool operator()(size_t lhs, size_t rhs) const { return CompareHelper<T>::greater(parent.data[lhs], parent.data[rhs]); }
 	};
 
-	void getPermutation(bool reverse, size_t limit, Permutation & res) const
+	void getPermutation(bool reverse, size_t limit, Permutation & res) const override
 	{
 		size_t s = data.size();
 		res.resize(s);
@@ -168,36 +169,36 @@ public:
 		}
 	}
 
-	void reserve(size_t n)
+	void reserve(size_t n) override
 	{
 		data.reserve(n);
 	}
 
-	std::string getName() const { return "ColumnVector<" + TypeName<T>::get() + ">"; }
+	std::string getName() const override { return "ColumnVector<" + TypeName<T>::get() + ">"; }
 
-	ColumnPtr cloneEmpty() const
+	ColumnPtr cloneEmpty() const override
 	{
 		return new ColumnVector<T>;
 	}
 
-	Field operator[](size_t n) const
+	Field operator[](size_t n) const override
 	{
 		return typename NearestFieldType<T>::Type(data[n]);
 	}
 
-	void get(size_t n, Field & res) const
+	void get(size_t n, Field & res) const override
 	{
 		res = typename NearestFieldType<T>::Type(data[n]);
 	}
 
-	UInt64 get64(size_t n) const;
+	UInt64 get64(size_t n) const override;
 
-	void insert(const Field & x)
+	void insert(const Field & x) override
 	{
 		data.push_back(DB::get<typename NearestFieldType<T>::Type>(x));
 	}
 
-	ColumnPtr cut(size_t start, size_t length) const
+	ColumnPtr cut(size_t start, size_t length) const override
 	{
 		if (start + length > data.size())
 			throw Exception("Parameters start = "
@@ -211,7 +212,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr filter(const IColumn::Filter & filt) const
+	ColumnPtr filter(const IColumn::Filter & filt) const override
 	{
 		size_t size = data.size();
 		if (size != filt.size())
@@ -269,7 +270,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr permute(const IColumn::Permutation & perm, size_t limit) const
+	ColumnPtr permute(const IColumn::Permutation & perm, size_t limit) const override
 	{
 		size_t size = data.size();
 
@@ -290,7 +291,7 @@ public:
 		return res;
 	}
 
-	ColumnPtr replicate(const IColumn::Offsets_t & offsets) const
+	ColumnPtr replicate(const IColumn::Offsets_t & offsets) const override
 	{
 		size_t size = data.size();
 		if (size != offsets.size())
@@ -317,7 +318,7 @@ public:
 		return res;
 	}
 
-	void getExtremes(Field & min, Field & max) const
+	void getExtremes(Field & min, Field & max) const override
 	{
 		size_t size = data.size();
 

dbms/include/DB/Columns/IColumnDummy.h

@@ -17,30 +17,30 @@ public:
 
 	virtual ColumnPtr cloneDummy(size_t s_) const = 0;
 
-	ColumnPtr cloneResized(size_t s_) const { return cloneDummy(s_); }
-	bool isConst() const { return true; }
-	size_t size() const { return s; }
-	void insertDefault() { ++s; }
-	size_t byteSize() const { return 0; }
-	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const { return 0; }
+	ColumnPtr cloneResized(size_t s_) const override { return cloneDummy(s_); }
+	bool isConst() const override { return true; }
+	size_t size() const override { return s; }
+	void insertDefault() override { ++s; }
+	size_t byteSize() const override { return 0; }
+	int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const override { return 0; }
 
-	Field operator[](size_t n) const { throw Exception("Cannot get value from " + getName(), ErrorCodes::NOT_IMPLEMENTED); }
-	void get(size_t n, Field & res) const { throw Exception("Cannot get value from " + getName(), ErrorCodes::NOT_IMPLEMENTED); };
-	void insert(const Field & x) { throw Exception("Cannot insert element into " + getName(), ErrorCodes::NOT_IMPLEMENTED); }
-	StringRef getDataAt(size_t n) const { throw Exception("Method getDataAt is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED); }
-	void insertData(const char * pos, size_t length) { throw Exception("Method insertData is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED); }
+	Field operator[](size_t n) const override { throw Exception("Cannot get value from " + getName(), ErrorCodes::NOT_IMPLEMENTED); }
+	void get(size_t n, Field & res) const override { throw Exception("Cannot get value from " + getName(), ErrorCodes::NOT_IMPLEMENTED); };
+	void insert(const Field & x) override { throw Exception("Cannot insert element into " + getName(), ErrorCodes::NOT_IMPLEMENTED); }
+	StringRef getDataAt(size_t n) const override { throw Exception("Method getDataAt is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED); }
+	void insertData(const char * pos, size_t length) override { throw Exception("Method insertData is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED); }
 
-	void getExtremes(Field & min, Field & max) const
+	void getExtremes(Field & min, Field & max) const override
 	{
 		throw Exception("Method getExtremes is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 	}
 
-	ColumnPtr cut(size_t start, size_t length) const
+	ColumnPtr cut(size_t start, size_t length) const override
 	{
 		return cloneDummy(length);
 	}
 
-	ColumnPtr filter(const Filter & filt) const
+	ColumnPtr filter(const Filter & filt) const override
 	{
 		size_t new_size = 0;
 		for (Filter::const_iterator it = filt.begin(); it != filt.end(); ++it)
@@ -50,7 +50,7 @@ public:
 			return cloneDummy(new_size);
 	}
 
-	ColumnPtr permute(const Permutation & perm, size_t limit) const
+	ColumnPtr permute(const Permutation & perm, size_t limit) const override
 	{
 		if (s != perm.size())
 			throw Exception("Size of permutation doesn't match size of column.", ErrorCodes::SIZES_OF_COLUMNS_DOESNT_MATCH);
@@ -58,14 +58,14 @@ public:
 		return cloneDummy(limit ? std::min(s, limit) : s);
 	}
 
-	void getPermutation(bool reverse, size_t limit, Permutation & res) const
+	void getPermutation(bool reverse, size_t limit, Permutation & res) const override
 	{
 		res.resize(s);
 		for (size_t i = 0; i < s; ++i)
 			res[i] = i;
 	}
 
-	ColumnPtr replicate(const Offsets_t & offsets) const
+	ColumnPtr replicate(const Offsets_t & offsets) const override
 	{
 		if (s != offsets.size())
 			throw Exception("Size of offsets doesn't match size of column.", ErrorCodes::SIZES_OF_COLUMNS_DOESNT_MATCH);

dbms/include/DB/Common/HashTable/HashTable.h

@@ -455,7 +455,7 @@ public:
 	const_iterator begin() const
 	{
 		if (this->hasZero())
-			return const_iterator(this, this->zeroValue());
+			return iteratorToZero();
 
 		const Cell * ptr = buf;
 		while (ptr < buf + grower.bufSize() && ptr->isZero(*this))
@@ -467,7 +467,7 @@ public:
 	iterator begin()
 	{
 		if (this->hasZero())
-			return iterator(this, this->zeroValue());
+			return iteratorToZero();
 
 		Cell * ptr = buf;
 		while (ptr < buf + grower.bufSize() && ptr->isZero(*this))
@@ -481,6 +481,10 @@ public:
 
 
 protected:
+	const_iterator iteratorToZero() const 	{ return const_iterator(this, this->zeroValue()); }
+	iterator iteratorToZero() 				{ return iterator(this, this->zeroValue()); }
+
+
 	/// Если ключ нулевой - вставить его в специальное место и вернуть true.
 	bool emplaceIfZero(Key x, iterator & it, bool & inserted)
 	{
@@ -490,17 +494,17 @@ protected:
 
 		if (Cell::isZero(x, *this))
 		{
+			it = iteratorToZero();
 			if (!this->hasZero())
 			{
 				++m_size;
 				this->setHasZero();
+				it.ptr->setHash(hash(x));
 				inserted = true;
 			}
 			else
 				inserted = false;
 
-			it = begin();
-			it.ptr->setHash(hash(x));
 			return true;
 		}
 
@@ -545,7 +549,7 @@ public:
 
 		if (res.second)
 			res.first.ptr->setMapped(x);
-		
+
 		return res;
 	}
 
@@ -583,7 +587,7 @@ public:
 	iterator find(Key x)
 	{
 		if (Cell::isZero(x, *this))
-			return this->hasZero() ? begin() : end();
+			return this->hasZero() ? iteratorToZero() : end();
 
 		size_t place_value = findCell(x, grower.place(hash(x)));
 
@@ -594,7 +598,7 @@ public:
 	const_iterator find(Key x) const
 	{
 		if (Cell::isZero(x, *this))
-			return this->hasZero() ? begin() : end();
+			return this->hasZero() ? iteratorToZero() : end();
 
 		size_t place_value = findCell(x, grower.place(hash(x)));
 

dbms/include/DB/Common/SipHash.h

@@ -30,7 +30,7 @@ class SipHash
 private:
 	typedef uint64_t u64;
 	typedef uint8_t u8;
-	
+
 	/// Состояние.
 	u64 v0;
 	u64 v1;
@@ -131,7 +131,7 @@ public:
 	}
 
 	/// Получить результат в некотором виде. Это можно сделать только один раз!
-	
+
 	void get128(char * out)
 	{
 		finalize();
@@ -158,7 +158,14 @@ public:
 #undef SIPROUND
 
 
-inline uint64_t sipHash64(const char * data, size_t size)
+inline void sipHash128(const char * data, const size_t size, char * out)
+{
+	SipHash hash;
+	hash.update(data, size);
+	hash.get128(out);
+}
+
+inline uint64_t sipHash64(const char * data, const size_t size)
 {
 	SipHash hash;
 	hash.update(data, size);

dbms/include/DB/Common/UnicodeBar.h

@@ -0,0 +1,62 @@
+#pragma once
+
+#include <cstring>
+#include <cmath>
+#include <string>
+
+
+#define UNICODE_BAR_CHAR_SIZE (strlen("█"))
+
+
+/** Позволяет нарисовать unicode-art полоску, ширина которой отображается с разрешением 1/8 символа.
+  */
+
+
+namespace UnicodeBar
+{
+	template <typename T>
+	double getWidth(T x, int64_t min, int64_t max, double max_width)
+	{
+		if (x <= min)
+			return 0;
+
+		if (x >= max)
+			return max_width;
+
+		return (x - min) * max_width / (max - min);
+	}
+
+	inline size_t getWidthInBytes(double width)
+	{
+		return ceil(width - 1.0 / 8) * UNICODE_BAR_CHAR_SIZE;
+	}
+
+	/// В dst должно быть место для barWidthInBytes(width) символов и завершающего нуля.
+	inline void render(double width, char * dst)
+	{
+		size_t floor_width = floor(width);
+
+		for (size_t i = 0; i < floor_width; ++i)
+		{
+			memcpy(dst, "█", UNICODE_BAR_CHAR_SIZE);
+			dst += UNICODE_BAR_CHAR_SIZE;
+		}
+
+		size_t remainder = floor((width - floor_width) * 8);
+
+		if (remainder)
+		{
+			memcpy(dst, &"▏▎▍▌▋▋▊▉"[(remainder - 1) * UNICODE_BAR_CHAR_SIZE], UNICODE_BAR_CHAR_SIZE);
+			dst += UNICODE_BAR_CHAR_SIZE;
+		}
+
+		*dst = 0;
+	}
+
+	inline std::string render(double width)
+	{
+		std::string res(getWidthInBytes(width), '\0');
+		render(width, &res[0]);
+		return res;
+	}
+}

dbms/include/DB/Common/formatReadable.h

@@ -0,0 +1,13 @@
+#pragma once
+
+#include <string>
+
+
+/// Выводит переданный размер в байтах в виде 123.45 GiB.
+std::string formatReadableSizeWithBinarySuffix(double value, int precision = 2);
+
+/// Выводит переданный размер в байтах в виде 132.55 GB.
+std::string formatReadableSizeWithDecimalSuffix(double value, int precision = 2);
+
+/// Выводит число в виде 123.45 billion.
+std::string formatReadableQuantity(double value, int precision = 2);

dbms/include/DB/Core/Defines.h

@@ -64,7 +64,9 @@
 #define DBMS_MIN_REVISION_WITH_TOTALS_EXTREMES					35265
 #define DBMS_MIN_REVISION_WITH_STRING_QUERY_ID					39002
 #define DBMS_MIN_REVISION_WITH_TEMPORARY_TABLES					50264
+#define DBMS_MIN_REVISION_WITH_TOTAL_ROWS_IN_PROGRESS			51554
 
 #define DBMS_DISTRIBUTED_DIRECTORY_MONITOR_SLEEP_TIME_MS		100
 
 #define ALWAYS_INLINE 	__attribute__((__always_inline__))
+#define NO_INLINE 		__attribute__((__noinline__))

dbms/include/DB/Core/Progress.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <DB/Core/Defines.h>
 #include <DB/IO/ReadBuffer.h>
 #include <DB/IO/WriteBuffer.h>
 #include <DB/IO/ReadHelpers.h>
@@ -10,25 +11,72 @@ namespace DB
 {
 
 
-/// Прогресс выполнения запроса
+/** Прогресс выполнения запроса.
+  * Передаваемые по сети значения представляют собой разницу - сколько было сделано после предыдущего отправленного значения.
+  * Тот же объект используется для суммирования полученных значений.
+  */
 struct Progress
 {
-	size_t rows;	/// Строк обработано.
-	size_t bytes;	/// Байт обработано.
+	size_t rows = 0;		/// Строк обработано.
+	size_t bytes = 0;		/// Байт обработано.
 
-	Progress() : rows(0), bytes(0) {}
-	Progress(size_t rows_, size_t bytes_) : rows(rows_), bytes(bytes_) {}
+	/** Сколько ещё строк надо обработать, приблизительно. Передаётся не ноль, когда возникает информация о какой-то новой части работы.
+	  * Полученные значения надо суммровать, чтобы получить оценку общего количества строк для обработки.
+	  * Используется для отображения прогресс-бара на клиенте.
+	  */
+	size_t total_rows = 0;
 
-	void read(ReadBuffer & in)
+	Progress() {}
+	Progress(size_t rows_, size_t bytes_, size_t total_rows_ = 0)
+		: rows(rows_), bytes(bytes_), total_rows(total_rows_) {}
+
+	void read(ReadBuffer & in, UInt64 server_revision)
 	{
 		readVarUInt(rows, in);
 		readVarUInt(bytes, in);
+
+		if (server_revision >= DBMS_MIN_REVISION_WITH_TOTAL_ROWS_IN_PROGRESS)
+			readVarUInt(total_rows, in);
 	}
 
-	void write(WriteBuffer & out)
+	void write(WriteBuffer & out, UInt64 client_revision) const
 	{
 		writeVarUInt(rows, out);
 		writeVarUInt(bytes, out);
+
+		if (client_revision >= DBMS_MIN_REVISION_WITH_TOTAL_ROWS_IN_PROGRESS)
+			writeVarUInt(total_rows, out);
+	}
+
+	void increment(const Progress & rhs)
+	{
+		rows += rhs.rows;
+		bytes += rhs.bytes;
+		total_rows += rhs.total_rows;
+	}
+
+	/// Каждое значение по-отдельности изменяется атомарно.
+	void incrementPiecewiseAtomically(const Progress & rhs)
+	{
+		__sync_add_and_fetch(&rows, rhs.rows);
+		__sync_add_and_fetch(&bytes, rhs.bytes);
+		__sync_add_and_fetch(&total_rows, rhs.total_rows);
+	}
+
+	void reset()
+	{
+		*this = Progress();
+	}
+
+	Progress fetchAndResetPiecewiseAtomically()
+	{
+		Progress res;
+
+		res.rows = __sync_fetch_and_and(&rows, 0);
+		res.bytes = __sync_fetch_and_and(&bytes, 0);
+		res.total_rows = __sync_fetch_and_and(&total_rows, 0);
+
+		return res;
 	}
 };
 

dbms/include/DB/DataStreams/IBlockInputStream.h

@@ -5,6 +5,7 @@
 #include <Poco/SharedPtr.h>
 
 #include <DB/Core/Block.h>
+#include <DB/Core/Progress.h>
 #include <DB/Storages/IStorage.h>
 
 
@@ -18,7 +19,7 @@ using Poco::SharedPtr;
   * Функция принимает количество строк в последнем блоке, количество байт в последнем блоке.
   * Следует иметь ввиду, что колбэк может вызываться из разных потоков.
   */
-typedef std::function<void(size_t, size_t)> ProgressCallback;
+typedef std::function<void(const Progress & progress)> ProgressCallback;
 
 
 /** Интерфейс потока для чтения данных по блокам из БД.
@@ -29,9 +30,9 @@ class IBlockInputStream : private boost::noncopyable
 public:
 	typedef SharedPtr<IBlockInputStream> BlockInputStreamPtr;
 	typedef std::vector<BlockInputStreamPtr> BlockInputStreams;
-	
+
 	IBlockInputStream() {}
-	
+
 	/** Прочитать следующий блок.
 	  * Если блоков больше нет - вернуть пустой блок (для которого operator bool возвращает false).
 	  */
@@ -61,7 +62,7 @@ public:
 	virtual String getID() const = 0;
 
 	BlockInputStreams & getChildren() { return children; }
-	
+
 	void dumpTree(std::ostream & ostr, size_t indent = 0, size_t multiplier = 1);
 
 	/// Получить листовые источники (не считая этот).

dbms/include/DB/DataStreams/IProfilingBlockInputStream.h

@@ -5,6 +5,7 @@
 #include <statdaemons/Stopwatch.h>
 
 #include <DB/Core/Names.h>
+#include <DB/Core/Progress.h>
 
 #include <DB/Interpreters/Limits.h>
 #include <DB/Interpreters/Quota.h>
@@ -109,8 +110,8 @@ public:
 	  * - проверяются ограничения и квоты, которые должны быть проверены не в рамках одного источника,
 	  *   а над общим количеством потраченных ресурсов во всех источниках сразу (информация в ProcessList-е).
 	  */
-	virtual void progress(size_t rows, size_t bytes) { progressImpl(rows, bytes); }
-	void progressImpl(size_t rows, size_t bytes);
+	virtual void progress(const Progress & value) { progressImpl(value); }
+	void progressImpl(const Progress & value);
 
 
 	/** Установить указатель на элемент списка процессов.

dbms/include/DB/DataStreams/RemoteBlockInputStream.h

@@ -66,7 +66,7 @@ public:
 	/** Отменяем умолчальное уведомление о прогрессе,
 	  * так как колбэк прогресса вызывается самостоятельно.
 	  */
-	void progress(size_t rows, size_t bytes) {}
+	void progress(const Progress & value) override {}
 
 
 	void cancel()
@@ -156,7 +156,7 @@ protected:
 					  *  ограничений (например, минимальная скорость выполнения запроса)
 					  *  и квот (например, на количество строчек для чтения).
 					  */
-					progressImpl(packet.progress.rows, packet.progress.bytes);
+					progressImpl(packet.progress);
 
 					if (!was_cancelled && !finished && isCancelled())
 						cancel();

dbms/include/DB/DataTypes/DataTypeString.h

@@ -16,6 +16,8 @@ using Poco::SharedPtr;
 class DataTypeString : public IDataType
 {
 public:
+	using FieldType = String;
+
 	std::string getName() const
 	{
 		return "String";
@@ -39,7 +41,7 @@ public:
 
 	void serializeTextQuoted(const Field & field, WriteBuffer & ostr) const;
 	void deserializeTextQuoted(Field & field, ReadBuffer & istr) const;
-	
+
 	void serializeTextJSON(const Field & field, WriteBuffer & ostr) const;
 
 	ColumnPtr createColumn() const;

dbms/include/DB/Functions/FunctionsArray.h

@@ -2,6 +2,10 @@
 
 #include <DB/DataTypes/DataTypeArray.h>
 #include <DB/DataTypes/DataTypesNumberFixed.h>
+#include <DB/DataTypes/DataTypeDate.h>
+#include <DB/DataTypes/DataTypeDateTime.h>
+#include <DB/DataTypes/DataTypeString.h>
+
 
 #include <DB/Columns/ColumnArray.h>
 #include <DB/Columns/ColumnString.h>
@@ -1190,6 +1194,43 @@ private:
 };
 
 
+template <typename Type> struct TypeToColumnType { using ColumnType = ColumnVector<Type>; };
+template <> struct TypeToColumnType<String> { using ColumnType = ColumnString; };
+
+template <typename DataType> struct DataTypeToName : TypeName<typename DataType::FieldType> { };
+template <> struct DataTypeToName<DataTypeDate> { static std::string get() { return "Date"; } };
+template <> struct DataTypeToName<DataTypeDateTime> { static std::string get() { return "DateTime"; } };
+
+template <typename DataType>
+struct EmptyArray : public IFunction
+{
+	String getName() const
+	{
+		return "emptyArray" + DataTypeToName<DataType>::get();
+	}
+
+	DataTypePtr getReturnType(const DataTypes & arguments) const
+	{
+		if (arguments.size() != 0)
+			throw Exception("Number of arguments for function " + getName() + " doesn't match: passed "
+				+ toString(arguments.size()) + ", should be 0.",
+				ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+		return new DataTypeArray{new DataType{}};
+	}
+
+	void execute(Block & block, const ColumnNumbers & arguments, size_t result)
+	{
+		using UnderlyingColumnType = typename TypeToColumnType<typename DataType::FieldType>::ColumnType;
+
+		block.getByPosition(result).column = new ColumnArray{
+			new UnderlyingColumnType,
+			new ColumnArray::ColumnOffsets_t{block.rowsInFirstColumn(), 0}
+		};
+	}
+};
+
+
 struct NameHas			{ static const char * get() { return "has"; } };
 struct NameIndexOf		{ static const char * get() { return "indexOf"; } };
 struct NameCountEqual	{ static const char * get() { return "countEqual"; } };
@@ -1198,5 +1239,19 @@ typedef FunctionArrayIndex<IndexToOne, 		NameHas>	FunctionHas;
 typedef FunctionArrayIndex<IndexIdentity, 	NameIndexOf>	FunctionIndexOf;
 typedef FunctionArrayIndex<IndexCount, 	NameCountEqual>	FunctionCountEqual;
 
+using FunctionEmptyArrayUInt8 = EmptyArray<DataTypeUInt8>;
+using FunctionEmptyArrayUInt16 = EmptyArray<DataTypeUInt16>;
+using FunctionEmptyArrayUInt32 = EmptyArray<DataTypeUInt32>;
+using FunctionEmptyArrayUInt64 = EmptyArray<DataTypeUInt64>;
+using FunctionEmptyArrayInt8 = EmptyArray<DataTypeInt8>;
+using FunctionEmptyArrayInt16 = EmptyArray<DataTypeInt16>;
+using FunctionEmptyArrayInt32 = EmptyArray<DataTypeInt32>;
+using FunctionEmptyArrayInt64 = EmptyArray<DataTypeInt64>;
+using FunctionEmptyArrayFloat32 = EmptyArray<DataTypeFloat32>;
+using FunctionEmptyArrayFloat64 = EmptyArray<DataTypeFloat64>;
+using FunctionEmptyArrayDate = EmptyArray<DataTypeDate>;
+using FunctionEmptyArrayDateTime = EmptyArray<DataTypeDateTime>;
+using FunctionEmptyArrayString = EmptyArray<DataTypeString>;
+
 
 }

dbms/include/DB/Functions/FunctionsCoding.h

@@ -13,6 +13,10 @@
 #include <DB/Columns/ColumnConst.h>
 #include <DB/Functions/IFunction.h>
 
+#include <arpa/inet.h>
+#include <statdaemons/ext/range.hpp>
+#include <array>
+
 
 namespace DB
 {
@@ -35,6 +39,426 @@ namespace DB
 /// Включая нулевой символ в конце.
 #define MAX_UINT_HEX_LENGTH 20
 
+const auto ipv4_bytes_length = 4;
+const auto ipv6_bytes_length = 16;
+
+class FunctionIPv6NumToString : public IFunction
+{
+public:
+	String getName() const { return "IPv6NumToString"; }
+
+	DataTypePtr getReturnType(const DataTypes & arguments) const
+	{
+		if (arguments.size() != 1)
+			throw Exception("Number of arguments for function " + getName() + " doesn't match: passed "
+			+ toString(arguments.size()) + ", should be 1.",
+			ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+		const auto ptr = typeid_cast<const DataTypeFixedString *>(arguments[0].get());
+		if (!ptr || ptr->getN() != ipv6_bytes_length)
+			throw Exception("Illegal type " + arguments[0]->getName() +
+							" of argument of function " + getName() +
+							", expected FixedString(" + toString(ipv6_bytes_length) + ")",
+							ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+
+		return new DataTypeString;
+	}
+
+	/// integer logarithm, return ceil(log(value, base)) (the smallest integer greater or equal  than log(value, base)
+	static constexpr uint32_t int_log(const uint32_t value, const uint32_t base, const bool carry = false)
+	{
+		return value >= base ? 1 + int_log(value / base, base, value % base || carry) : value % base > 1 || carry;
+	}
+
+	/// mapping of digits up to base 16
+	static constexpr auto && digits = "0123456789abcdef";
+
+	/// print integer in desired base, faster than sprintf
+	template <uint32_t base, typename T, uint32_t buffer_size = sizeof(T) * int_log(256, base)>
+	static void print_integer(char *& out, T value)
+	{
+		if (value == 0)
+			*out++ = '0';
+		else
+		{
+			char buf[buffer_size];
+			auto ptr = buf;
+
+			while (value > 0)
+			{
+				*ptr++ = digits[value % base];
+				value /= base;
+			}
+
+			while (ptr != buf)
+				*out++ = *--ptr;
+		}
+	}
+
+	/// print IPv4 address as %u.%u.%u.%u
+	static void ipv4_format(const unsigned char * src, char *& dst)
+	{
+		constexpr auto size = sizeof(UInt32);
+
+		for (const auto i : ext::range(0, size))
+		{
+			print_integer<10, UInt8>(dst, src[i]);
+
+			if (i != size - 1)
+				*dst++ = '.';
+		}
+	}
+
+	/** rewritten inet_ntop6 from http://svn.apache.org/repos/asf/apr/apr/trunk/network_io/unix/inet_pton.c
+	 *	performs significantly faster than the reference implementation due to the absence of sprintf calls,
+	 *	bounds checking, unnecessary string copying and length calculation */
+	static const void ipv6_format(const unsigned char * src, char *& dst)
+	{
+		struct { int base, len; } best{-1}, cur{-1};
+		std::array<uint16_t, ipv6_bytes_length / sizeof(uint16_t)> words{};
+
+		/** Preprocess:
+		 *	Copy the input (bytewise) array into a wordwise array.
+		 *	Find the longest run of 0x00's in src[] for :: shorthanding. */
+		for (const auto i : ext::range(0, ipv6_bytes_length))
+			words[i / 2] |= src[i] << ((1 - (i % 2)) << 3);
+
+		for (const auto i : ext::range(0, words.size()))
+		{
+			if (words[i] == 0) {
+				if (cur.base == -1)
+					cur.base = i, cur.len = 1;
+				else
+					cur.len++;
+			}
+			else
+			{
+				if (cur.base != -1)
+				{
+					if (best.base == -1 || cur.len > best.len)
+						best = cur;
+					cur.base = -1;
+				}
+			}
+		}
+
+		if (cur.base != -1)
+		{
+			if (best.base == -1 || cur.len > best.len)
+				best = cur;
+		}
+
+		if (best.base != -1 && best.len < 2)
+			best.base = -1;
+
+		/// Format the result.
+		for (const int i : ext::range(0, words.size()))
+		{
+			/// Are we inside the best run of 0x00's?
+			if (best.base != -1 && i >= best.base && i < (best.base + best.len))
+			{
+				if (i == best.base)
+					*dst++ = ':';
+				continue;
+			}
+
+			/// Are we following an initial run of 0x00s or any real hex?
+			if (i != 0)
+				*dst++ = ':';
+
+			/// Is this address an encapsulated IPv4?
+			if (i == 6 && best.base == 0 && (best.len == 6 || (best.len == 5 && words[5] == 0xffffu)))
+			{
+				ipv4_format(src + 12, dst);
+				break;
+			}
+
+			print_integer<16>(dst, words[i]);
+		}
+
+		/// Was it a trailing run of 0x00's?
+		if (best.base != -1 && (best.base + best.len) == words.size())
+			*dst++ = ':';
+
+		*dst++ = '\0';
+	}
+
+	void execute(Block & block, const ColumnNumbers & arguments, const size_t result)
+	{
+		const auto & col_name_type = block.getByPosition(arguments[0]);
+		const ColumnPtr & column = col_name_type.column;
+
+		if (const auto col_in = typeid_cast<const ColumnFixedString *>(column.get()))
+		{
+			if (col_in->getN() != ipv6_bytes_length)
+				throw Exception("Illegal type " + col_name_type.type->getName() +
+								" of column " + col_in->getName() +
+								" argument of function " + getName() +
+								", expected FixedString(" + toString(ipv6_bytes_length) + ")",
+								ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+
+			const auto size = col_in->size();
+			const auto & vec_in = col_in->getChars();
+
+			auto col_res = new ColumnString;
+			block.getByPosition(result).column = col_res;
+
+			ColumnString::Chars_t & vec_res = col_res->getChars();
+			ColumnString::Offsets_t & offsets_res = col_res->getOffsets();
+			vec_res.resize(size * INET6_ADDRSTRLEN);
+			offsets_res.resize(size);
+
+			auto begin = reinterpret_cast<char *>(&vec_res[0]);
+			auto pos = begin;
+
+			for (size_t offset = 0, i = 0; offset < vec_in.size(); offset += ipv6_bytes_length, ++i)
+			{
+				ipv6_format(&vec_in[offset], pos);
+				offsets_res[i] = pos - begin;
+			}
+
+			vec_res.resize(pos - begin);
+		}
+		else if (const auto col_in = typeid_cast<const ColumnConst<String> *>(column.get()))
+		{
+			const auto data_type_fixed_string = typeid_cast<const DataTypeFixedString *>(col_in->getDataType().get());
+			if (!data_type_fixed_string || data_type_fixed_string->getN() != ipv6_bytes_length)
+				throw Exception("Illegal type " + col_name_type.type->getName() +
+								" of column " + col_in->getName() +
+								" argument of function " + getName() +
+								", expected FixedString(" + toString(ipv6_bytes_length) + ")",
+								ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+
+			const auto & data_in = col_in->getData();
+
+			char buf[INET6_ADDRSTRLEN];
+			char * dst = buf;
+			ipv6_format(reinterpret_cast<const unsigned char *>(data_in.data()), dst);
+
+			block.getByPosition(result).column = new ColumnConstString{col_in->size(), buf};
+		}
+		else
+			throw Exception("Illegal column " + block.getByPosition(arguments[0]).column->getName()
+			+ " of argument of function " + getName(),
+			ErrorCodes::ILLEGAL_COLUMN);
+	}
+};
+
+class FunctionIPv6StringToNum : public IFunction
+{
+public:
+	String getName() const { return "IPv6StringToNum"; }
+
+	DataTypePtr getReturnType(const DataTypes & arguments) const
+	{
+		if (arguments.size() != 1)
+			throw Exception("Number of arguments for function " + getName() + " doesn't match: passed "
+			+ toString(arguments.size()) + ", should be 1.",
+							ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+		if (!typeid_cast<const DataTypeString *>(&*arguments[0]))
+			throw Exception("Illegal type " + arguments[0]->getName() + " of argument of function " + getName(),
+			ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+
+		return new DataTypeFixedString{ipv6_bytes_length};
+	}
+
+
+	static bool isDigit(char c) { return c >= '0' && c <= '9'; }
+
+	static bool ipv4_scan(const char * src, unsigned char * dst)
+	{
+		constexpr auto size = sizeof(UInt32);
+		char bytes[size]{};
+
+		for (const auto i : ext::range(0, size))
+		{
+			UInt32 value = 0;
+			size_t len = 0;
+			while (isDigit(*src) && len <= 3)
+			{
+				value = value * 10 + (*src - '0');
+				++len;
+				++src;
+			}
+
+			if (len == 0 || value > 255 || (i < size - 1 && *src != '.'))
+			{
+				memset(dst, 0, size);
+				return false;
+			}
+			bytes[i] = value;
+			++src;
+		}
+
+		if (src[-1] != '\0')
+		{
+			memset(dst, 0, size);
+			return false;
+		}
+
+		memcpy(dst, bytes, sizeof(bytes));
+		return true;
+	}
+
+	/// slightly altered implementation from http://svn.apache.org/repos/asf/apr/apr/trunk/network_io/unix/inet_pton.c
+	static void ipv6_scan(const char *  src, unsigned char * dst)
+	{
+		const auto clear_dst = [dst] {
+			memset(dst, '\0', ipv6_bytes_length);
+		};
+
+		/// Leading :: requires some special handling.
+		if (*src == ':')
+			if (*++src != ':')
+				return clear_dst();
+
+		/// get integer value for a hexademical char digit, or -1
+		const auto number_by_char = [] (const char ch) {
+			if ('A' <= ch && ch <= 'F')
+				return 10 + ch - 'A';
+
+			if ('a' <= ch && ch <= 'f')
+				return 10 + ch - 'a';
+
+			if ('0' <= ch && ch <= '9')
+				return ch - '0';
+
+			return -1;
+		};
+
+		unsigned char tmp[ipv6_bytes_length]{};
+		auto tp = tmp;
+		auto endp = tp + ipv6_bytes_length;
+		auto curtok = src;
+		auto saw_xdigit = false;
+		uint16_t val{};
+		unsigned char * colonp = nullptr;
+
+		while (const auto ch = *src++)
+		{
+			const auto num = number_by_char(ch);
+
+			if (num != -1)
+			{
+				val <<= 4;
+				val |= num;
+				if (val > 0xffffu)
+					return clear_dst();
+
+				saw_xdigit = 1;
+				continue;
+			}
+
+			if (ch == ':')
+			{
+				curtok = src;
+				if (!saw_xdigit)
+				{
+					if (colonp)
+						return clear_dst();
+
+					colonp = tp;
+					continue;
+				}
+
+				if (tp + sizeof(uint16_t) > endp)
+					return clear_dst();
+
+				*tp++ = static_cast<unsigned char>((val >> 8) & 0xffu);
+				*tp++ = static_cast<unsigned char>(val & 0xffu);
+				saw_xdigit = false;
+				val = 0;
+				continue;
+			}
+
+			if (ch == '.' && (tp + ipv4_bytes_length) <= endp)
+			{
+				if (!ipv4_scan(curtok, tp))
+					return clear_dst();
+
+				tp += ipv4_bytes_length;
+				saw_xdigit = false;
+				break;	/* '\0' was seen by ipv4_scan(). */
+			}
+
+			return clear_dst();
+		}
+
+		if (saw_xdigit)
+		{
+			if (tp + sizeof(uint16_t) > endp)
+				return clear_dst();
+
+			*tp++ = static_cast<unsigned char>((val >> 8) & 0xffu);
+			*tp++ = static_cast<unsigned char>(val & 0xffu);
+		}
+
+		if (colonp)
+		{
+			/*
+			 * Since some memmove()'s erroneously fail to handle
+			 * overlapping regions, we'll do the shift by hand.
+			 */
+			const auto n = tp - colonp;
+
+			for (int i = 1; i <= n; i++)
+			{
+				endp[- i] = colonp[n - i];
+				colonp[n - i] = 0;
+			}
+			tp = endp;
+		}
+
+		if (tp != endp)
+			return clear_dst();
+
+		memcpy(dst, tmp, sizeof(tmp));
+	}
+
+	void execute(Block & block, const ColumnNumbers & arguments, size_t result)
+	{
+		const ColumnPtr & column = block.getByPosition(arguments[0]).column;
+
+		if (const auto col_in = typeid_cast<const ColumnString *>(&*column))
+		{
+		    const auto col_res = new ColumnFixedString{ipv6_bytes_length};
+			block.getByPosition(result).column = col_res;
+
+			auto & vec_res = col_res->getChars();
+			vec_res.resize(col_in->size() * ipv6_bytes_length);
+
+			const ColumnString::Chars_t & vec_src = col_in->getChars();
+			const ColumnString::Offsets_t & offsets_src = col_in->getOffsets();
+			size_t src_offset = 0;
+
+			for (size_t out_offset = 0, i = 0;
+				 out_offset < vec_res.size();
+				 out_offset += ipv6_bytes_length, ++i)
+			{
+				ipv6_scan(reinterpret_cast<const char* >(&vec_src[src_offset]), &vec_res[out_offset]);
+				src_offset = offsets_src[i];
+			}
+		}
+		else if (const auto col_in = typeid_cast<const ColumnConstString *>(&*column))
+		{
+			String out(ipv6_bytes_length, 0);
+			ipv6_scan(col_in->getData().data(), reinterpret_cast<unsigned char *>(&out[0]));
+
+			block.getByPosition(result).column = new ColumnConst<String>{
+				col_in->size(),
+				out,
+				new DataTypeFixedString{ipv6_bytes_length}
+			};
+		}
+		else
+			throw Exception("Illegal column " + block.getByPosition(arguments[0]).column->getName()
+			+ " of argument of function " + getName(),
+							ErrorCodes::ILLEGAL_COLUMN);
+	}
+};
+
+
 class FunctionIPv4NumToString : public IFunction
 {
 public:
@@ -108,7 +532,7 @@ public:
 			ColumnString::Chars_t & vec_res = col_res->getChars();
 			ColumnString::Offsets_t & offsets_res = col_res->getOffsets();
 
-			vec_res.resize(vec_in.size() * 16); /// самое длинное значение: 255.255.255.255\0
+			vec_res.resize(vec_in.size() * INET_ADDRSTRLEN); /// самое длинное значение: 255.255.255.255\0
 			offsets_res.resize(vec_in.size());
 			char * begin = reinterpret_cast<char *>(&vec_res[0]);
 			char * pos = begin;
@@ -161,7 +585,7 @@ public:
 		return new DataTypeUInt32;
 	}
 
-	static inline bool isDigit(char c)
+	static bool isDigit(char c)
 	{
 		return c >= '0' && c <= '9';
 	}

dbms/include/DB/Functions/FunctionsConversion.h

@@ -453,7 +453,11 @@ public:
 			if (column_const->getData().size() > n)
 				throw Exception("String too long for type FixedString(" + toString(n) + ")",
 					ErrorCodes::TOO_LARGE_STRING_SIZE);
-			block.getByPosition(result).column = new ColumnConst<String>(column_const->size(), column_const->getData(), new DataTypeFixedString(n));
+
+			auto resized_string = column_const->getData();
+			resized_string.resize(n);
+
+			block.getByPosition(result).column = new ColumnConst<String>(column_const->size(), std::move(resized_string), new DataTypeFixedString(n));
 		}
 		else if(const ColumnString * column_string = typeid_cast<const ColumnString *>(&*column))
 		{

dbms/include/DB/Functions/FunctionsHashing.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <openssl/md5.h>
+#include <openssl/sha.h>
 #include <city.h>
 
 #include <Poco/ByteOrder.h>
@@ -64,6 +65,62 @@ struct HalfMD5Impl
 	}
 };
 
+struct MD5Impl
+{
+	static constexpr auto name = "MD5";
+	static constexpr auto length = 16;
+
+	static void apply(const char * begin, const size_t size, unsigned char * out_char_data)
+	{
+		MD5_CTX ctx;
+		MD5_Init(&ctx);
+		MD5_Update(&ctx, reinterpret_cast<const unsigned char *>(begin), size);
+		MD5_Final(out_char_data, &ctx);
+	}
+};
+
+struct SHA1Impl
+{
+	static constexpr auto name = "SHA1";
+	static constexpr auto length = 20;
+
+	static void apply(const char * begin, const size_t size, unsigned char * out_char_data)
+	{
+		SHA_CTX ctx;
+		SHA1_Init(&ctx);
+		SHA1_Update(&ctx, reinterpret_cast<const unsigned char *>(begin), size);
+		SHA1_Final(out_char_data, &ctx);
+	}
+};
+
+struct SHA224Impl
+{
+	static constexpr auto name = "SHA224";
+	static constexpr auto length = 28;
+
+	static void apply(const char * begin, const size_t size, unsigned char * out_char_data)
+	{
+		SHA256_CTX ctx;
+		SHA224_Init(&ctx);
+		SHA224_Update(&ctx, reinterpret_cast<const unsigned char *>(begin), size);
+		SHA224_Final(out_char_data, &ctx);
+	}
+};
+
+struct SHA256Impl
+{
+	static constexpr auto name = "SHA256";
+	static constexpr auto length = 32;
+
+	static void apply(const char * begin, const size_t size, unsigned char * out_char_data)
+	{
+		SHA256_CTX ctx;
+		SHA256_Init(&ctx);
+		SHA256_Update(&ctx, reinterpret_cast<const unsigned char *>(begin), size);
+		SHA256_Final(out_char_data, &ctx);
+	}
+};
+
 struct SipHash64Impl
 {
 	static UInt64 apply(const char * begin, size_t size)
@@ -72,6 +129,17 @@ struct SipHash64Impl
 	}
 };
 
+struct SipHash128Impl
+{
+	static constexpr auto name = "SipHash128";
+	static constexpr auto length = 16;
+
+	static void apply(const char * begin, const size_t size, unsigned char * out_char_data)
+	{
+		sipHash128(begin, size, reinterpret_cast<char*>(out_char_data));
+	}
+};
+
 struct IntHash32Impl
 {
 	typedef UInt32 ReturnType;
@@ -152,6 +220,72 @@ public:
 };
 
 
+template <typename Impl>
+class FunctionStringHashFixedString : public IFunction
+{
+public:
+	/// Получить имя функции.
+	String getName() const
+	{
+		return Impl::name;
+	}
+
+	/// Получить тип результата по типам аргументов. Если функция неприменима для данных аргументов - кинуть исключение.
+	DataTypePtr getReturnType(const DataTypes & arguments) const
+	{
+		if (arguments.size() != 1)
+			throw Exception("Number of arguments for function " + getName() + " doesn't match: passed "
+				+ toString(arguments.size()) + ", should be 1.",
+				ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+		if (!typeid_cast<const DataTypeString *>(&*arguments[0]))
+			throw Exception("Illegal type " + arguments[0]->getName() + " of argument of function " + getName(),
+				ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+
+		return new DataTypeFixedString{Impl::length};
+	}
+
+	/// Выполнить функцию над блоком.
+	void execute(Block & block, const ColumnNumbers & arguments, size_t result)
+	{
+		if (const ColumnString * col_from = typeid_cast<const ColumnString *>(&*block.getByPosition(arguments[0]).column))
+		{
+			auto col_to = new ColumnFixedString{Impl::length};
+			block.getByPosition(result).column = col_to;
+
+			const typename ColumnString::Chars_t & data = col_from->getChars();
+			const typename ColumnString::Offsets_t & offsets = col_from->getOffsets();
+			auto & chars_to = col_to->getChars();
+			const auto size = offsets.size();
+			chars_to.resize(size * Impl::length);
+
+			for (size_t i = 0; i < size; ++i)
+				Impl::apply(
+					reinterpret_cast<const char *>(&data[i == 0 ? 0 : offsets[i - 1]]),
+					i == 0 ? offsets[i] - 1 : (offsets[i] - 1 - offsets[i - 1]),
+					&chars_to[i * Impl::length]);
+		}
+		else if (const ColumnConstString * col_from = typeid_cast<const ColumnConstString *>(&*block.getByPosition(arguments[0]).column))
+		{
+			const auto & data = col_from->getData();
+
+			String hash(Impl::length, 0);
+			Impl::apply(data.data(), data.size(), reinterpret_cast<unsigned char *>(&hash[0]));
+
+			block.getByPosition(result).column = new ColumnConst<String>{
+				col_from->size(),
+				hash,
+				new DataTypeFixedString{Impl::length}
+			};
+		}
+		else
+			throw Exception("Illegal column " + block.getByPosition(arguments[0]).column->getName()
+					+ " of first argument of function " + getName(),
+				ErrorCodes::ILLEGAL_COLUMN);
+	}
+};
+
+
 template <typename Impl, typename Name>
 class FunctionIntHash : public IFunction
 {
@@ -465,6 +599,10 @@ typedef FunctionStringHash64<HalfMD5Impl,		NameHalfMD5> 		FunctionHalfMD5;
 typedef FunctionStringHash64<SipHash64Impl,		NameSipHash64> 		FunctionSipHash64;
 typedef FunctionIntHash<IntHash32Impl,			NameIntHash32> 		FunctionIntHash32;
 typedef FunctionIntHash<IntHash64Impl,			NameIntHash64> 		FunctionIntHash64;
-
+typedef FunctionStringHashFixedString<MD5Impl>						FunctionMD5;
+typedef FunctionStringHashFixedString<SHA1Impl>						FunctionSHA1;
+typedef FunctionStringHashFixedString<SHA224Impl>					FunctionSHA224;
+typedef FunctionStringHashFixedString<SHA256Impl>					FunctionSHA256;
+typedef FunctionStringHashFixedString<SipHash128Impl>				FunctionSipHash128;
 
 }

dbms/include/DB/Functions/FunctionsHigherOrder.h

@@ -288,6 +288,95 @@ struct ArraySumImpl
 	}
 };
 
+struct ArrayFirstImpl
+{
+	static bool needBoolean() { return false; }
+	static bool needExpression() { return true; }
+	static bool needOneArray() { return false; }
+
+	static DataTypePtr getReturnType(const DataTypePtr & expression_return, const DataTypePtr & array_element)
+	{
+		return array_element;
+	}
+
+	static ColumnPtr execute(const ColumnArray * array, ColumnPtr mapped)
+	{
+		auto column_filter = typeid_cast<ColumnVector<UInt8> *>(&*mapped);
+		if (!column_filter)
+			throw Exception("Unexpected type of filter column", ErrorCodes::ILLEGAL_COLUMN);
+
+		const auto & filter = column_filter->getData();
+		const auto & offsets = array->getOffsets();
+		const auto & data = array->getData();
+		ColumnPtr out{data.cloneEmpty()};
+
+		size_t pos{};
+		for (size_t i = 0; i < offsets.size(); ++i)
+		{
+			auto exists = false;
+			for (; pos < offsets[i]; ++pos)
+			{
+				if (filter[pos])
+				{
+					out->insert(data[pos]);
+					exists = true;
+					pos = offsets[i];
+					break;
+				}
+			}
+
+			if (!exists)
+				out->insertDefault();
+		}
+
+		return out;
+	}
+};
+
+struct ArrayFirstIndexImpl
+{
+	static bool needBoolean() { return false; }
+	static bool needExpression() { return true; }
+	static bool needOneArray() { return false; }
+
+	static DataTypePtr getReturnType(const DataTypePtr & expression_return, const DataTypePtr & array_element)
+	{
+		return new DataTypeUInt32;
+	}
+
+	static ColumnPtr execute(const ColumnArray * array, ColumnPtr mapped)
+	{
+		auto column_filter = typeid_cast<ColumnVector<UInt8> *>(&*mapped);
+		if (!column_filter)
+			throw Exception("Unexpected type of filter column", ErrorCodes::ILLEGAL_COLUMN);
+
+		const auto & filter = column_filter->getData();
+		const auto & offsets = array->getOffsets();
+		auto out_column = new ColumnVector<UInt32>{offsets.size()};
+		ColumnPtr out_column_ptr{out_column};
+		auto & out_index = out_column->getData();
+
+		size_t pos{};
+		for (size_t i = 0; i < offsets.size(); ++i)
+		{
+			UInt32 index{};
+			for (size_t idx{1}; pos < offsets[i]; ++pos, ++idx)
+			{
+				if (filter[pos])
+				{
+					index = idx;
+					pos = offsets[i];
+					break;
+				}
+			}
+
+			out_index[i] = index;
+		}
+
+		return out_column_ptr;
+	}
+};
+
 template <typename Impl, typename Name>
 class FunctionArrayMapped : public IFunction
 {
@@ -508,6 +597,8 @@ struct NameArrayCount	{ static const char * get() { return "arrayCount"; } };
 struct NameArrayExists	{ static const char * get() { return "arrayExists"; } };
 struct NameArrayAll		{ static const char * get() { return "arrayAll"; } };
 struct NameArraySum		{ static const char * get() { return "arraySum"; } };
+struct NameArrayFirst	{ static const char * get() { return "arrayFirst"; } };
+struct NameArrayFirstIndex	{ static const char * get() { return "arrayFirstIndex"; } };
 
 typedef FunctionArrayMapped<ArrayMapImpl, 		NameArrayMap>		FunctionArrayMap;
 typedef FunctionArrayMapped<ArrayFilterImpl, 	NameArrayFilter>	FunctionArrayFilter;
@@ -515,5 +606,7 @@ typedef FunctionArrayMapped<ArrayCountImpl, 	NameArrayCount>		FunctionArrayCount
 typedef FunctionArrayMapped<ArrayExistsImpl, 	NameArrayExists>	FunctionArrayExists;
 typedef FunctionArrayMapped<ArrayAllImpl,	 	NameArrayAll>		FunctionArrayAll;
 typedef FunctionArrayMapped<ArraySumImpl,	 	NameArraySum>		FunctionArraySum;
+typedef FunctionArrayMapped<ArrayFirstImpl,	 	NameArrayFirst>		FunctionArrayFirst;
+typedef FunctionArrayMapped<ArrayFirstIndexImpl,	 	NameArrayFirstIndex>		FunctionArrayFirstIndex;
 
 }

dbms/include/DB/Functions/FunctionsMiscellaneous.h

@@ -20,6 +20,7 @@
 #include <DB/Columns/ColumnTuple.h>
 #include <DB/Columns/ColumnArray.h>
 #include <DB/Columns/ColumnReplicated.h>
+#include <DB/Common/UnicodeBar.h>
 #include <DB/Functions/IFunction.h>
 
 
@@ -672,47 +673,6 @@ private:
 		return apply_visitor(FieldVisitorConvertToNumber<T>(), column[0]);
 	}
 
-	static constexpr size_t BAR_CHAR_SIZE = strlen("█");
-
-	template <typename T>
-	static Float64 barWidth(T x, Int64 min, Int64 max, Float64 max_width)
-	{
-		if (x <= min)
-			return 0;
-
-		if (x >= max)
-			return max_width;
-
-		return (x - min) * max_width / (max - min);
-	}
-
-	static size_t barWidthInBytes(Float64 width)
-	{
-		return ceil(width - 1.0 / 8) * BAR_CHAR_SIZE;
-	}
-
-	/// В dst должно быть место для barWidthInBytes(width) символов и завершающего нуля.
-	static void renderBar(Float64 width, char * dst)
-	{
-		size_t floor_width = floor(width);
-
-		for (size_t i = 0; i < floor_width; ++i)
-		{
-			memcpy(dst, "█", BAR_CHAR_SIZE);
-			dst += BAR_CHAR_SIZE;
-		}
-
-		size_t remainder = floor((width - floor_width) * 8);
-
-		if (remainder)
-		{
-			memcpy(dst, &"▏▎▍▌▋▋▊▉"[(remainder - 1) * BAR_CHAR_SIZE], BAR_CHAR_SIZE);
-			dst += BAR_CHAR_SIZE;
-		}
-
-		*dst = 0;
-	}
-
 	template <typename T>
 	static void fill(const PODArray<T> & src, ColumnString::Chars_t & dst_chars, ColumnString::Offsets_t & dst_offsets,
 		Int64 min, Int64 max, Float64 max_width)
@@ -721,14 +681,14 @@ private:
 		size_t current_offset = 0;
 
 		dst_offsets.resize(size);
-		dst_chars.reserve(size * (barWidthInBytes(max_width) + 1));	/// строки 0-terminated.
+		dst_chars.reserve(size * (UnicodeBar::getWidthInBytes(max_width) + 1));	/// строки 0-terminated.
 
 		for (size_t i = 0; i < size; ++i)
 		{
-			Float64 width = barWidth(src[i], min, max, max_width);
-			size_t next_size = current_offset + barWidthInBytes(width) + 1;
+			Float64 width = UnicodeBar::getWidth(src[i], min, max, max_width);
+			size_t next_size = current_offset + UnicodeBar::getWidthInBytes(width) + 1;
 			dst_chars.resize(next_size);
-			renderBar(width, reinterpret_cast<char *>(&dst_chars[current_offset]));
+			UnicodeBar::render(width, reinterpret_cast<char *>(&dst_chars[current_offset]));
 			current_offset = next_size;
 			dst_offsets[i] = current_offset;
 		}
@@ -738,9 +698,9 @@ private:
 	static void fill(T src, String & dst_chars,
 		Int64 min, Int64 max, Float64 max_width)
 	{
-		Float64 width = barWidth(src, min, max, max_width);
-		dst_chars.resize(barWidthInBytes(width));
-		renderBar(width, &dst_chars[0]);
+		Float64 width = UnicodeBar::getWidth(src, min, max, max_width);
+		dst_chars.resize(UnicodeBar::getWidthInBytes(width));
+		UnicodeBar::render(width, &dst_chars[0]);
 	}
 
 	template <typename T>

dbms/include/DB/Functions/FunctionsURL.h

@@ -113,6 +113,94 @@ struct ExtractDomain
 	}
 };
 
+struct ExtractFirstSignificantSubdomain
+{
+	static size_t getReserveLengthForElement() { return 10; }
+
+	static void execute(const Pos data, const size_t size, Pos & res_data, size_t & res_size, Pos * out_domain_end = nullptr)
+	{
+		res_data = data;
+		res_size = 0;
+
+		Pos tmp;
+		size_t domain_length;
+		ExtractDomain<true>::execute(data, size, tmp, domain_length);
+
+		if (domain_length == 0)
+			return;
+
+		if (out_domain_end)
+			*out_domain_end = tmp + domain_length;
+
+		/// cut useless dot
+		if (tmp[domain_length - 1] == '.')
+			--domain_length;
+
+		res_data = tmp;
+		res_size = domain_length;
+
+		auto begin = tmp;
+		auto end = begin + domain_length;
+		const char * last_3_periods[3]{};
+		auto pos = static_cast<const char *>(memchr(begin, '.', domain_length));
+
+		while (pos)
+		{
+			last_3_periods[2] = last_3_periods[1];
+			last_3_periods[1] = last_3_periods[0];
+			last_3_periods[0] = pos;
+			pos = static_cast<const char *>(memchr(pos + 1, '.', end - pos - 1));
+		}
+
+		if (!last_3_periods[0])
+			return;
+
+		if (!last_3_periods[1])
+		{
+			res_size = last_3_periods[0] - begin;
+			return;
+		}
+
+		if (!last_3_periods[2])
+			last_3_periods[2] = begin - 1;
+
+		if (!strncmp(last_3_periods[1] + 1, "com", 3) ||
+			!strncmp(last_3_periods[1] + 1, "net", 3) ||
+			!strncmp(last_3_periods[1] + 1, "org", 3) ||
+			!strncmp(last_3_periods[1] + 1, "co", 2))
+		{
+			res_data += last_3_periods[2] + 1 - begin;
+			res_size = last_3_periods[1] - last_3_periods[2] - 1;
+			return;
+		}
+
+		res_data += last_3_periods[1] + 1 - begin;
+		res_size = last_3_periods[0] - last_3_periods[1] - 1;
+	}
+};
+
+struct CutToFirstSignificantSubdomain
+{
+	static size_t getReserveLengthForElement() { return 15; }
+
+	static void execute(const Pos data, const size_t size, Pos & res_data, size_t & res_size)
+	{
+		res_data = data;
+		res_size = 0;
+
+		Pos tmp_data;
+		size_t tmp_length;
+		Pos domain_end;
+		ExtractFirstSignificantSubdomain::execute(data, size, tmp_data, tmp_length, &domain_end);
+
+		if (tmp_length == 0)
+			return;
+
+		res_data = tmp_data;
+		res_size = domain_end - tmp_data;
+	}
+};
+
 struct ExtractTopLevelDomain
 {
 	static size_t getReserveLengthForElement() { return 5; }
@@ -839,12 +927,15 @@ struct CutSubstringImpl
 struct NameProtocol 					{ static const char * get() { return "protocol"; } };
 struct NameDomain 						{ static const char * get() { return "domain"; } };
 struct NameDomainWithoutWWW 			{ static const char * get() { return "domainWithoutWWW"; } };
+struct NameFirstSignificantSubdomain	{ static const char * get() { return "firstSignificantSubdomain"; } };
 struct NameTopLevelDomain 				{ static const char * get() { return "topLevelDomain"; } };
 struct NamePath 						{ static const char * get() { return "path"; } };
 struct NameQueryString					{ static const char * get() { return "queryString"; } };
 struct NameFragment 					{ static const char * get() { return "fragment"; } };
 struct NameQueryStringAndFragment		{ static const char * get() { return "queryStringAndFragment"; } };
 
+struct NameCutToFirstSignificantSubdomain { static const char * get() { return "cutToFirstSignificantSubdomain"; } };
+
 struct NameCutWWW 						{ static const char * get() { return "cutWWW"; } };
 struct NameCutQueryString				{ static const char * get() { return "cutQueryString"; } };
 struct NameCutFragment 					{ static const char * get() { return "cutFragment"; } };
@@ -856,12 +947,15 @@ struct NameCutURLParameter 				{ static const char * get() { return "cutURLParam
 typedef FunctionStringToString<ExtractSubstringImpl<ExtractProtocol>, 			NameProtocol>	 		FunctionProtocol;
 typedef FunctionStringToString<ExtractSubstringImpl<ExtractDomain<false> >, 		NameDomain>	 			FunctionDomain;
 typedef FunctionStringToString<ExtractSubstringImpl<ExtractDomain<true>  >, 		NameDomainWithoutWWW>	FunctionDomainWithoutWWW;
+typedef FunctionStringToString<ExtractSubstringImpl<ExtractFirstSignificantSubdomain>, NameFirstSignificantSubdomain>	FunctionFirstSignificantSubdomain;
 typedef FunctionStringToString<ExtractSubstringImpl<ExtractTopLevelDomain>, 		NameTopLevelDomain>		FunctionTopLevelDomain;
 typedef FunctionStringToString<ExtractSubstringImpl<ExtractPath>, 				NamePath>				FunctionPath;
 typedef FunctionStringToString<ExtractSubstringImpl<ExtractQueryString<true> >, 	NameQueryString>		FunctionQueryString;
 typedef FunctionStringToString<ExtractSubstringImpl<ExtractFragment<true> >, 		NameFragment>			FunctionFragment;
 typedef FunctionStringToString<ExtractSubstringImpl<ExtractQueryStringAndFragment<true> >, NameQueryStringAndFragment>	FunctionQueryStringAndFragment;
 
+typedef FunctionStringToString<ExtractSubstringImpl<CutToFirstSignificantSubdomain>, NameCutToFirstSignificantSubdomain> FunctionCutToFirstSignificantSubdomain;
+
 typedef FunctionStringToString<CutSubstringImpl<ExtractWWW>, 						NameCutWWW>				FunctionCutWWW;
 typedef FunctionStringToString<CutSubstringImpl<ExtractQueryString<false> >, 		NameCutQueryString>		FunctionCutQueryString;
 typedef FunctionStringToString<CutSubstringImpl<ExtractFragment<false> >, 		NameCutFragment>		FunctionCutFragment;

dbms/include/DB/Interpreters/Aggregator.h

@@ -22,6 +22,7 @@
 #include <DB/Columns/ColumnString.h>
 #include <DB/Columns/ColumnFixedString.h>
 #include <DB/Columns/ColumnAggregateFunction.h>
+#include <DB/Columns/ColumnVector.h>
 
 
 
@@ -53,25 +54,78 @@ typedef HashMap<UInt128, AggregateDataPtr, UInt128Hash> AggregatedDataWithKeys12
 typedef HashMap<UInt128, std::pair<StringRef*, AggregateDataPtr>, UInt128TrivialHash> AggregatedDataHashed;
 
 
-/// Для случая, когда есть один числовой ключ.
-struct AggregationMethodKey64
+/// Специализации для UInt8, UInt16.
+struct TrivialHash
 {
-	typedef AggregatedDataWithUInt64Key Data;
-	typedef Data::key_type Key;
-	typedef Data::mapped_type Mapped;
-	typedef Data::iterator iterator;
-	typedef Data::const_iterator const_iterator;
+	template <typename T>
+	size_t operator() (T key) const
+	{
+		return key;
+	}
+};
+
+/// Превращает хэш-таблицу в что-то типа lookup-таблицы. Остаётся неоптимальность - в ячейках хранятся ключи.
+template <size_t key_bits>
+struct HashTableFixedGrower
+{
+	size_t bufSize() const				{ return 1 << key_bits; }
+	size_t place(size_t x) const 		{ return x; }
+	/// Тут можно было бы написать __builtin_unreachable(), но компилятор не до конца всё оптимизирует, и получается менее эффективно.
+	size_t next(size_t pos) const		{ return pos + 1; }
+	bool overflow(size_t elems) const	{ return false; }
+
+	void increaseSize() { __builtin_unreachable(); }
+	void set(size_t num_elems) {}
+	void setBufSize(size_t buf_size_) {}
+};
+
+typedef HashMap<UInt64, AggregateDataPtr, TrivialHash, HashTableFixedGrower<8>> AggregatedDataWithUInt8Key;
+typedef HashMap<UInt64, AggregateDataPtr, TrivialHash, HashTableFixedGrower<16>> AggregatedDataWithUInt16Key;
+
+template <typename FieldType>
+struct AggregatedDataWithUIntKey
+{
+	using Type = AggregatedDataWithUInt64Key;
+	static constexpr bool never_overflows = false;
+};
+
+template <>
+struct AggregatedDataWithUIntKey<UInt8>
+{
+	using Type = AggregatedDataWithUInt8Key;
+	static constexpr bool never_overflows = true;	/// Говорит о том, что в результате агрегации не может быть много записей.
+};
+
+template <>
+struct AggregatedDataWithUIntKey<UInt16>
+{
+	using Type = AggregatedDataWithUInt16Key;
+	static constexpr bool never_overflows = true;
+};
+
+
+/// Для случая, когда есть один числовой ключ.
+template <typename FieldType>	/// UInt8/16/32/64 для любых типов соответствующей битности.
+struct AggregationMethodOneNumber
+{
+	typedef typename AggregatedDataWithUIntKey<FieldType>::Type Data;
+	typedef typename Data::key_type Key;
+	typedef typename Data::mapped_type Mapped;
+	typedef typename Data::iterator iterator;
+	typedef typename Data::const_iterator const_iterator;
+
+	static constexpr bool never_overflows = AggregatedDataWithUIntKey<FieldType>::never_overflows;
 
 	Data data;
 
-	const IColumn * column;
+	const FieldType * column;
 
 	/** Вызывается в начале обработки каждого блока.
 	  * Устанавливает переменные, необходимые для остальных методов, вызываемых во внутренних циклах.
 	  */
 	void init(ConstColumnPlainPtrs & key_columns)
 	{
-		column = key_columns[0];
+		column = &static_cast<const ColumnVector<FieldType> *>(key_columns[0])->getData()[0];
 	}
 
 	/// Достать из ключевых столбцов ключ для вставки в хэш-таблицу.
@@ -82,7 +136,7 @@ struct AggregationMethodKey64
 		const Sizes & key_sizes,	/// Если ключи фиксированной длины - их длины. Не используется в методах агрегации по ключам переменной длины.
 		StringRefs & keys) const	/// Сюда могут быть записаны ссылки на данные ключей в столбцах. Они могут быть использованы в дальнейшем.
 	{
-		return column->get64(i);
+		return get64(column[i]);
 	}
 
 	/// Из значения в хэш-таблице получить AggregateDataPtr.
@@ -99,10 +153,43 @@ struct AggregationMethodKey64
 	  */
 	static void insertKeyIntoColumns(const_iterator & it, ColumnPlainPtrs & key_columns, size_t keys_size, const Sizes & key_sizes)
 	{
-		key_columns[0]->insertData(reinterpret_cast<const char *>(&it->first), sizeof(it->first));
+		static_cast<ColumnVector<FieldType> *>(key_columns[0])->insertData(reinterpret_cast<const char *>(&it->first), sizeof(it->first));
+	}
+
+private:
+	UInt64 get64(FieldType x) const
+	{
+		return x;
 	}
 };
 
+template <>
+inline UInt64 AggregationMethodOneNumber<Float64>::get64(Float64 x) const
+{
+	union
+	{
+		Float64 src;
+		UInt64 res;
+	};
+
+	src = x;
+	return res;
+}
+
+template <>
+inline UInt64 AggregationMethodOneNumber<Float32>::get64(Float32 x) const
+{
+	union
+	{
+		Float32 src;
+		UInt64 res;
+	};
+
+	res = 0;
+	src = x;
+	return res;
+}
+
 
 /// Для случая, когда есть один строковый ключ.
 struct AggregationMethodString
@@ -113,6 +200,8 @@ struct AggregationMethodString
 	typedef Data::iterator iterator;
 	typedef Data::const_iterator const_iterator;
 
+	static constexpr bool never_overflows = false;
+
 	Data data;
 
 	const ColumnString::Offsets_t * offsets;
@@ -160,6 +249,8 @@ struct AggregationMethodFixedString
 	typedef Data::iterator iterator;
 	typedef Data::const_iterator const_iterator;
 
+	static constexpr bool never_overflows = false;
+
 	Data data;
 
 	size_t n;
@@ -207,6 +298,8 @@ struct AggregationMethodKeys128
 	typedef Data::iterator iterator;
 	typedef Data::const_iterator const_iterator;
 
+	static constexpr bool never_overflows = false;
+
 	Data data;
 
 	void init(ConstColumnPlainPtrs & key_columns)
@@ -252,6 +345,8 @@ struct AggregationMethodHashed
 	typedef Data::iterator iterator;
 	typedef Data::const_iterator const_iterator;
 
+	static constexpr bool never_overflows = false;
+
 	Data data;
 
 	void init(ConstColumnPlainPtrs & key_columns)
@@ -317,7 +412,10 @@ struct AggregatedDataVariants : private boost::noncopyable
 	  */
 	AggregatedDataWithoutKey without_key = nullptr;
 
-	std::unique_ptr<AggregationMethodKey64> 		key64;
+	std::unique_ptr<AggregationMethodOneNumber<UInt8>>	key8;
+	std::unique_ptr<AggregationMethodOneNumber<UInt16>>	key16;
+	std::unique_ptr<AggregationMethodOneNumber<UInt32>>	key32;
+	std::unique_ptr<AggregationMethodOneNumber<UInt64>>	key64;
 	std::unique_ptr<AggregationMethodString> 		key_string;
 	std::unique_ptr<AggregationMethodFixedString> 	key_fixed_string;
 	std::unique_ptr<AggregationMethodKeys128> 		keys128;
@@ -325,13 +423,16 @@ struct AggregatedDataVariants : private boost::noncopyable
 
 	enum Type
 	{
-		EMPTY 				= 0,
-		WITHOUT_KEY 		= 1,
-		KEY_64				= 2,
-		KEY_STRING			= 3,
-		KEY_FIXED_STRING 	= 4,
-		KEYS_128			= 5,
-		HASHED				= 6,
+		EMPTY = 0,
+		WITHOUT_KEY,
+		KEY_8,
+		KEY_16,
+		KEY_32,
+		KEY_64,
+		KEY_STRING,
+		KEY_FIXED_STRING,
+		KEYS_128,
+		HASHED,
 	};
 	Type type = EMPTY;
 
@@ -348,11 +449,14 @@ struct AggregatedDataVariants : private boost::noncopyable
 		{
 			case EMPTY:				break;
 			case WITHOUT_KEY:		break;
-			case KEY_64:			key64			.reset(new AggregationMethodKey64); 		break;
-			case KEY_STRING:		key_string		.reset(new AggregationMethodString); 		break;
-			case KEY_FIXED_STRING:	key_fixed_string.reset(new AggregationMethodFixedString); 	break;
-			case KEYS_128:			keys128			.reset(new AggregationMethodKeys128); 		break;
-			case HASHED:			hashed			.reset(new AggregationMethodHashed);	 	break;
+			case KEY_8:				key8			.reset(new decltype(key8)::element_type); 			break;
+			case KEY_16:			key16			.reset(new decltype(key16)::element_type); 			break;
+			case KEY_32:			key32			.reset(new decltype(key32)::element_type); 			break;
+			case KEY_64:			key64			.reset(new decltype(key64)::element_type); 			break;
+			case KEY_STRING:		key_string		.reset(new decltype(key_string)::element_type); 	break;
+			case KEY_FIXED_STRING:	key_fixed_string.reset(new decltype(key_fixed_string)::element_type); break;
+			case KEYS_128:			keys128			.reset(new decltype(keys128)::element_type); 		break;
+			case HASHED:			hashed			.reset(new decltype(hashed)::element_type);	 		break;
 
 			default:
 				throw Exception("Unknown aggregated data variant.", ErrorCodes::UNKNOWN_AGGREGATED_DATA_VARIANT);
@@ -365,6 +469,9 @@ struct AggregatedDataVariants : private boost::noncopyable
 		{
 			case EMPTY:				return 0;
 			case WITHOUT_KEY:		return 1;
+			case KEY_8:				return key8->data.size() 				+ (without_key != nullptr);
+			case KEY_16:			return key16->data.size() 				+ (without_key != nullptr);
+			case KEY_32:			return key32->data.size() 				+ (without_key != nullptr);
 			case KEY_64:			return key64->data.size() 				+ (without_key != nullptr);
 			case KEY_STRING:		return key_string->data.size() 			+ (without_key != nullptr);
 			case KEY_FIXED_STRING:	return key_fixed_string->data.size() 	+ (without_key != nullptr);
@@ -382,6 +489,9 @@ struct AggregatedDataVariants : private boost::noncopyable
 		{
 			case EMPTY:				return "EMPTY";
 			case WITHOUT_KEY:		return "WITHOUT_KEY";
+			case KEY_8:				return "KEY_8";
+			case KEY_16:			return "KEY_16";
+			case KEY_32:			return "KEY_32";
 			case KEY_64:			return "KEY_64";
 			case KEY_STRING:		return "KEY_STRING";
 			case KEY_FIXED_STRING:	return "KEY_FIXED_STRING";

dbms/include/DB/Interpreters/ProcessList.h

@@ -8,6 +8,7 @@
 #include <Poco/Net/IPAddress.h>
 #include <statdaemons/Stopwatch.h>
 #include <DB/Core/Defines.h>
+#include <DB/Core/Progress.h>
 #include <DB/Core/Exception.h>
 #include <DB/Core/ErrorCodes.h>
 #include <DB/Common/MemoryTracker.h>
@@ -35,8 +36,7 @@ public:
 
 		Stopwatch watch;
 
-		volatile size_t rows_processed = 0;
-		volatile size_t bytes_processed = 0;
+		Progress progress;
 
 		MemoryTracker memory_tracker;
 
@@ -56,10 +56,9 @@ public:
 			current_memory_tracker = nullptr;
 		}
 
-		bool update(size_t rows, size_t bytes) volatile
+		bool update(const Progress & value)
 		{
-			__sync_add_and_fetch(&rows_processed, rows);
-			__sync_add_and_fetch(&bytes_processed, bytes);
+			progress.incrementPiecewiseAtomically(value);
 			return !is_cancelled;
 		}
 	};

dbms/include/DB/Interpreters/SplittingAggregator.h

@@ -94,6 +94,9 @@ private:
 	void calculateHashesThread(Block & block, size_t begin, size_t end, ExceptionPtr & exception, MemoryTracker * memory_tracker);
 	void aggregateThread(Block & block, AggregatedDataVariants & result, size_t thread_no, ExceptionPtr & exception, MemoryTracker * memory_tracker);
 	void convertToBlockThread(AggregatedDataVariants & data_variant, Block & block, bool final, ExceptionPtr & exception, MemoryTracker * memory_tracker);
+
+	template <typename FieldType>
+	void aggregateOneNumber(AggregatedDataVariants & result, size_t thread_no, bool no_more_keys);
 };
 
 

dbms/include/DB/Storages/MergeTree/MergeTreeBlockInputStream.h

@@ -41,10 +41,10 @@ public:
 			///  (от storage ожидают получить только столбцы таблицы).
 			remove_prewhere_column = !pre_name_set.count(prewhere_column);
 			Names post_column_names;
-			for (size_t i = 0; i < column_names.size(); ++i)
+			for (const auto & name : column_names)
 			{
-				if (!pre_name_set.count(column_names[i]))
-					post_column_names.push_back(column_names[i]);
+				if (!pre_name_set.count(name))
+					post_column_names.push_back(name);
 			}
 			column_names = post_column_names;
 		}
@@ -68,8 +68,16 @@ public:
 			columns = owned_data_part->columns.addTypes(column_names);
 		}
 
+		/// Оценим общее количество строк - для прогресс-бара.
+		for (const auto & range : all_mark_ranges)
+			total_rows += range.end - range.begin;
+		total_rows *= storage.index_granularity;
+
 		LOG_TRACE(log, "Reading " << all_mark_ranges.size() << " ranges from part " << owned_data_part->name
-			<< ", up to " << (all_mark_ranges.back().end - all_mark_ranges.front().begin) * storage.index_granularity
+			<< ", approx. " << total_rows
+			<< (all_mark_ranges.size() > 1
+				? ", up to " + toString((all_mark_ranges.back().end - all_mark_ranges.front().begin) * storage.index_granularity)
+				: "")
 			<< " rows starting from " << all_mark_ranges.front().begin * storage.index_granularity);
 	}
 
@@ -97,7 +105,7 @@ public:
 
 protected:
 	/// Будем вызывать progressImpl самостоятельно.
-	void progress(size_t rows, size_t bytes) {}
+	void progress(const Progress & value) override {}
 
 	Block readImpl()
 	{
@@ -108,6 +116,10 @@ protected:
 
 		if (!reader)
 		{
+			/// Отправим информацию о том, что собираемся читать примерно столько-то строк.
+			/// NOTE В конструкторе это делать не получилось бы, потому что тогда ещё не установлен progress_callback.
+			progressImpl(Progress(0, 0, total_rows));
+
 			UncompressedCache * uncompressed_cache = use_uncompressed_cache ? storage.context.getUncompressedCache() : NULL;
 			reader.reset(new MergeTreeReader(path, owned_data_part->name, columns, uncompressed_cache, storage, all_mark_ranges));
 			if (prewhere_actions)
@@ -135,7 +147,7 @@ protected:
 					if (range.begin == range.end)
 						remaining_mark_ranges.pop_back();
 				}
-				progressImpl(res.rows(), res.bytes());
+				progressImpl(Progress(res.rows(), res.bytes()));
 				pre_reader->fillMissingColumns(res);
 
 				/// Вычислим выражение в PREWHERE.
@@ -164,7 +176,7 @@ protected:
 						reader->readRange(range.begin, range.end, res);
 					}
 
-					progressImpl(0, res.bytes() - pre_bytes);
+					progressImpl(Progress(0, res.bytes() - pre_bytes));
 				}
 				else if (ColumnUInt8 * column_vec = typeid_cast<ColumnUInt8 *>(&*column))
 				{
@@ -216,7 +228,7 @@ protected:
 						continue;
 					}
 
-					progressImpl(0, res.bytes() - pre_bytes);
+					progressImpl(Progress(0, res.bytes() - pre_bytes));
 
 					post_filter.resize(post_filter_pos);
 
@@ -259,7 +271,7 @@ protected:
 					remaining_mark_ranges.pop_back();
 			}
 
-			progressImpl(res.rows(), res.bytes());
+			progressImpl(Progress(res.rows(), res.bytes()));
 
 			reader->fillMissingColumns(res);
 		}
@@ -297,6 +309,7 @@ private:
 	ExpressionActionsPtr prewhere_actions;
 	String prewhere_column;
 	bool remove_prewhere_column;
+	size_t total_rows = 0;	/// Приблизительное общее количество строк - для прогресс-бара.
 
 	Logger * log;
 };

dbms/include/DB/Storages/StorageBuffer.h

@@ -0,0 +1,128 @@
+#pragma once
+
+#include <mutex>
+#include <thread>
+
+#include <DB/Core/NamesAndTypes.h>
+#include <DB/Storages/IStorage.h>
+#include <DB/DataStreams/IProfilingBlockInputStream.h>
+#include <DB/DataStreams/IBlockOutputStream.h>
+#include <DB/Interpreters/Context.h>
+
+
+namespace DB
+{
+
+
+/** При вставке, буферизует данные в оперативке, пока не превышены некоторые пороги.
+  * Когда пороги превышены - сбрасывает данные в другую таблицу.
+  * При чтении, читает как из своих буферов, так и из подчинённой таблицы.
+  *
+  * Буфер представляет собой набор из num_shards блоков.
+  * При записи, выбирается номер блока по остатку от деления ThreadNumber на num_shards (или один из других),
+  *  и в соответствующий блок добавляются строчки.
+  * При использовании блока, он блокируется некоторым mutex-ом. Если при записи, соответствующий блок уже занят
+  *  - пробуем заблокировать следующий по кругу блок, и так не более num_shards раз (далее блокируемся).
+  * Пороги проверяются при вставке, а также, периодически, в фоновом потоке (чтобы реализовать пороги по времени).
+  * Пороги действуют независимо для каждого shard-а. Каждый shard может быть сброшен независимо от других.
+  * Если в таблицу вставляется блок, который сам по себе превышает max-пороги, то он записывается сразу в подчинённую таблицу без буферизации.
+  * Пороги могут быть превышены. Например, если max_rows = 1 000 000, в буфере уже было 500 000 строк,
+  *  и добавляется кусок из 800 000 строк, то в буфере окажется 1 300 000 строк, и затем такой блок будет записан в подчинённую таблицу
+  *
+  * При уничтожении таблицы типа Buffer и при завершении работы, все данные сбрасываются.
+  * Данные в буфере не реплицируются, не логгируются на диск, не индексируются. При грубом перезапуске сервера, данные пропадают.
+  */
+class StorageBuffer : public IStorage
+{
+friend class BufferBlockInputStream;
+friend class BufferBlockOutputStream;
+
+public:
+	/// Пороги.
+	struct Thresholds
+	{
+		time_t time;	/// Количество секунд от момента вставки первой строчки в блок.
+		size_t rows;	/// Количество строк в блоке.
+		size_t bytes;	/// Количество (несжатых) байт в блоке.
+	};
+
+	/** num_shards - уровень внутреннего параллелизма (количество независимых буферов)
+	  * Буфер сбрасывается, если превышены все минимальные пороги или хотя бы один из максимальных.
+	  */
+	static StoragePtr create(const std::string & name_, NamesAndTypesListPtr columns_, Context & context_,
+		size_t num_shards_, const Thresholds & min_thresholds_, const Thresholds & max_thresholds_,
+		const String & destination_database_, const String & destination_table_);
+
+	std::string getName() const override { return "Buffer"; }
+	std::string getTableName() const override { return name; }
+
+	const NamesAndTypesList & getColumnsList() const override { return *columns; }
+
+	BlockInputStreams read(
+		const Names & column_names,
+		ASTPtr query,
+		const Settings & settings,
+		QueryProcessingStage::Enum & processed_stage,
+		size_t max_block_size = DEFAULT_BLOCK_SIZE,
+		unsigned threads = 1) override;
+
+	BlockOutputStreamPtr write(ASTPtr query) override;
+
+	/// Сбрасывает все буферы в подчинённую таблицу.
+	void shutdown() override;
+	bool optimize() override;
+
+	void rename(const String & new_path_to_db, const String & new_database_name, const String & new_table_name) override { name = new_table_name; }
+
+	bool supportsSampling() const override { return true; }
+	bool supportsFinal() const override { return true; }
+	bool supportsPrewhere() const override { return true; }
+
+	/// Структура подчинённой таблицы не проверяется и не изменяется.
+	void alter(const AlterCommands & params, const String & database_name, const String & table_name, Context & context) override;
+
+private:
+	String name;
+	NamesAndTypesListPtr columns;
+
+	Context & context;
+
+	struct Buffer
+	{
+		time_t first_write_time = 0;
+		Block data;
+		std::mutex mutex;
+	};
+
+	/// Имеется num_shards независимых буферов.
+	const size_t num_shards;
+	std::vector<Buffer> buffers;
+
+	const Thresholds min_thresholds;
+	const Thresholds max_thresholds;
+
+	const String destination_database;
+	const String destination_table;
+	bool no_destination;	/// Если задано - не записывать данные из буфера, а просто опустошать буфер.
+
+	Logger * log;
+
+	/// Выполняет сброс данных по таймауту.
+	std::thread flush_thread;
+
+	StorageBuffer(const std::string & name_, NamesAndTypesListPtr columns_, Context & context_,
+		size_t num_shards_, const Thresholds & min_thresholds_, const Thresholds & max_thresholds_,
+		const String & destination_database_, const String & destination_table_);
+
+	/// Сбросить буфер. Если выставлено check_thresholds - сбрасывает только если превышены пороги.
+	void flushBuffer(Buffer & buffer, bool check_thresholds);
+	bool checkThresholds(Buffer & buffer, time_t current_time, size_t additional_rows = 0, size_t additional_bytes = 0);
+
+	/// Аргумент table передаётся, так как иногда вычисляется заранее. Он должен соответствовать destination-у.
+	void writeBlockToDestination(const Block & block, StoragePtr table);
+
+	Poco::Event shutdown_event;
+	void flushThread();
+};
+
+}

dbms/include/DB/Storages/StorageMemory.h

@@ -24,13 +24,13 @@ public:
 		std::stringstream res;
 		res << "Memory(" << &*begin << ", " << &*end;
 
-		for (size_t i = 0; i < column_names.size(); ++i)
-			res << ", " << column_names[i];
+		for (const auto & name : column_names)
+			res << ", " << name;
 
 		res << ")";
 		return res.str();
 	}
-	
+
 protected:
 	Block readImpl();
 private:
@@ -92,7 +92,7 @@ private:
 	BlocksList data;
 
 	Poco::FastMutex mutex;
-	
+
 	StorageMemory(const std::string & name_, NamesAndTypesListPtr columns_);
 };
 

dbms/include/DB/Storages/StorageView.h

@@ -7,7 +7,8 @@
 namespace DB
 {
 
-class StorageView : public IStorage {
+class StorageView : public IStorage
+{
 
 public:
 	static StoragePtr create(const String & table_name_, const String & database_name_,
@@ -16,7 +17,7 @@ public:
 	std::string getName() const override { return "View"; }
 	std::string getTableName() const override { return table_name; }
 	const NamesAndTypesList & getColumnsList() const override { return *columns; }
-	DB::ASTPtr getInnerQuery() const { return inner_query.clone(); };
+	ASTPtr getInnerQuery() const { return inner_query.clone(); };
 
 	/// Пробрасывается внутрь запроса и решается на его уровне.
 	bool supportsSampling() const override { return true; }

dbms/src/Client/Benchmark.cpp

@@ -239,9 +239,8 @@ private:
 		Stopwatch watch;
 		RemoteBlockInputStream stream(connection, query, nullptr);
 
-		size_t read_rows = 0;
-		size_t read_bytes = 0;
-		stream.setProgressCallback([&](size_t rows_inc, size_t bytes_inc) { read_rows += rows_inc; read_bytes += bytes_inc; });
+		Progress progress;
+		stream.setProgressCallback([&progress](const Progress & value) { progress.incrementPiecewiseAtomically(value); });
 
 		stream.readPrefix();
 		while (Block block = stream.read())
@@ -253,8 +252,8 @@ private:
 		double seconds = watch.elapsedSeconds();
 
 		Poco::ScopedLock<Poco::FastMutex> lock(mutex);
-		info_per_interval.add(seconds, read_rows, read_bytes, info.rows, info.bytes);
-		info_total.add(seconds, read_rows, read_bytes, info.rows, info.bytes);
+		info_per_interval.add(seconds, progress.rows, progress.bytes, info.rows, info.bytes);
+		info_total.add(seconds, progress.rows, progress.bytes, info.rows, info.bytes);
 	}
 
 

dbms/src/Client/Client.cpp

@@ -48,6 +48,8 @@
 #include "InterruptListener.h"
 
 #include <DB/Common/ExternalTable.h>
+#include <DB/Common/UnicodeBar.h>
+#include <DB/Common/formatReadable.h>
 
 
 /// http://en.wikipedia.org/wiki/ANSI_escape_code
@@ -86,6 +88,8 @@ private:
 	bool is_interactive = true;			/// Использовать readline интерфейс или batch режим.
 	bool stdin_is_not_tty = false;		/// stdin - не терминал.
 
+	winsize terminal_size {};			/// Размер терминала - для вывода прогресс-бара.
+
 	SharedPtr<Connection> connection;	/// Соединение с БД.
 	String query;						/// Текущий запрос.
 
@@ -121,8 +125,10 @@ private:
 
 	Stopwatch watch;
 
-	size_t rows_read_on_server = 0;
-	size_t bytes_read_on_server = 0;
+	/// С сервера периодически приходит информация, о том, сколько прочитано данных за прошедшее время.
+	Progress progress;
+	bool show_progress_bar = false;
+
 	size_t written_progress_chars = 0;
 	bool written_first_block = false;
 
@@ -363,6 +369,9 @@ private:
 
 				try
 				{
+					/// Выясняем размер терминала.
+					ioctl(0, TIOCGWINSZ, &terminal_size);
+
 					if (!process(query))
 						break;
 				}
@@ -470,8 +479,8 @@ private:
 			return true;
 
 		processed_rows = 0;
-		rows_read_on_server = 0;
-		bytes_read_on_server = 0;
+		progress.reset();
+		show_progress_bar = false;
 		written_progress_chars = 0;
 		written_first_block = false;
 
@@ -511,7 +520,7 @@ private:
 			std::cout << std::endl
 				<< processed_rows << " rows in set. Elapsed: " << watch.elapsedSeconds() << " sec. ";
 
-			if (rows_read_on_server >= 1000)
+			if (progress.rows >= 1000)
 				writeFinalProgress();
 
 			std::cout << std::endl << std::endl;
@@ -809,11 +818,9 @@ private:
 	}
 
 
-	void onProgress(const Progress & progress)
+	void onProgress(const Progress & value)
 	{
-		rows_read_on_server += progress.rows;
-		bytes_read_on_server += progress.bytes;
-
+		progress.increment(value);
 		writeProgress();
 	}
 
@@ -851,31 +858,62 @@ private:
 		std::stringstream message;
 		message << indicators[increment % 8]
 			<< std::fixed << std::setprecision(3)
-			<< " Progress: " << rows_read_on_server << " rows, " << bytes_read_on_server / 1000000.0 << " MB";
+			<< " Progress: ";
+
+		message
+			<< formatReadableQuantity(progress.rows) << " rows, "
+			<< formatReadableSizeWithDecimalSuffix(progress.bytes);
 
 		size_t elapsed_ns = watch.elapsed();
 		if (elapsed_ns)
 			message << " ("
-				<< rows_read_on_server * 1000000000.0 / elapsed_ns << " rows/s., "
-				<< bytes_read_on_server * 1000.0 / elapsed_ns << " MB/s.) ";
+				<< formatReadableQuantity(progress.rows * 1000000000.0 / elapsed_ns) << " rows/s., "
+				<< formatReadableSizeWithDecimalSuffix(progress.bytes * 1000000000.0 / elapsed_ns) << "/s.) ";
 		else
 			message << ". ";
 
-		written_progress_chars = message.str().size() - 13;
-		std::cerr << DISABLE_LINE_WRAPPING << message.rdbuf() << ENABLE_LINE_WRAPPING;
+		written_progress_chars = message.str().size() - (increment % 8 == 7 ? 10 : 13);
+		std::cerr << DISABLE_LINE_WRAPPING << message.rdbuf();
+
+		/** Если известно приблизительное общее число строк, которых нужно обработать - можно вывести прогрессбар.
+		  * Чтобы не было "мерцания", выводим его только если с момента начала выполнения запроса прошло хотя бы пол секунды,
+		  *  и если к этому моменту запрос обработан менее чем наполовину.
+		  */
+		ssize_t width_of_progress_bar = static_cast<ssize_t>(terminal_size.ws_col) - written_progress_chars - strlen(" 99%");
+
+		if (show_progress_bar
+			|| (width_of_progress_bar > 0
+				&& progress.total_rows
+				&& elapsed_ns > 500000000
+				&& progress.rows * 2 < progress.total_rows))
+		{
+			show_progress_bar = true;
+
+			size_t total_rows_corrected = std::max(progress.rows, progress.total_rows);
+
+			std::string bar = UnicodeBar::render(UnicodeBar::getWidth(progress.rows, 0, total_rows_corrected, width_of_progress_bar));
+			std::cerr << "\033[0;32m" << bar << "\033[0m";
+			if (width_of_progress_bar > static_cast<ssize_t>(bar.size() / UNICODE_BAR_CHAR_SIZE))
+				std::cerr << std::string(width_of_progress_bar - bar.size() / UNICODE_BAR_CHAR_SIZE, ' ');
+			std::cerr << ' ' << (99 * progress.rows / total_rows_corrected) << '%';	/// Чуть-чуть занижаем процент, чтобы не показывать 100%.
+		}
+
+		std::cerr << ENABLE_LINE_WRAPPING;
 		++increment;
 	}
 
 
 	void writeFinalProgress()
 	{
-		std::cout << "Processed " << rows_read_on_server << " rows, " << bytes_read_on_server / 1000000.0 << " MB";
+		std::cout << "Processed "
+			<< formatReadableQuantity(progress.rows) << " rows, "
+			<< formatReadableSizeWithDecimalSuffix(progress.bytes);
 
 		size_t elapsed_ns = watch.elapsed();
 		if (elapsed_ns)
 			std::cout << " ("
-				<< rows_read_on_server * 1000000000.0 / elapsed_ns << " rows/s., "
-				<< bytes_read_on_server * 1000.0 / elapsed_ns << " MB/s.) ";
+				<< formatReadableQuantity(progress.rows * 1000000000.0 / elapsed_ns) << " rows/s., "
+				<< formatReadableSizeWithDecimalSuffix(progress.bytes * 1000000000.0 / elapsed_ns) << "/s.) ";
 		else
 			std::cout << ". ";
 	}

dbms/src/Client/Connection.cpp

@@ -470,7 +470,7 @@ Progress Connection::receiveProgress()
 	//LOG_TRACE(log_wrapper.get(), "Receiving progress (" << getServerAddress() << ")");
 
 	Progress progress;
-	progress.read(*in);
+	progress.read(*in, server_revision);
 	return progress;
 }
 

dbms/src/Common/MemoryTracker.cpp

@@ -1,30 +1,16 @@
 #include <Yandex/likely.h>
 #include <Yandex/logger_useful.h>
 #include <DB/Core/Exception.h>
+#include <DB/Common/formatReadable.h>
 #include <DB/IO/WriteHelpers.h>
 #include <iomanip>
 
 #include <DB/Common/MemoryTracker.h>
 
 
-static std::string formatReadableSize(double size)
-{
-	const char* units[] = {"B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB", "ZiB", "YiB"};
-	size_t i = 0;
-	while (i + 1 < sizeof(units) / sizeof(units[0]) &&
-		fabs(size) >= 1024)
-	{
-		size /= 1024;
-        ++i;
-    }
-    std::stringstream ss;
-    ss << std::fixed << std::setprecision(i) << size << ' ' << units[i];
-    return ss.str();
-}
-
 MemoryTracker::~MemoryTracker()
 {
-	LOG_DEBUG(&Logger::get("MemoryTracker"), "Peak memory usage for query: " << formatReadableSize(peak) << ".");
+	LOG_DEBUG(&Logger::get("MemoryTracker"), "Peak memory usage for query: " << formatReadableSizeWithBinarySuffix(peak) << ".");
 }
 
 void MemoryTracker::alloc(Int64 size)
@@ -34,9 +20,9 @@ void MemoryTracker::alloc(Int64 size)
 	if (unlikely(limit && will_be > limit))
 	{
 		free(size);
-		throw DB::Exception("Memory limit exceeded: would use " + formatReadableSize(will_be) + ""
+		throw DB::Exception("Memory limit exceeded: would use " + formatReadableSizeWithBinarySuffix(will_be) + ""
 			" (attempt to allocate chunk of " + DB::toString(size) + " bytes)"
-			", maximum: " + formatReadableSize(limit), DB::ErrorCodes::MEMORY_LIMIT_EXCEEDED);
+			", maximum: " + formatReadableSizeWithBinarySuffix(limit), DB::ErrorCodes::MEMORY_LIMIT_EXCEEDED);
 	}
 
 	if (will_be > peak)

dbms/src/Common/formatReadable.cpp

@@ -0,0 +1,36 @@
+#include <cmath>
+#include <sstream>
+#include <iomanip>
+
+#include <DB/Common/formatReadable.h>
+
+
+static std::string formatReadable(double size, int precision, const char ** units, size_t units_size, double delimiter)
+{
+	size_t i = 0;
+	for (; i + 1 < units_size && fabs(size) >= delimiter; ++i)
+		size /= delimiter;
+
+    std::stringstream ss;
+    ss << std::fixed << std::setprecision(precision) << size << units[i];
+    return ss.str();
+}
+
+
+std::string formatReadableSizeWithBinarySuffix(double value, int precision)
+{
+	const char * units[]  = {" B", " KiB", " MiB", " GiB", " TiB", " PiB", " EiB", " ZiB", " YiB"};
+	return formatReadable(value, precision, units, sizeof(units) / sizeof(units[0]), 1024);
+}
+
+std::string formatReadableSizeWithDecimalSuffix(double value, int precision)
+{
+	const char * units[] = {" B", " KB", " MB", " GB", " TB", " PB", " EB", " ZB", " YB"};
+	return formatReadable(value, precision, units, sizeof(units) / sizeof(units[0]), 1000);
+}
+
+std::string formatReadableQuantity(double value, int precision)
+{
+	const char * units[] = {"", " thousand", " million", " billion", " trillion", " quadrillion"};
+	return formatReadable(value, precision, units, sizeof(units) / sizeof(units[0]), 1000);
+}

dbms/src/DataStreams/IProfilingBlockInputStream.cpp

@@ -166,7 +166,7 @@ Block IProfilingBlockInputStream::read()
 		cancel();
 	}
 
-	progress(res.rowsInFirstColumn(), res.bytes());
+	progress(Progress(res.rowsInFirstColumn(), res.bytes()));
 
 	return res;
 }
@@ -295,36 +295,45 @@ void IProfilingBlockInputStream::checkQuota(Block & block)
 }
 
 
-void IProfilingBlockInputStream::progressImpl(size_t rows, size_t bytes)
+void IProfilingBlockInputStream::progressImpl(const Progress & value)
 {
 	/// Данные для прогресса берутся из листовых источников.
 	if (children.empty())
 	{
 		if (progress_callback)
-			progress_callback(rows, bytes);
+			progress_callback(value);
 
 		if (process_list_elem)
 		{
-			if (!process_list_elem->update(rows, bytes))
+			if (!process_list_elem->update(value))
 				cancel();
 
-			/// Общее количество данных, обработанных во всех листовых источниках, возможно, на удалённых серверах.
+			/// Общее количество данных, обработанных или предполагаемых к обработке во всех листовых источниках, возможно, на удалённых серверах.
 
-			size_t total_rows = process_list_elem->rows_processed;
-			size_t total_bytes = process_list_elem->bytes_processed;
+			size_t rows_processed = process_list_elem->progress.rows;
+			size_t bytes_processed = process_list_elem->progress.bytes;
+
+			size_t total_rows_estimate = std::max(process_list_elem->progress.rows, process_list_elem->progress.total_rows);
 
 			/** Проверяем ограничения на объём данных для чтения, скорость выполнения запроса, квоту на объём данных для чтения.
 			  * NOTE: Может быть, имеет смысл сделать, чтобы они проверялись прямо в ProcessList?
 			  */
 
 			if (limits.mode == LIMITS_TOTAL
-				&& ((limits.max_rows_to_read && total_rows > limits.max_rows_to_read)
-					|| (limits.max_bytes_to_read && total_bytes > limits.max_bytes_to_read)))
+				&& ((limits.max_rows_to_read && total_rows_estimate > limits.max_rows_to_read)
+					|| (limits.max_bytes_to_read && bytes_processed > limits.max_bytes_to_read)))
 			{
 				if (limits.read_overflow_mode == OverflowMode::THROW)
-					throw Exception("Limit for rows to read exceeded: read " + toString(total_rows)
-						+ " rows, maximum: " + toString(limits.max_rows_to_read),
-						ErrorCodes::TOO_MUCH_ROWS);
+				{
+					if (limits.max_rows_to_read && total_rows_estimate > limits.max_rows_to_read)
+						throw Exception("Limit for rows to read exceeded: " + toString(total_rows_estimate)
+							+ " rows read (or to read), maximum: " + toString(limits.max_rows_to_read),
+							ErrorCodes::TOO_MUCH_ROWS);
+					else
+						throw Exception("Limit for (uncompressed) bytes to read exceeded: " + toString(bytes_processed)
+							+ " bytes read, maximum: " + toString(limits.max_bytes_to_read),
+							ErrorCodes::TOO_MUCH_ROWS);
+				}
 				else if (limits.read_overflow_mode == OverflowMode::BREAK)
 					cancel();
 				else
@@ -336,9 +345,9 @@ void IProfilingBlockInputStream::progressImpl(size_t rows, size_t bytes)
 				double total_elapsed = info.total_stopwatch.elapsedSeconds();
 
 				if (total_elapsed > limits.timeout_before_checking_execution_speed.totalMicroseconds() / 1000000.0
-					&& total_rows / total_elapsed < limits.min_execution_speed)
+					&& rows_processed / total_elapsed < limits.min_execution_speed)
 				{
-					throw Exception("Query is executing too slow: " + toString(total_rows / total_elapsed)
+					throw Exception("Query is executing too slow: " + toString(rows_processed / total_elapsed)
 						+ " rows/sec., minimum: " + toString(limits.min_execution_speed),
 						ErrorCodes::TOO_SLOW);
 				}
@@ -346,7 +355,7 @@ void IProfilingBlockInputStream::progressImpl(size_t rows, size_t bytes)
 
 			if (quota != nullptr && limits.mode == LIMITS_TOTAL)
 			{
-				quota->checkAndAddReadRowsBytes(time(0), rows, bytes);
+				quota->checkAndAddReadRowsBytes(time(0), value.rows, value.bytes);
 			}
 		}
 	}

dbms/src/Functions/FunctionsArray.cpp

@@ -6,7 +6,7 @@ namespace DB
 
 void registerFunctionsArray(FunctionFactory & factory)
 {
-	#define F [](const Context & context) -> IFunction*
+	#define F [](const Context & context) -> IFunction *
 
 	factory.registerFunction("array", 				F { return new FunctionArray; });
 	factory.registerFunction("arrayElement", 		F { return new FunctionArrayElement; });
@@ -15,6 +15,20 @@ void registerFunctionsArray(FunctionFactory & factory)
 	factory.registerFunction("countEqual", 			F { return new FunctionCountEqual; });
 	factory.registerFunction("arrayEnumerate", 		F { return new FunctionArrayEnumerate; });
 	factory.registerFunction("arrayEnumerateUniq", 	F { return new FunctionArrayEnumerateUniq; });
+	factory.registerFunction("emptyArrayUInt8",		F { return new FunctionEmptyArrayUInt8; });
+	factory.registerFunction("emptyArrayUInt16",	F { return new FunctionEmptyArrayUInt16; });
+	factory.registerFunction("emptyArrayUInt32",	F { return new FunctionEmptyArrayUInt32; });
+	factory.registerFunction("emptyArrayUInt64",	F { return new FunctionEmptyArrayUInt64; });
+	factory.registerFunction("emptyArrayInt8",		F { return new FunctionEmptyArrayInt8; });
+	factory.registerFunction("emptyArrayInt16",		F { return new FunctionEmptyArrayInt16; });
+	factory.registerFunction("emptyArrayInt32",		F { return new FunctionEmptyArrayInt32; });
+	factory.registerFunction("emptyArrayInt64",		F { return new FunctionEmptyArrayInt64; });
+	factory.registerFunction("emptyArrayFloat32",	F { return new FunctionEmptyArrayFloat32; });
+	factory.registerFunction("emptyArrayFloat64",	F { return new FunctionEmptyArrayFloat64; });
+	factory.registerFunction("emptyArrayDate",		F { return new FunctionEmptyArrayDate; });
+	factory.registerFunction("emptyArrayDateTime",	F { return new FunctionEmptyArrayDateTime; });
+	factory.registerFunction("emptyArrayString",	F { return new FunctionEmptyArrayString; });
+
 
 	#undef F
 }

dbms/src/Functions/FunctionsCoding.cpp

@@ -6,9 +6,11 @@ namespace DB
 
 void registerFunctionsCoding(FunctionFactory & factory)
 {
-	#define F [](const Context & context) -> IFunction*
+	#define F [](const Context & context) -> IFunction *
 
 	factory.registerFunction("toStringCutToZero", 	F { return new FunctionToStringCutToZero; });
+	factory.registerFunction("IPv6NumToString",		F { return new FunctionIPv6NumToString; });
+	factory.registerFunction("IPv6StringToNum",		F { return new FunctionIPv6StringToNum; });
 	factory.registerFunction("IPv4NumToString", 	F { return new FunctionIPv4NumToString; });
 	factory.registerFunction("IPv4StringToNum", 	F { return new FunctionIPv4StringToNum; });
 	factory.registerFunction("hex", 				F { return new FunctionHex; });

dbms/src/Functions/FunctionsHashing.cpp

@@ -7,10 +7,15 @@ namespace DB
 
 void registerFunctionsHashing(FunctionFactory & factory)
 {
-	#define F [](const Context & context) -> IFunction*
+	#define F [](const Context & context) -> IFunction *
 
 	factory.registerFunction("halfMD5", 	F { return new FunctionHalfMD5; });
+	factory.registerFunction("MD5",			F { return new FunctionMD5; });
+	factory.registerFunction("SHA1",		F { return new FunctionSHA1; });
+	factory.registerFunction("SHA224",		F { return new FunctionSHA224; });
+	factory.registerFunction("SHA256",		F { return new FunctionSHA256; });
 	factory.registerFunction("sipHash64", 	F { return new FunctionSipHash64; });
+	factory.registerFunction("sipHash128",	F { return new FunctionSipHash128; });
 	factory.registerFunction("cityHash64", 	F { return new FunctionCityHash64; });
 	factory.registerFunction("intHash32", 	F { return new FunctionIntHash32; });
 	factory.registerFunction("intHash64", 	F { return new FunctionIntHash64; });

dbms/src/Functions/FunctionsHigherOrder.cpp

@@ -6,7 +6,7 @@ namespace DB
 
 void registerFunctionsHigherOrder(FunctionFactory & factory)
 {
-	#define F [](const Context & context) -> IFunction*
+	#define F [](const Context & context) -> IFunction *
 
 	factory.registerFunction("arrayMap", 		F { return new FunctionArrayMap; });
 	factory.registerFunction("arrayFilter", 	F { return new FunctionArrayFilter; });
@@ -14,6 +14,8 @@ void registerFunctionsHigherOrder(FunctionFactory & factory)
 	factory.registerFunction("arrayExists", 	F { return new FunctionArrayExists; });
 	factory.registerFunction("arrayAll", 		F { return new FunctionArrayAll; });
 	factory.registerFunction("arraySum", 		F { return new FunctionArraySum; });
+	factory.registerFunction("arrayFirst", 		F { return new FunctionArrayFirst; });
+	factory.registerFunction("arrayFirstIndex", 		F { return new FunctionArrayFirstIndex; });
 
 	#undef F
 }

dbms/src/Functions/FunctionsURL.cpp

@@ -11,6 +11,7 @@ void registerFunctionsURL(FunctionFactory & factory)
 	factory.registerFunction("protocol", 					F { return new FunctionProtocol; });
 	factory.registerFunction("domain", 						F { return new FunctionDomain; });
 	factory.registerFunction("domainWithoutWWW", 			F { return new FunctionDomainWithoutWWW; });
+	factory.registerFunction("firstSignificantSubdomain",	F { return new FunctionFirstSignificantSubdomain; });
 	factory.registerFunction("topLevelDomain", 				F { return new FunctionTopLevelDomain; });
 	factory.registerFunction("path", 						F { return new FunctionPath; });
 	factory.registerFunction("queryString", 				F { return new FunctionQueryString; });
@@ -21,6 +22,7 @@ void registerFunctionsURL(FunctionFactory & factory)
 	factory.registerFunction("extractURLParameterNames", 	F { return new FunctionExtractURLParameterNames; });
 	factory.registerFunction("URLHierarchy", 				F { return new FunctionURLHierarchy; });
 	factory.registerFunction("URLPathHierarchy", 			F { return new FunctionURLPathHierarchy; });
+	factory.registerFunction("cutToFirstSignificantSubdomain", F { return new FunctionCutToFirstSignificantSubdomain; });
 	factory.registerFunction("cutWWW", 						F { return new FunctionCutWWW; });
 	factory.registerFunction("cutQueryString", 				F { return new FunctionCutQueryString; });
 	factory.registerFunction("cutFragment", 				F { return new FunctionCutFragment; });

dbms/src/Interpreters/Aggregator.cpp

@@ -125,7 +125,18 @@ AggregatedDataVariants::Type Aggregator::chooseAggregationMethod(const ConstColu
 
 	/// Если есть один числовой ключ, который помещается в 64 бита
 	if (keys_size == 1 && key_columns[0]->isNumeric())
-		return AggregatedDataVariants::KEY_64;
+	{
+		size_t size_of_field = key_columns[0]->sizeOfField();
+		if (size_of_field == 1)
+			return AggregatedDataVariants::KEY_8;
+		if (size_of_field == 2)
+			return AggregatedDataVariants::KEY_16;
+		if (size_of_field == 4)
+			return AggregatedDataVariants::KEY_32;
+		if (size_of_field == 8)
+			return AggregatedDataVariants::KEY_64;
+		throw Exception("Logical error: numeric column has sizeOfField not in 1, 2, 4, 8.", ErrorCodes::LOGICAL_ERROR);
+	}
 
 	/// Если ключи помещаются в 128 бит, будем использовать хэш-таблицу по упакованным в 128-бит ключам
 	if (keys_fit_128_bits)
@@ -167,8 +178,12 @@ void Aggregator::createAggregateStates(AggregateDataPtr & aggregate_data) const
 }
 
 
+/** Интересно - если убрать noinline, то gcc зачем-то инлайнит эту функцию, и производительность уменьшается (~10%).
+  * (Возможно из-за того, что после инлайна этой функции, перестают инлайниться более внутренние функции.)
+  * Инлайнить не имеет смысла, так как внутренний цикл находится целиком внутри этой функции.
+  */
 template <typename Method>
-void Aggregator::executeImpl(
+void NO_INLINE Aggregator::executeImpl(
 	Method & method,
 	Arena * aggregates_pool,
 	size_t rows,
@@ -191,7 +206,7 @@ void Aggregator::executeImpl(
 		/// Получаем ключ для вставки в хэш-таблицу.
 		typename Method::Key key = method.getKey(key_columns, keys_size, i, key_sizes, keys);
 
-		if (!no_more_keys)	/// Вставляем.
+		if (Method::never_overflows || !no_more_keys)	/// Вставляем.
 			method.data.emplace(key, it, inserted);
 		else
 		{
@@ -203,7 +218,7 @@ void Aggregator::executeImpl(
 		}
 
 		/// Если ключ не поместился, и данные не надо агрегировать в отдельную строку, то делать нечего.
-		if (overflow && !overflow_row)
+		if (!Method::never_overflows && overflow && !overflow_row)
 			continue;
 
 		/// Если вставили новый ключ - инициализируем состояния агрегатных функций, и возможно, что-нибудь связанное с ключом.
@@ -216,7 +231,7 @@ void Aggregator::executeImpl(
 			createAggregateStates(aggregate_data);
 		}
 
-		AggregateDataPtr value = !overflow ? Method::getAggregateData(it->second) : overflow_row;
+		AggregateDataPtr value = (Method::never_overflows || !overflow) ? Method::getAggregateData(it->second) : overflow_row;
 
 		/// Добавляем значения в агрегатные функции.
 		for (size_t j = 0; j < aggregates_size; ++j)
@@ -226,7 +241,7 @@ void Aggregator::executeImpl(
 
 
 template <typename Method>
-void Aggregator::convertToBlockImpl(
+void NO_INLINE Aggregator::convertToBlockImpl(
 	Method & method,
 	ColumnPlainPtrs & key_columns,
 	AggregateColumnsData & aggregate_columns,
@@ -262,7 +277,7 @@ void Aggregator::convertToBlockImpl(
 
 
 template <typename Method>
-void Aggregator::mergeDataImpl(
+void NO_INLINE Aggregator::mergeDataImpl(
 	Method & method_dst,
 	Method & method_src) const
 {
@@ -294,7 +309,7 @@ void Aggregator::mergeDataImpl(
 
 
 template <typename Method>
-void Aggregator::mergeStreamsImpl(
+void NO_INLINE Aggregator::mergeStreamsImpl(
 	Method & method,
 	Arena * aggregates_pool,
 	size_t start_row,
@@ -336,7 +351,7 @@ void Aggregator::mergeStreamsImpl(
 
 
 template <typename Method>
-void Aggregator::destroyImpl(
+void NO_INLINE Aggregator::destroyImpl(
 	Method & method) const
 {
 	for (typename Method::const_iterator it = method.data.begin(); it != method.data.end(); ++it)
@@ -372,8 +387,14 @@ bool Aggregator::executeOnBlock(Block & block, AggregatedDataVariants & result,
 
 	/// Запоминаем столбцы, с которыми будем работать
 	for (size_t i = 0; i < keys_size; ++i)
+	{
 		key_columns[i] = block.getByPosition(keys[i]).column;
 
+		if (key_columns[i]->isConst())
+			throw Exception("Constants is not allowed as GROUP BY keys"
+				" (but all of them must be eliminated in ExpressionAnalyzer)", ErrorCodes::ILLEGAL_COLUMN);
+	}
+
 	for (size_t i = 0; i < aggregates_size; ++i)
 	{
 		for (size_t j = 0; j < aggregate_columns[i].size(); ++j)
@@ -434,7 +455,16 @@ bool Aggregator::executeOnBlock(Block & block, AggregatedDataVariants & result,
 
 	AggregateDataPtr overflow_row_ptr = overflow_row ? result.without_key : nullptr;
 
-	if (result.type == AggregatedDataVariants::KEY_64)
+	if (result.type == AggregatedDataVariants::KEY_8)
+		executeImpl(*result.key8, result.aggregates_pool, rows, key_columns, aggregate_columns,
+			result.key_sizes, key, no_more_keys, overflow_row_ptr);
+	else if (result.type == AggregatedDataVariants::KEY_16)
+		executeImpl(*result.key16, result.aggregates_pool, rows, key_columns, aggregate_columns,
+			result.key_sizes, key, no_more_keys, overflow_row_ptr);
+	else if (result.type == AggregatedDataVariants::KEY_32)
+		executeImpl(*result.key32, result.aggregates_pool, rows, key_columns, aggregate_columns,
+			result.key_sizes, key, no_more_keys, overflow_row_ptr);
+	else if (result.type == AggregatedDataVariants::KEY_64)
 		executeImpl(*result.key64, result.aggregates_pool, rows, key_columns, aggregate_columns,
 			result.key_sizes, key, no_more_keys, overflow_row_ptr);
 	else if (result.type == AggregatedDataVariants::KEY_STRING)
@@ -590,7 +620,16 @@ Block Aggregator::convertToBlock(AggregatedDataVariants & data_variants, bool fi
 
 		size_t start_row = overflow_row ? 1 : 0;
 
-		if (data_variants.type == AggregatedDataVariants::KEY_64)
+		if (data_variants.type == AggregatedDataVariants::KEY_8)
+			convertToBlockImpl(*data_variants.key8, key_columns, aggregate_columns,
+							final_aggregate_columns, data_variants.key_sizes, start_row, final);
+		else if (data_variants.type == AggregatedDataVariants::KEY_16)
+			convertToBlockImpl(*data_variants.key16, key_columns, aggregate_columns,
+							final_aggregate_columns, data_variants.key_sizes, start_row, final);
+		else if (data_variants.type == AggregatedDataVariants::KEY_32)
+			convertToBlockImpl(*data_variants.key32, key_columns, aggregate_columns,
+							final_aggregate_columns, data_variants.key_sizes, start_row, final);
+		else if (data_variants.type == AggregatedDataVariants::KEY_64)
 			convertToBlockImpl(*data_variants.key64, key_columns, aggregate_columns,
 							final_aggregate_columns, data_variants.key_sizes, start_row, final);
 		else if (data_variants.type == AggregatedDataVariants::KEY_STRING)
@@ -694,7 +733,13 @@ AggregatedDataVariantsPtr Aggregator::merge(ManyAggregatedDataVariants & data_va
 			current_data = nullptr;
 		}
 
-		if (res->type == AggregatedDataVariants::KEY_64)
+		if (res->type == AggregatedDataVariants::KEY_8)
+			mergeDataImpl(*res->key8, *current.key8);
+		else if (res->type == AggregatedDataVariants::KEY_16)
+			mergeDataImpl(*res->key16, *current.key16);
+		else if (res->type == AggregatedDataVariants::KEY_32)
+			mergeDataImpl(*res->key32, *current.key32);
+		else if (res->type == AggregatedDataVariants::KEY_64)
 			mergeDataImpl(*res->key64, *current.key64);
 		else if (res->type == AggregatedDataVariants::KEY_STRING)
 			mergeDataImpl(*res->key_string, *current.key_string);
@@ -782,7 +827,13 @@ void Aggregator::merge(BlockInputStreamPtr stream, AggregatedDataVariants & resu
 
 		size_t start_row = overflow_row ? 1 : 0;
 
-		if (result.type == AggregatedDataVariants::KEY_64)
+		if (result.type == AggregatedDataVariants::KEY_8)
+			mergeStreamsImpl(*result.key8, result.aggregates_pool, start_row, rows, key_columns, aggregate_columns, key_sizes, key);
+		else if (result.type == AggregatedDataVariants::KEY_16)
+			mergeStreamsImpl(*result.key16, result.aggregates_pool, start_row, rows, key_columns, aggregate_columns, key_sizes, key);
+		else if (result.type == AggregatedDataVariants::KEY_32)
+			mergeStreamsImpl(*result.key32, result.aggregates_pool, start_row, rows, key_columns, aggregate_columns, key_sizes, key);
+		else if (result.type == AggregatedDataVariants::KEY_64)
 			mergeStreamsImpl(*result.key64, result.aggregates_pool, start_row, rows, key_columns, aggregate_columns, key_sizes, key);
 		else if (result.type == AggregatedDataVariants::KEY_STRING)
 			mergeStreamsImpl(*result.key_string, result.aggregates_pool, start_row, rows, key_columns, aggregate_columns, key_sizes, key);
@@ -818,7 +869,13 @@ void Aggregator::destroyAllAggregateStates(AggregatedDataVariants & result)
 					aggregate_functions[i]->destroy(res_data + offsets_of_aggregate_states[i]);
 	}
 
-	if (result.type == AggregatedDataVariants::KEY_64)
+	if (result.type == AggregatedDataVariants::KEY_8)
+		destroyImpl(*result.key8);
+	else if (result.type == AggregatedDataVariants::KEY_16)
+		destroyImpl(*result.key16);
+	else if (result.type == AggregatedDataVariants::KEY_32)
+		destroyImpl(*result.key32);
+	else if (result.type == AggregatedDataVariants::KEY_64)
 		destroyImpl(*result.key64);
 	else if (result.type == AggregatedDataVariants::KEY_STRING)
 		destroyImpl(*result.key_string);

dbms/src/Interpreters/InterpreterCreateQuery.cpp

@@ -77,7 +77,7 @@ StoragePtr InterpreterCreateQuery::execute(bool assume_metadata_exists)
 
 	SharedPtr<InterpreterSelectQuery> interpreter_select;
 	Block select_sample;
-	/// Для таблиц типа вью, чтобы получить столбцы, может понадобиться sample block.
+	/// Для таблиц типа view, чтобы получить столбцы, может понадобиться sample_block.
 	if (create.select && (!create.attach || (!create.columns && (create.is_view || create.is_materialized_view))))
 	{
 		interpreter_select = new InterpreterSelectQuery(create.select, context);

dbms/src/Interpreters/SplittingAggregator.cpp

@@ -48,7 +48,10 @@ void SplittingAggregator::execute(BlockInputStreamPtr stream, ManyAggregatedData
 			method = chooseAggregationMethod(key_columns, key_sizes);
 
 		/// Подготавливаем массивы, куда будут складываться ключи или хэши от ключей.
-		if (method == AggregatedDataVariants::KEY_64)
+		if (method == AggregatedDataVariants::KEY_8			/// TODO не использовать SplittingAggregator для маленьких ключей.
+			|| method == AggregatedDataVariants::KEY_16
+			|| method == AggregatedDataVariants::KEY_32
+			|| method == AggregatedDataVariants::KEY_64)
 		{
 			keys64.resize(rows);
 		}
@@ -96,7 +99,7 @@ void SplittingAggregator::execute(BlockInputStreamPtr stream, ManyAggregatedData
 
 		pool.wait();
 
-		rethrowFirstException(exceptions);
+		rethrowFirstException(exceptions);		/// TODO Заменить на future, packaged_task
 
 		/// Параллельно агрегируем в независимые хэш-таблицы
 
@@ -150,14 +153,17 @@ void SplittingAggregator::calculateHashesThread(Block & block, size_t begin, siz
 
 	try
 	{
-		if (method == AggregatedDataVariants::KEY_64)
+		if (method == AggregatedDataVariants::KEY_8
+			|| method == AggregatedDataVariants::KEY_16
+			|| method == AggregatedDataVariants::KEY_32
+			|| method == AggregatedDataVariants::KEY_64)
 		{
 			const IColumn & column = *key_columns[0];
 
 			for (size_t i = begin; i < end; ++i)
 			{
-				keys64[i] = column.get64(i);
-				thread_nums[i] = intHash32<0xd1f93e3190506c7cULL>(keys64[i]) % threads;
+				keys64[i] = column.get64(i);											/// TODO Убрать виртуальный вызов
+				thread_nums[i] = intHash32<0xd1f93e3190506c7cULL>(keys64[i]) % threads;	/// TODO более эффективная хэш-функция
 			}
 		}
 		else if (method == AggregatedDataVariants::KEY_STRING)
@@ -216,6 +222,45 @@ void SplittingAggregator::calculateHashesThread(Block & block, size_t begin, siz
 }
 
 
+template <typename FieldType>
+void SplittingAggregator::aggregateOneNumber(AggregatedDataVariants & result, size_t thread_no, bool no_more_keys)
+{
+	AggregatedDataWithUInt64Key & res = result.key64->data;
+
+	for (size_t i = 0; i < rows; ++i)
+	{
+		if (thread_nums[i] != thread_no)
+			continue;
+
+		/// Берём ключ
+		UInt64 key = keys64[i];
+
+		AggregatedDataWithUInt64Key::iterator it;
+		bool inserted;
+
+		if (!no_more_keys)
+			res.emplace(key, it, inserted);
+		else
+		{
+			inserted = false;
+			it = res.find(key);
+			if (res.end() == it)
+				continue;
+		}
+
+		if (inserted)
+		{
+			it->second = result.aggregates_pool->alloc(total_size_of_aggregate_states);
+			createAggregateStates(it->second);
+		}
+
+		/// Добавляем значения
+		for (size_t j = 0; j < aggregates_size; ++j)
+			aggregate_functions[j]->add(it->second + offsets_of_aggregate_states[j], &aggregate_columns[j][0], i);
+	}
+}
+
+
 void SplittingAggregator::aggregateThread(
 	Block & block, AggregatedDataVariants & result, size_t thread_no, ExceptionPtr & exception, MemoryTracker * memory_tracker)
 {
@@ -233,42 +278,14 @@ void SplittingAggregator::aggregateThread(
 		bool no_more_keys = max_rows_to_group_by && size_of_all_results > max_rows_to_group_by;
 		size_t old_result_size = result.size();
 
-		if (method == AggregatedDataVariants::KEY_64)
-		{
-			AggregatedDataWithUInt64Key & res = result.key64->data;
-
-			for (size_t i = 0; i < rows; ++i)
-			{
-				if (thread_nums[i] != thread_no)
-					continue;
-
-				/// Берём ключ
-				UInt64 key = keys64[i];
-
-				AggregatedDataWithUInt64Key::iterator it;
-				bool inserted;
-
-				if (!no_more_keys)
-					res.emplace(key, it, inserted);
-				else
-				{
-					inserted = false;
-					it = res.find(key);
-					if (res.end() == it)
-						continue;
-				}
-
-				if (inserted)
-				{
-					it->second = result.aggregates_pool->alloc(total_size_of_aggregate_states);
-					createAggregateStates(it->second);
-				}
-
-				/// Добавляем значения
-				for (size_t j = 0; j < aggregates_size; ++j)
-					aggregate_functions[j]->add(it->second + offsets_of_aggregate_states[j], &aggregate_columns[j][0], i);
-			}
-		}
+		if (method == AggregatedDataVariants::KEY_8)
+			aggregateOneNumber<UInt8>(result, thread_no, no_more_keys);
+		else if (method == AggregatedDataVariants::KEY_16)
+			aggregateOneNumber<UInt16>(result, thread_no, no_more_keys);
+		else if (method == AggregatedDataVariants::KEY_32)
+			aggregateOneNumber<UInt32>(result, thread_no, no_more_keys);
+		else if (method == AggregatedDataVariants::KEY_64)
+			aggregateOneNumber<UInt64>(result, thread_no, no_more_keys);
 		else if (method == AggregatedDataVariants::KEY_STRING)
 		{
 			AggregatedDataWithStringKey & res = result.key_string->data;

dbms/src/Server/TCPHandler.cpp

@@ -1,7 +1,5 @@
 #include <iomanip>
 
-#include <boost/bind.hpp>
-
 #include <Poco/Net/NetException.h>
 
 #include <Yandex/Revision.h>
@@ -85,9 +83,7 @@ void TCPHandler::runImpl()
 
 	sendHello();
 
-	connection_context.setProgressCallback([this] (const size_t rows, const size_t bytes) {
-		return this->updateProgress(rows, bytes);
-	});
+	connection_context.setProgressCallback([this] (const Progress & value) { return this->updateProgress(value); });
 
 	while (1)
 	{
@@ -125,6 +121,7 @@ void TCPHandler::runImpl()
 			/// Очищаем, так как, получая данные внешних таблиц, мы получили пустой блок.
 			/// А значит, stream помечен как cancelled и читать из него нельзя.
 			state.block_in = nullptr;
+			state.maybe_compressed_in = nullptr;	/// Для более корректного учёта MemoryTracker-ом.
 
 			/// Обрабатываем Query
 			state.io = executeQuery(state.query, query_context, false, state.stage);
@@ -286,7 +283,7 @@ void TCPHandler::processOrdinaryQuery()
 				}
 				else
 				{
-					if (state.rows_processed && after_send_progress.elapsed() / 1000 >= query_context.getSettingsRef().interactive_delay)
+					if (state.progress.rows && after_send_progress.elapsed() / 1000 >= query_context.getSettingsRef().interactive_delay)
 					{
 						/// Прошло некоторое время и есть прогресс.
 						after_send_progress.restart();
@@ -691,21 +688,17 @@ void TCPHandler::sendEndOfStream()
 }
 
 
-void TCPHandler::updateProgress(size_t rows, size_t bytes)
+void TCPHandler::updateProgress(const Progress & value)
 {
-	__sync_fetch_and_add(&state.rows_processed, rows);
-	__sync_fetch_and_add(&state.bytes_processed, bytes);
+	state.progress.incrementPiecewiseAtomically(value);
 }
 
 
 void TCPHandler::sendProgress()
 {
-	size_t rows_processed = __sync_fetch_and_and(&state.rows_processed, 0);
-	size_t bytes_processed = __sync_fetch_and_and(&state.bytes_processed, 0);
-
 	writeVarUInt(Protocol::Server::Progress, *out);
-	Progress progress(rows_processed, bytes_processed);
-	progress.write(*out);
+	Progress increment = state.progress.fetchAndResetPiecewiseAtomically();
+	increment.write(*out, client_revision);
 	out->next();
 }
 

dbms/src/Server/TCPHandler.h

@@ -23,8 +23,8 @@ struct QueryState
 	/// Идентификатор запроса.
 	String query_id;
 
-	QueryProcessingStage::Enum stage;
-	Protocol::Compression::Enum compression;
+	QueryProcessingStage::Enum stage = QueryProcessingStage::Complete;
+	Protocol::Compression::Enum compression = Protocol::Compression::Disable;
 
 	/// Откуда читать данные для INSERT-а.
 	SharedPtr<ReadBuffer> maybe_compressed_in;
@@ -40,24 +40,29 @@ struct QueryState
 	BlockIO io;
 
 	/// Отменен ли запрос
-	bool is_cancelled;
+	bool is_cancelled = false;
 	/// Пустой или нет
-	bool is_empty;
+	bool is_empty = true;
 	/// Данные были отправлены.
-	bool sent_all_data;
+	bool sent_all_data = false;
 	/// Запрос на вставку или нет.
-	bool is_insert;
+	bool is_insert = false;
 
 	/// Для вывода прогресса - разница после предыдущей отправки прогресса.
-	volatile size_t rows_processed;
-	volatile size_t bytes_processed;
+	Progress progress;
 
 
-	QueryState() : query_id(""), stage(QueryProcessingStage::Complete), compression(Protocol::Compression::Disable),
-		is_cancelled(false), is_empty(true), sent_all_data(false), is_insert(false), rows_processed(0), bytes_processed(0) {}
-	
 	void reset()
 	{
+		/** process_list_entry также включает/выключает учёт памяти MemoryTracker-ом.
+		  * Члены maybe_compressed_in, block_in, maybe_compressed_out, block_out
+		  *  могли быть инициализированы до io, и выделенная в них память могла не быть учтена MemoryTracker-ом.
+		  * Если эти члены будут уничтожены раньше, то освобождение памяти будет учтено MemoryTracker-ом,
+		  *  и вычисленный расход памяти может оказаться отрицательным (это не проблема, но некрасиво).
+		  * Поэтому, сначала уничтожим process_list_entry.
+		  */
+		io.process_list_entry = nullptr;
+
 		*this = QueryState();
 	}
 
@@ -133,7 +138,7 @@ private:
 	bool isQueryCancelled();
 
 	/// Эта функция вызывается из разных потоков.
-	void updateProgress(size_t rows, size_t bytes);
+	void updateProgress(const Progress & value);
 
 	/// Вывести информацию о скорости выполнения SELECT запроса.
 	void logProfileInfo(Stopwatch & watch, IBlockInputStream & in);

dbms/src/Storages/MergeTree/MergeTreeDataMerger.cpp

@@ -350,11 +350,13 @@ MergeTreeData::DataPartPtr MergeTreeDataMerger::mergeParts(
 		auto input = stdext::make_unique<MergeTreeBlockInputStream>(
 			data.getFullPath() + parts[i]->name + '/', DEFAULT_MERGE_BLOCK_SIZE, union_column_names, data,
 			parts[i], ranges, false, nullptr, "");
-		input->setProgressCallback([&merge_entry, rows_total] (const std::size_t rows, const std::size_t bytes) {
-			const auto new_rows_read = __sync_add_and_fetch(&merge_entry->rows_read, rows);
-			merge_entry->progress = static_cast<Float64>(new_rows_read) / rows_total;
-			__sync_add_and_fetch(&merge_entry->bytes_read_uncompressed, bytes);
-		});
+
+		input->setProgressCallback([&merge_entry, rows_total] (const Progress & value)
+			{
+				const auto new_rows_read = __sync_add_and_fetch(&merge_entry->rows_read, value.rows);
+				merge_entry->progress = static_cast<Float64>(new_rows_read) / rows_total;
+				__sync_add_and_fetch(&merge_entry->bytes_read_uncompressed, value.bytes);
+			});
 
 		src_streams.push_back(new ExpressionBlockInputStream(input.release(), data.getPrimaryExpression()));
 		sum_rows_approx += parts[i]->size * data.index_granularity;

dbms/src/Storages/StorageBuffer.cpp

@@ -0,0 +1,432 @@
+#include <DB/Interpreters/InterpreterSelectQuery.h>
+#include <DB/Interpreters/InterpreterInsertQuery.h>
+#include <DB/Interpreters/InterpreterAlterQuery.h>
+#include <DB/Storages/StorageBuffer.h>
+#include <DB/Parsers/ASTInsertQuery.h>
+#include <Poco/Ext/ThreadNumber.h>
+
+#include <statdaemons/ext/range.hpp>
+
+
+namespace DB
+{
+
+
+StoragePtr StorageBuffer::create(const std::string & name_, NamesAndTypesListPtr columns_, Context & context_,
+	size_t num_shards_, const Thresholds & min_thresholds_, const Thresholds & max_thresholds_,
+	const String & destination_database_, const String & destination_table_)
+{
+	return (new StorageBuffer{
+		name_, columns_, context_, num_shards_, min_thresholds_, max_thresholds_, destination_database_, destination_table_})->thisPtr();
+}
+
+
+StorageBuffer::StorageBuffer(const std::string & name_, NamesAndTypesListPtr columns_, Context & context_,
+	size_t num_shards_, const Thresholds & min_thresholds_, const Thresholds & max_thresholds_,
+	const String & destination_database_, const String & destination_table_)
+	: name(name_), columns(columns_), context(context_),
+	num_shards(num_shards_), buffers(num_shards_),
+	min_thresholds(min_thresholds_), max_thresholds(max_thresholds_),
+	destination_database(destination_database_), destination_table(destination_table_),
+	no_destination(destination_database.empty() && destination_table.empty()),
+	log(&Logger::get("StorageBuffer (" + name + ")")),
+	flush_thread([this] { flushThread(); })
+{
+}
+
+
+/// Читает из одного буфера (из одного блока) под его mutex-ом.
+class BufferBlockInputStream : public IProfilingBlockInputStream
+{
+public:
+	BufferBlockInputStream(const Names & column_names_, StorageBuffer::Buffer & buffer_)
+		: column_names(column_names_.begin(), column_names_.end()), buffer(buffer_) {}
+
+	String getName() const { return "BufferBlockInputStream"; }
+
+	String getID() const
+	{
+		std::stringstream res;
+		res << "Buffer(" << &buffer;
+
+		for (const auto & name : column_names)
+			res << ", " << name;
+
+		res << ")";
+		return res.str();
+	}
+
+protected:
+	Block readImpl()
+	{
+		Block res;
+
+		if (has_been_read)
+			return res;
+		has_been_read = true;
+
+		std::lock_guard<std::mutex> lock(buffer.mutex);
+
+		if (!buffer.data)
+			return res;
+
+		for (size_t i = 0, size = buffer.data.columns(); i < size; ++i)
+		{
+			auto & col = buffer.data.unsafeGetByPosition(i);
+			if (column_names.count(col.name))
+				res.insert(col);
+		}
+
+		return res;
+	}
+
+private:
+	NameSet column_names;
+	StorageBuffer::Buffer & buffer;
+	bool has_been_read = false;
+};
+
+
+BlockInputStreams StorageBuffer::read(
+	const Names & column_names,
+	ASTPtr query,
+	const Settings & settings,
+	QueryProcessingStage::Enum & processed_stage,
+	size_t max_block_size,
+	unsigned threads)
+{
+	processed_stage = QueryProcessingStage::FetchColumns;
+
+	BlockInputStreams streams_from_dst;
+
+	if (!no_destination)
+		streams_from_dst = context.getTable(destination_database, destination_table)->read(
+			column_names, query, settings, processed_stage, max_block_size, threads);
+
+	BlockInputStreams streams_from_buffers;
+	streams_from_buffers.reserve(num_shards);
+	for (auto & buf : buffers)
+		streams_from_buffers.push_back(new BufferBlockInputStream(column_names, buf));
+
+	/** Если источники из таблицы были обработаны до какой-то не начальной стадии выполнения запроса,
+	  * то тогда источники из буферов надо тоже обернуть в конвейер обработки до той же стадии.
+	  */
+	if (processed_stage > QueryProcessingStage::FetchColumns)
+		for (auto & stream : streams_from_buffers)
+			stream = InterpreterSelectQuery(query, context, processed_stage, 0, stream).execute();
+
+	streams_from_dst.insert(streams_from_dst.end(), streams_from_buffers.begin(), streams_from_buffers.end());
+	return streams_from_dst;
+}
+
+
+static void appendBlock(const Block & from, Block & to)
+{
+	size_t rows = from.rows();
+	for (size_t column_no = 0, columns = to.columns(); column_no < columns; ++column_no)
+	{
+		const IColumn & col_from = *from.getByPosition(column_no).column.get();
+		IColumn & col_to = *to.getByPosition(column_no).column.get();
+
+		if (col_from.getName() != col_to.getName())
+			throw Exception("Cannot append block to another: different type of columns at index " + toString(column_no)
+				+ ". Block 1: " + from.dumpStructure() + ". Block 2: " + to.dumpStructure(), ErrorCodes::BLOCKS_HAS_DIFFERENT_STRUCTURE);
+
+		for (size_t row_no = 0; row_no < rows; ++row_no)
+			col_to.insertFrom(col_from, row_no);
+	}
+}
+
+
+class BufferBlockOutputStream : public IBlockOutputStream
+{
+public:
+	BufferBlockOutputStream(StorageBuffer & storage_) : storage(storage_) {}
+
+	void write(const Block & block)
+	{
+		if (!block)
+			return;
+
+		size_t rows = block.rowsInFirstColumn();
+		if (!rows)
+			return;
+
+		StoragePtr destination;
+		if (!storage.no_destination)
+		{
+			destination = storage.context.tryGetTable(storage.destination_database, storage.destination_table);
+
+			/// Проверяем структуру таблицы.
+			try
+			{
+				destination->check(block, true);
+			}
+			catch (Exception & e)
+			{
+				e.addMessage("(when looking at destination table " + storage.destination_database + "." + storage.destination_table + ")");
+				throw;
+			}
+		}
+
+		size_t bytes = block.bytes();
+
+		/// Если блок уже превышает максимальные ограничения, то пишем минуя буфер.
+		if (rows > storage.max_thresholds.rows || bytes > storage.max_thresholds.bytes)
+		{
+			if (!storage.no_destination)
+			{
+				LOG_TRACE(storage.log, "Writing block with " << rows << " rows, " << bytes << " bytes directly.");
+				storage.writeBlockToDestination(block, destination);
+ 			}
+			return;
+		}
+
+		/// Распределяем нагрузку по шардам по номеру потока.
+		const auto start_shard_num = Poco::ThreadNumber::get() % storage.num_shards;
+
+		/// Перебираем буферы по кругу, пытаясь заблокировать mutex. Не более одного круга.
+		auto shard_num = start_shard_num;
+		size_t try_no = 0;
+		for (; try_no != storage.num_shards; ++try_no)
+		{
+			std::unique_lock<std::mutex> lock(storage.buffers[shard_num].mutex, std::try_to_lock_t());
+			if (lock.owns_lock())
+			{
+				insertIntoBuffer(block, storage.buffers[shard_num], std::move(lock));
+				break;
+			}
+
+			++shard_num;
+			if (shard_num == storage.num_shards)
+				shard_num = 0;
+		}
+
+		/// Если так и не удалось ничего сразу заблокировать, то будем ждать на mutex-е.
+		if (try_no == storage.num_shards)
+			insertIntoBuffer(block, storage.buffers[start_shard_num], std::unique_lock<std::mutex>(storage.buffers[start_shard_num].mutex));
+	}
+private:
+	StorageBuffer & storage;
+
+	void insertIntoBuffer(const Block & block, StorageBuffer::Buffer & buffer, std::unique_lock<std::mutex> && lock)
+	{
+		if (!buffer.data)
+		{
+			buffer.first_write_time = time(0);
+			buffer.data = block.cloneEmpty();
+		}
+
+		/// Если после вставки в буфер, ограничения будут превышены, то будем сбрасывать буфер.
+		if (storage.checkThresholds(buffer, time(0), block.rowsInFirstColumn(), block.bytes()))
+		{
+			/// Вытащим из буфера блок, заменим буфер на пустой. После этого можно разблокировать mutex.
+			Block block_to_write;
+			buffer.data.swap(block_to_write);
+			buffer.first_write_time = 0;
+			lock.unlock();
+
+			if (!storage.no_destination)
+			{
+				appendBlock(block, block_to_write);
+				storage.writeBlockToDestination(block_to_write,
+					storage.context.tryGetTable(storage.destination_database, storage.destination_table));
+			}
+		}
+		else
+			appendBlock(block, buffer.data);
+	}
+};
+
+
+BlockOutputStreamPtr StorageBuffer::write(ASTPtr query)
+{
+	return new BufferBlockOutputStream(*this);
+}
+
+
+void StorageBuffer::shutdown()
+{
+	shutdown_event.set();
+
+	if (flush_thread.joinable())
+		flush_thread.join();
+
+	optimize();
+}
+
+
+bool StorageBuffer::optimize()
+{
+	for (auto & buf : buffers)
+		flushBuffer(buf, false);
+
+	return true;
+}
+
+
+bool StorageBuffer::checkThresholds(Buffer & buffer, time_t current_time, size_t additional_rows, size_t additional_bytes)
+{
+	time_t time_passed = 0;
+	if (buffer.first_write_time)
+		time_passed = current_time - buffer.first_write_time;
+
+	size_t rows = buffer.data.rowsInFirstColumn() + additional_rows;
+	size_t bytes = buffer.data.bytes() + additional_bytes;
+
+	bool res =
+	       (time_passed > min_thresholds.time && rows > min_thresholds.rows && bytes > min_thresholds.bytes)
+		|| (time_passed > max_thresholds.time || rows > max_thresholds.rows || bytes > max_thresholds.bytes);
+
+	if (res)
+		LOG_TRACE(log, "Flushing buffer with " << rows << " rows, " << bytes << " bytes, age " << time_passed << " seconds.");
+
+	return res;
+}
+
+
+void StorageBuffer::flushBuffer(Buffer & buffer, bool check_thresholds)
+{
+	Block block_to_write;
+	time_t current_time = check_thresholds ? time(0) : 0;
+
+	/** Довольно много проблем из-за того, что хотим блокировать буфер лишь на короткое время.
+	  * Под блокировкой, получаем из буфера блок, и заменяем в нём блок на новый пустой.
+	  * Затем пытаемся записать полученный блок в подчинённую таблицу.
+	  * Если этого не получилось - кладём данные обратно в буфер.
+	  * Замечание: может быть, стоит избавиться от такой сложности.
+	  */
+	{
+		std::lock_guard<std::mutex> lock(buffer.mutex);
+
+		if (check_thresholds && !checkThresholds(buffer, current_time))
+			return;
+
+		buffer.data.swap(block_to_write);
+		buffer.first_write_time = 0;
+	}
+
+	if (no_destination)
+		return;
+
+	try
+	{
+		writeBlockToDestination(block_to_write, context.tryGetTable(destination_database, destination_table));
+	}
+	catch (...)
+	{
+		/// Возвращаем блок на место в буфер.
+
+		std::lock_guard<std::mutex> lock(buffer.mutex);
+
+		if (buffer.data)
+		{
+			/** Так как структура таблицы не изменилась, можно склеить два блока.
+			  * Замечание: остаётся проблема - из-за того, что в разных попытках вставляются разные блоки,
+			  *  теряется идемпотентность вставки в ReplicatedMergeTree.
+			  */
+			appendBlock(block_to_write, buffer.data);
+			buffer.data.swap(block_to_write);
+		}
+
+		if (!buffer.first_write_time)
+			buffer.first_write_time = current_time;
+
+		/// Через некоторое время будет следующая попытка записать.
+		throw;
+	}
+}
+
+
+void StorageBuffer::writeBlockToDestination(const Block & block, StoragePtr table)
+{
+	if (no_destination || !block)
+		return;
+
+	if (!table)
+	{
+		LOG_ERROR(log, "Destination table " << destination_database << "." << destination_table << " doesn't exist. Block of data is discarded.");
+		return;
+	}
+
+	ASTInsertQuery * insert = new ASTInsertQuery;
+	ASTPtr ast_ptr = insert;
+
+	insert->database = destination_database;
+	insert->table = destination_table;
+
+	/** Будем вставлять столбцы, являющиеся пересечением множества столбцов таблицы-буфера и подчинённой таблицы.
+	  * Это позволит поддержать часть случаев (но не все), когда структура таблицы не совпадает.
+	  */
+	Block structure_of_destination_table = table->getSampleBlock();
+	Names columns_intersection;
+	columns_intersection.reserve(block.columns());
+	for (size_t i : ext::range(0, structure_of_destination_table.columns()))
+	{
+		auto dst_col = structure_of_destination_table.unsafeGetByPosition(i);
+		if (block.has(dst_col.name))
+		{
+			if (block.getByName(dst_col.name).type->getName() != dst_col.type->getName())
+			{
+				LOG_ERROR(log, "Destination table " << destination_database << "." << destination_table
+					<< " have different type of column " << dst_col.name << ". Block of data is discarded.");
+				return;
+			}
+
+			columns_intersection.push_back(dst_col.name);
+		}
+	}
+
+	if (columns_intersection.empty())
+	{
+		LOG_ERROR(log, "Destination table " << destination_database << "." << destination_table << " have no common columns with block in buffer. Block of data is discarded.");
+		return;
+	}
+
+	if (columns_intersection.size() != block.columns())
+		LOG_WARNING(log, "Not all columns from block in buffer exist in destination table "
+			<< destination_database << "." << destination_table << ". Some columns are discarded.");
+
+	ASTExpressionList * list_of_columns = new ASTExpressionList;
+	insert->columns = list_of_columns;
+	list_of_columns->children.reserve(columns_intersection.size());
+	for (const String & column : columns_intersection)
+		list_of_columns->children.push_back(new ASTIdentifier(StringRange(), column, ASTIdentifier::Column));
+
+	InterpreterInsertQuery interpreter{ast_ptr, context};
+
+	auto block_io = interpreter.execute();
+	block_io.out->writePrefix();
+	block_io.out->write(block);
+	block_io.out->writeSuffix();
+}
+
+
+void StorageBuffer::flushThread()
+{
+	do
+	{
+		try
+		{
+			optimize();
+		}
+		catch (...)
+		{
+			tryLogCurrentException(__PRETTY_FUNCTION__);
+		}
+	} while (!shutdown_event.tryWait(1000));
+}
+
+
+void StorageBuffer::alter(const AlterCommands & params, const String & database_name, const String & table_name, Context & context)
+{
+	auto lock = lockStructureForAlter();
+
+	/// Чтобы не осталось блоков старой структуры.
+	optimize();
+
+	params.apply(*columns);
+	InterpreterAlterQuery::updateMetadata(database_name, table_name, *columns, context);
+}
+
+}

dbms/src/Storages/StorageFactory.cpp

@@ -11,6 +11,7 @@
 #include <DB/Storages/StorageLog.h>
 #include <DB/Storages/StorageTinyLog.h>
 #include <DB/Storages/StorageMemory.h>
+#include <DB/Storages/StorageBuffer.h>
 #include <DB/Storages/StorageNull.h>
 #include <DB/Storages/StorageMerge.h>
 #include <DB/Storages/StorageMergeTree.h>
@@ -190,6 +191,46 @@ StoragePtr StorageFactory::get(
 		return StorageDistributed::create(
 			table_name, columns, remote_database, remote_table, cluster_name, context, sharding_key, data_path);
 	}
+	else if (name == "Buffer")
+	{
+		/** Buffer(db, table, num_buckets, min_time, max_time, min_rows, max_rows, min_bytes, max_bytes)
+		  *
+		  * db, table - в какую таблицу сбрасывать данные из буфера.
+		  * num_buckets - уровень параллелизма.
+		  * min_time, max_time, min_rows, max_rows, min_bytes, max_bytes - условия вытеснения из буфера.
+		  */
+
+		ASTs & args_func = typeid_cast<ASTFunction &>(*typeid_cast<ASTCreateQuery &>(*query).storage).children;
+
+		if (args_func.size() != 1)
+			throw Exception("Storage Buffer requires 9 parameters: "
+				" destination database, destination table, num_buckets, min_time, max_time, min_rows, max_rows, min_bytes, max_bytes.",
+				ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+		ASTs & args = typeid_cast<ASTExpressionList &>(*args_func.at(0)).children;
+
+		if (args.size() != 9)
+			throw Exception("Storage Buffer requires 9 parameters: "
+				" destination_database, destination_table, num_buckets, min_time, max_time, min_rows, max_rows, min_bytes, max_bytes.",
+				ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+		String destination_database = reinterpretAsIdentifier(args[0], local_context).name;
+		String destination_table 	= typeid_cast<ASTIdentifier &>(*args[1]).name;
+
+		size_t num_buckets = apply_visitor(FieldVisitorConvertToNumber<size_t>(), typeid_cast<ASTLiteral &>(*args[2]).value);
+
+		time_t min_time = apply_visitor(FieldVisitorConvertToNumber<size_t>(), typeid_cast<ASTLiteral &>(*args[3]).value);
+		time_t max_time = apply_visitor(FieldVisitorConvertToNumber<size_t>(), typeid_cast<ASTLiteral &>(*args[4]).value);
+		size_t min_rows = apply_visitor(FieldVisitorConvertToNumber<size_t>(), typeid_cast<ASTLiteral &>(*args[5]).value);
+		size_t max_rows = apply_visitor(FieldVisitorConvertToNumber<size_t>(), typeid_cast<ASTLiteral &>(*args[6]).value);
+		size_t min_bytes = apply_visitor(FieldVisitorConvertToNumber<size_t>(), typeid_cast<ASTLiteral &>(*args[7]).value);
+		size_t max_bytes = apply_visitor(FieldVisitorConvertToNumber<size_t>(), typeid_cast<ASTLiteral &>(*args[8]).value);
+
+		return StorageBuffer::create(
+			table_name, columns, context,
+			num_buckets, {min_time, min_rows, min_bytes}, {max_time, max_rows, max_bytes},
+			destination_database, destination_table);
+	}
 	else if (endsWith(name, "MergeTree"))
 	{
 		/** Движки [Replicated][Summing|Collapsing|Aggregating|]MergeTree  (8 комбинаций)

dbms/src/Storages/StorageLog.cpp

@@ -43,8 +43,8 @@ String LogBlockInputStream::getID() const
 	std::stringstream res;
 	res << "Log(" << storage.getTableName() << ", " << &storage << ", " << mark_number << ", " << rows_limit;
 
-	for (size_t i = 0; i < column_names.size(); ++i)
-		res << ", " << column_names[i];
+	for (const auto & name : column_names)
+		res << ", " << name;
 
 	res << ")";
 	return res.str();

dbms/src/Storages/StorageSystemOne.cpp

@@ -34,9 +34,9 @@ BlockInputStreams StorageSystemOne::read(
 	ColumnWithNameAndType col;
 	col.name = "dummy";
 	col.type = new DataTypeUInt8;
-	col.column = new ColumnConstUInt8(1, 0);
+	col.column = ColumnConstUInt8(1, 0).convertToFullColumn();
 	block.insert(col);
-	
+
 	return BlockInputStreams(1, new OneBlockInputStream(block));
 }
 

dbms/src/Storages/StorageSystemProcesses.cpp

@@ -17,6 +17,7 @@ StorageSystemProcesses::StorageSystemProcesses(const std::string & name_, const
 		{ "elapsed", 		new DataTypeFloat64	},
 		{ "rows_read", 		new DataTypeUInt64	},
 		{ "bytes_read",		new DataTypeUInt64	},
+		{ "total_rows_approx", new DataTypeUInt64 },
 		{ "memory_usage",	new DataTypeUInt64	},
 		{ "query", 			new DataTypeString	},
 		{ "query_id", 		new DataTypeString	}
@@ -42,20 +43,19 @@ BlockInputStreams StorageSystemProcesses::read(
 	ColumnWithNameAndType col_elapsed{new ColumnFloat64, new DataTypeFloat64, "elapsed"};
 	ColumnWithNameAndType col_rows_read{new ColumnUInt64, new DataTypeUInt64, "rows_read"};
 	ColumnWithNameAndType col_bytes_read{new ColumnUInt64, new DataTypeUInt64, "bytes_read"};
+	ColumnWithNameAndType col_total_rows_approx{new ColumnUInt64, new DataTypeUInt64, "total_rows_approx"};
 	ColumnWithNameAndType col_memory_usage{new ColumnUInt64, new DataTypeUInt64, "memory_usage"};
 	ColumnWithNameAndType col_query{new ColumnString, new DataTypeString, "query"};
 	ColumnWithNameAndType col_query_id{new ColumnString, new DataTypeString, "query_id"};
 
 	for (const auto & process : context.getProcessList().get())
 	{
-		const size_t rows_read = process.rows_processed;
-		const size_t bytes_read = process.bytes_processed;
-
 		col_user.column->insert(process.user);
 		col_address.column->insert(process.ip_address.toString());
 		col_elapsed.column->insert(process.watch.elapsedSeconds());
-		col_rows_read.column->insert(rows_read);
-		col_bytes_read.column->insert(bytes_read);
+		col_rows_read.column->insert(process.progress.rows);
+		col_bytes_read.column->insert(process.progress.bytes);
+		col_total_rows_approx.column->insert(process.progress.total_rows);
 		col_memory_usage.column->insert(static_cast<UInt64>(process.memory_tracker.get()));
 		col_query.column->insert(process.query);
 		col_query_id.column->insert(process.query_id);
@@ -67,6 +67,7 @@ BlockInputStreams StorageSystemProcesses::read(
 		col_elapsed,
 		col_rows_read,
 		col_bytes_read,
+		col_total_rows_approx,
 		col_memory_usage,
 		col_query,
 		col_query_id

dbms/src/Storages/StorageTinyLog.cpp

@@ -38,8 +38,8 @@ String TinyLogBlockInputStream::getID() const
 	std::stringstream res;
 	res << "TinyLog(" << storage.getTableName() << ", " << &storage;
 
-	for (size_t i = 0; i < column_names.size(); ++i)
-		res << ", " << column_names[i];
+	for (const auto & name : column_names)
+		res << ", " << name;
 
 	res << ")";
 	return res.str();

dbms/tests/queries/0_stateless/00019_quantiles_totals_distributed.sql

@@ -1 +1 @@
-SELECT quantilesTiming(0.1, 0.5, 0.9)(materialize(dummy)) FROM remote('127.0.0.{1,2}', system, one) GROUP BY 1 WITH TOTALS
+SELECT quantilesTiming(0.1, 0.5, 0.9)(dummy) FROM remote('127.0.0.{1,2}', system, one) GROUP BY 1 WITH TOTALS

dbms/tests/queries/0_stateless/00076_ip_coding_functions.reference

@@ -0,0 +1,18 @@
+1
+1
+1
+1
+1	1
+1	1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1

dbms/tests/queries/0_stateless/00076_ip_coding_functions.sql

@@ -0,0 +1,19 @@
+select IPv4StringToNum('') == 0;
+select IPv4StringToNum(materialize('')) == 0;
+select IPv4StringToNum('not an ip string') == 0;
+select IPv4StringToNum(materialize('not an ip string')) == 0;
+select IPv4StringToNum('127.0.0.1' as p) == (0x7f000001 as n), IPv4NumToString(n) == p;
+select IPv4StringToNum(materialize('127.0.0.1') as p) == (materialize(0x7f000001) as n), IPv4NumToString(n) == p;
+select IPv4NumToString(toUInt32(0)) == '0.0.0.0';
+select IPv4NumToString(materialize(toUInt32(0))) == materialize('0.0.0.0');
+
+select IPv6NumToString(toFixedString('', 16)) == '::';
+select IPv6NumToString(toFixedString(materialize(''), 16)) == materialize('::');
+select IPv6NumToString(IPv6StringToNum('::ffff:127.0.0.1' as p) as n) == p;
+select IPv6NumToString(IPv6StringToNum(materialize('::ffff:127.0.0.1') as p) as n) == p;
+select IPv6NumToString(toFixedString(unhex('20010DB800000003000001FF0000002E'), 16)) == '2001:db8:0:3:0:1ff:0:2e';
+select IPv6NumToString(toFixedString(unhex(materialize('20010DB800000003000001FF0000002E')), 16)) == materialize('2001:db8:0:3:0:1ff:0:2e');
+select IPv6StringToNum('') == toFixedString(materialize(''), 16);
+select IPv6StringToNum(materialize('')) == toFixedString(materialize(''), 16);
+select IPv6StringToNum('not an ip string') == toFixedString(materialize(''), 16);
+select IPv6StringToNum(materialize('not an ip string')) == toFixedString(materialize(''), 16);

libs/libmysqlxx/include/mysqlxx/DateTime.h

@@ -4,6 +4,7 @@
 #include <Yandex/DateLUT.h>
 
 #include <mysqlxx/Date.h>
+#include <iomanip>
 
 
 namespace mysqlxx
@@ -177,12 +178,15 @@ public:
 
 inline std::ostream & operator<< (std::ostream & ostr, const DateTime & datetime)
 {
-	return ostr << datetime.year()
-		<< '-' << (datetime.month() / 10) 	<< (datetime.month() % 10)
+	ostr << std::setfill('0') << std::setw(4) <<  datetime.year();
+
+	ostr << '-' << (datetime.month() / 10) 	<< (datetime.month() % 10)
 		<< '-' << (datetime.day() / 10) 	<< (datetime.day() % 10)
 		<< ' ' << (datetime.hour() / 10) 	<< (datetime.hour() % 10)
 		<< ':' << (datetime.minute() / 10) 	<< (datetime.minute() % 10)
 		<< ':' << (datetime.second() / 10) 	<< (datetime.second() % 10);
+
+	return ostr;
 }
 
 }

libs/libzkutil/include/zkutil/KeeperException.h

@@ -23,6 +23,11 @@ public:
 	const char * className() const throw() { return "zkutil::KeeperException"; }
 	KeeperException * clone() const { return new KeeperException(*this); }
 
+	/// при этих ошибках надо переинициализировать сессию с zookeeper
+	bool isUnrecoverable() const
+	{
+		return code == ZINVALIDSTATE || code == ZSESSIONEXPIRED;
+	}
 	int32_t code;
 
 private:
@@ -30,6 +35,7 @@ private:
 	{
 		ProfileEvents::increment(ProfileEvents::ZooKeeperExceptions);
 	}
+
 };
 
 };

libs/libzkutil/src/ZooKeeper.cpp

@@ -246,7 +246,7 @@ int32_t ZooKeeper::tryCreate(const std::string & path, const std::string & data,
 
 int32_t ZooKeeper::tryCreateWithRetries(const std::string & path, const std::string & data, int32_t mode, std::string & pathCreated, size_t* attempt)
 {
-	return retry([&path, &data, mode, &pathCreated, this] { return tryCreate(path, data, mode, pathCreated); });
+	return retry([&path, &data, mode, &pathCreated, this] { return tryCreate(path, data, mode, pathCreated); }, attempt);
 }