ClickHouse/src/Functions/TransformDateTime64.h

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

142 lines
5.8 KiB
C++
Raw Normal View History

Extended range of DateTime64 to years 1925 - 2238 The Year 1925 is a starting point because most of the timezones switched to saner (mostly 15-minutes based) offsets somewhere during 1924 or before. And that significantly simplifies implementation. 2238 is to simplify arithmetics for sanitizing LUT index access; there are less than 0x1ffff days from 1925. * Extended DateLUTImpl internal LUT to 0x1ffff items, some of which represent negative (pre-1970) time values. As a collateral benefit, Date now correctly supports dates up to 2149 (instead of 2106). * Added a new strong typedef ExtendedDayNum, which represents dates pre-1970 and post 2149. * Functions that used to return DayNum now return ExtendedDayNum. * Refactored DateLUTImpl to untie DayNum from the dual role of being a value and an index (due to negative time). Index is now a different type LUTIndex with explicit conversion functions from DatNum, time_t, and ExtendedDayNum. * Updated DateLUTImpl to properly support values close to epoch start (1970-01-01 00:00), including negative ones. * Reduced resolution of DateLUTImpl::Values::time_at_offset_change to multiple of 15-minutes to allow storing 64-bits of time_t in DateLUTImpl::Value while keeping same size. * Minor performance updates to DateLUTImpl when building month LUT by skipping non-start-of-month days. * Fixed extractTimeZoneFromFunctionArguments to work correctly with DateTime64. * New unit-tests and stateless integration tests for both DateTime and DateTime64.
2020-04-17 13:26:44 +00:00
#pragma once
#include <Core/Types.h>
#include <Core/DecimalFunctions.h>
namespace DB
{
/** Tansform-type wrapper for DateTime64, simplifies DateTime64 support for given Transform.
*
* Depending on what overloads of Transform::execute() are available, when called with DateTime64 value,
* invokes Transform::execute() with either:
* * whole part of DateTime64 value, discarding fractional part (1)
* * DateTime64 value and scale factor (2)
* * DateTime64 broken down to components, result of execute is then re-assembled back into DateTime64 value (3)
*
* Suitable Transform-types are commonly used in Date/DateTime manipulation functions,
Extended range of DateTime64 to years 1925 - 2238 The Year 1925 is a starting point because most of the timezones switched to saner (mostly 15-minutes based) offsets somewhere during 1924 or before. And that significantly simplifies implementation. 2238 is to simplify arithmetics for sanitizing LUT index access; there are less than 0x1ffff days from 1925. * Extended DateLUTImpl internal LUT to 0x1ffff items, some of which represent negative (pre-1970) time values. As a collateral benefit, Date now correctly supports dates up to 2149 (instead of 2106). * Added a new strong typedef ExtendedDayNum, which represents dates pre-1970 and post 2149. * Functions that used to return DayNum now return ExtendedDayNum. * Refactored DateLUTImpl to untie DayNum from the dual role of being a value and an index (due to negative time). Index is now a different type LUTIndex with explicit conversion functions from DatNum, time_t, and ExtendedDayNum. * Updated DateLUTImpl to properly support values close to epoch start (1970-01-01 00:00), including negative ones. * Reduced resolution of DateLUTImpl::Values::time_at_offset_change to multiple of 15-minutes to allow storing 64-bits of time_t in DateLUTImpl::Value while keeping same size. * Minor performance updates to DateLUTImpl when building month LUT by skipping non-start-of-month days. * Fixed extractTimeZoneFromFunctionArguments to work correctly with DateTime64. * New unit-tests and stateless integration tests for both DateTime and DateTime64.
2020-04-17 13:26:44 +00:00
* and should implement static (or const) function with following signatures:
* 1:
* R execute(Int64 whole_value, ... )
* 2:
* R execute(DateTime64 value, Int64 scale_multiplier, ... )
* 3:
* R execute(DecimalUtils::DecimalComponents<DateTime64> components, ... )
*
* Where R could be of arbitrary type, in case of (3) if R is DecimalUtils::DecimalComponents<DateTime64>, result is re-assembed back into DateTime64.
*/
template <typename Transform>
class TransformDateTime64
{
private:
// Detect if Transform::execute is const or static method
// with signature defined by template args (ignoring result type).
template<typename = void, typename... Args>
struct TransformHasExecuteOverload : std::false_type {};
template<typename... Args>
struct TransformHasExecuteOverload<std::void_t<decltype(std::declval<Transform>().execute(std::declval<Args>()...))>, Args...>
: std::true_type {};
template<typename... Args>
static constexpr bool TransformHasExecuteOverload_v = TransformHasExecuteOverload<void, Args...>::value;
public:
static constexpr auto name = Transform::name;
// non-explicit constructor to allow creating from scale value (or with no scale at all), indispensable in some contexts.
TransformDateTime64(UInt32 scale_ = 0) /// NOLINT
Extended range of DateTime64 to years 1925 - 2238 The Year 1925 is a starting point because most of the timezones switched to saner (mostly 15-minutes based) offsets somewhere during 1924 or before. And that significantly simplifies implementation. 2238 is to simplify arithmetics for sanitizing LUT index access; there are less than 0x1ffff days from 1925. * Extended DateLUTImpl internal LUT to 0x1ffff items, some of which represent negative (pre-1970) time values. As a collateral benefit, Date now correctly supports dates up to 2149 (instead of 2106). * Added a new strong typedef ExtendedDayNum, which represents dates pre-1970 and post 2149. * Functions that used to return DayNum now return ExtendedDayNum. * Refactored DateLUTImpl to untie DayNum from the dual role of being a value and an index (due to negative time). Index is now a different type LUTIndex with explicit conversion functions from DatNum, time_t, and ExtendedDayNum. * Updated DateLUTImpl to properly support values close to epoch start (1970-01-01 00:00), including negative ones. * Reduced resolution of DateLUTImpl::Values::time_at_offset_change to multiple of 15-minutes to allow storing 64-bits of time_t in DateLUTImpl::Value while keeping same size. * Minor performance updates to DateLUTImpl when building month LUT by skipping non-start-of-month days. * Fixed extractTimeZoneFromFunctionArguments to work correctly with DateTime64. * New unit-tests and stateless integration tests for both DateTime and DateTime64.
2020-04-17 13:26:44 +00:00
: scale_multiplier(DecimalUtils::scaleMultiplier<DateTime64::NativeType>(scale_))
{}
TransformDateTime64(DateTime64::NativeType scale_multiplier_ = 1) /// NOLINT(google-explicit-constructor)
: scale_multiplier(scale_multiplier_)
{}
Extended range of DateTime64 to years 1925 - 2238 The Year 1925 is a starting point because most of the timezones switched to saner (mostly 15-minutes based) offsets somewhere during 1924 or before. And that significantly simplifies implementation. 2238 is to simplify arithmetics for sanitizing LUT index access; there are less than 0x1ffff days from 1925. * Extended DateLUTImpl internal LUT to 0x1ffff items, some of which represent negative (pre-1970) time values. As a collateral benefit, Date now correctly supports dates up to 2149 (instead of 2106). * Added a new strong typedef ExtendedDayNum, which represents dates pre-1970 and post 2149. * Functions that used to return DayNum now return ExtendedDayNum. * Refactored DateLUTImpl to untie DayNum from the dual role of being a value and an index (due to negative time). Index is now a different type LUTIndex with explicit conversion functions from DatNum, time_t, and ExtendedDayNum. * Updated DateLUTImpl to properly support values close to epoch start (1970-01-01 00:00), including negative ones. * Reduced resolution of DateLUTImpl::Values::time_at_offset_change to multiple of 15-minutes to allow storing 64-bits of time_t in DateLUTImpl::Value while keeping same size. * Minor performance updates to DateLUTImpl when building month LUT by skipping non-start-of-month days. * Fixed extractTimeZoneFromFunctionArguments to work correctly with DateTime64. * New unit-tests and stateless integration tests for both DateTime and DateTime64.
2020-04-17 13:26:44 +00:00
template <typename ... Args>
inline auto NO_SANITIZE_UNDEFINED execute(const DateTime64 & t, Args && ... args) const
Extended range of DateTime64 to years 1925 - 2238 The Year 1925 is a starting point because most of the timezones switched to saner (mostly 15-minutes based) offsets somewhere during 1924 or before. And that significantly simplifies implementation. 2238 is to simplify arithmetics for sanitizing LUT index access; there are less than 0x1ffff days from 1925. * Extended DateLUTImpl internal LUT to 0x1ffff items, some of which represent negative (pre-1970) time values. As a collateral benefit, Date now correctly supports dates up to 2149 (instead of 2106). * Added a new strong typedef ExtendedDayNum, which represents dates pre-1970 and post 2149. * Functions that used to return DayNum now return ExtendedDayNum. * Refactored DateLUTImpl to untie DayNum from the dual role of being a value and an index (due to negative time). Index is now a different type LUTIndex with explicit conversion functions from DatNum, time_t, and ExtendedDayNum. * Updated DateLUTImpl to properly support values close to epoch start (1970-01-01 00:00), including negative ones. * Reduced resolution of DateLUTImpl::Values::time_at_offset_change to multiple of 15-minutes to allow storing 64-bits of time_t in DateLUTImpl::Value while keeping same size. * Minor performance updates to DateLUTImpl when building month LUT by skipping non-start-of-month days. * Fixed extractTimeZoneFromFunctionArguments to work correctly with DateTime64. * New unit-tests and stateless integration tests for both DateTime and DateTime64.
2020-04-17 13:26:44 +00:00
{
/// Type conversion from float to integer may be required.
/// We are Ok with implementation specific result for out of range and denormals conversion.
Extended range of DateTime64 to years 1925 - 2238 The Year 1925 is a starting point because most of the timezones switched to saner (mostly 15-minutes based) offsets somewhere during 1924 or before. And that significantly simplifies implementation. 2238 is to simplify arithmetics for sanitizing LUT index access; there are less than 0x1ffff days from 1925. * Extended DateLUTImpl internal LUT to 0x1ffff items, some of which represent negative (pre-1970) time values. As a collateral benefit, Date now correctly supports dates up to 2149 (instead of 2106). * Added a new strong typedef ExtendedDayNum, which represents dates pre-1970 and post 2149. * Functions that used to return DayNum now return ExtendedDayNum. * Refactored DateLUTImpl to untie DayNum from the dual role of being a value and an index (due to negative time). Index is now a different type LUTIndex with explicit conversion functions from DatNum, time_t, and ExtendedDayNum. * Updated DateLUTImpl to properly support values close to epoch start (1970-01-01 00:00), including negative ones. * Reduced resolution of DateLUTImpl::Values::time_at_offset_change to multiple of 15-minutes to allow storing 64-bits of time_t in DateLUTImpl::Value while keeping same size. * Minor performance updates to DateLUTImpl when building month LUT by skipping non-start-of-month days. * Fixed extractTimeZoneFromFunctionArguments to work correctly with DateTime64. * New unit-tests and stateless integration tests for both DateTime and DateTime64.
2020-04-17 13:26:44 +00:00
if constexpr (TransformHasExecuteOverload_v<DateTime64, decltype(scale_multiplier), Args...>)
{
return wrapped_transform.execute(t, scale_multiplier, std::forward<Args>(args)...);
}
else if constexpr (TransformHasExecuteOverload_v<DecimalUtils::DecimalComponents<DateTime64>, Args...>)
{
auto components = DecimalUtils::splitWithScaleMultiplier(t, scale_multiplier);
const auto result = wrapped_transform.execute(components, std::forward<Args>(args)...);
using ResultType = std::decay_t<decltype(result)>;
if constexpr (std::is_same_v<DecimalUtils::DecimalComponents<DateTime64>, ResultType>)
{
return DecimalUtils::decimalFromComponentsWithMultiplier<DateTime64>(result, scale_multiplier);
}
else
{
return result;
}
}
else
{
const auto components = DecimalUtils::splitWithScaleMultiplier(t, scale_multiplier);
return wrapped_transform.execute(static_cast<Int64>(components.whole), std::forward<Args>(args)...);
}
}
2021-03-04 00:11:40 +00:00
template <typename T, typename ... Args, typename = std::enable_if_t<!std::is_same_v<T, DateTime64>>>
Extended range of DateTime64 to years 1925 - 2238 The Year 1925 is a starting point because most of the timezones switched to saner (mostly 15-minutes based) offsets somewhere during 1924 or before. And that significantly simplifies implementation. 2238 is to simplify arithmetics for sanitizing LUT index access; there are less than 0x1ffff days from 1925. * Extended DateLUTImpl internal LUT to 0x1ffff items, some of which represent negative (pre-1970) time values. As a collateral benefit, Date now correctly supports dates up to 2149 (instead of 2106). * Added a new strong typedef ExtendedDayNum, which represents dates pre-1970 and post 2149. * Functions that used to return DayNum now return ExtendedDayNum. * Refactored DateLUTImpl to untie DayNum from the dual role of being a value and an index (due to negative time). Index is now a different type LUTIndex with explicit conversion functions from DatNum, time_t, and ExtendedDayNum. * Updated DateLUTImpl to properly support values close to epoch start (1970-01-01 00:00), including negative ones. * Reduced resolution of DateLUTImpl::Values::time_at_offset_change to multiple of 15-minutes to allow storing 64-bits of time_t in DateLUTImpl::Value while keeping same size. * Minor performance updates to DateLUTImpl when building month LUT by skipping non-start-of-month days. * Fixed extractTimeZoneFromFunctionArguments to work correctly with DateTime64. * New unit-tests and stateless integration tests for both DateTime and DateTime64.
2020-04-17 13:26:44 +00:00
inline auto execute(const T & t, Args && ... args) const
{
return wrapped_transform.execute(t, std::forward<Args>(args)...);
}
template <typename ... Args>
inline auto NO_SANITIZE_UNDEFINED executeExtendedResult(const DateTime64 & t, Args && ... args) const
{
/// Type conversion from float to integer may be required.
/// We are Ok with implementation specific result for out of range and denormals conversion.
if constexpr (TransformHasExecuteOverload_v<DateTime64, decltype(scale_multiplier), Args...>)
{
return wrapped_transform.executeExtendedResult(t, scale_multiplier, std::forward<Args>(args)...);
}
else if constexpr (TransformHasExecuteOverload_v<DecimalUtils::DecimalComponents<DateTime64>, Args...>)
{
auto components = DecimalUtils::splitWithScaleMultiplier(t, scale_multiplier);
const auto result = wrapped_transform.executeExtendedResult(components, std::forward<Args>(args)...);
using ResultType = std::decay_t<decltype(result)>;
if constexpr (std::is_same_v<DecimalUtils::DecimalComponents<DateTime64>, ResultType>)
{
return DecimalUtils::decimalFromComponentsWithMultiplier<DateTime64>(result, scale_multiplier);
}
else
{
return result;
}
}
else
{
const auto components = DecimalUtils::splitWithScaleMultiplier(t, scale_multiplier);
return wrapped_transform.executeExtendedResult(static_cast<Int64>(components.whole), std::forward<Args>(args)...);
}
}
template <typename T, typename ... Args, typename = std::enable_if_t<!std::is_same_v<T, DateTime64>>>
inline auto executeExtendedResult(const T & t, Args && ... args) const
{
return wrapped_transform.executeExtendedResult(t, std::forward<Args>(args)...);
}
DateTime64::NativeType getScaleMultiplier() const { return scale_multiplier; }
Extended range of DateTime64 to years 1925 - 2238 The Year 1925 is a starting point because most of the timezones switched to saner (mostly 15-minutes based) offsets somewhere during 1924 or before. And that significantly simplifies implementation. 2238 is to simplify arithmetics for sanitizing LUT index access; there are less than 0x1ffff days from 1925. * Extended DateLUTImpl internal LUT to 0x1ffff items, some of which represent negative (pre-1970) time values. As a collateral benefit, Date now correctly supports dates up to 2149 (instead of 2106). * Added a new strong typedef ExtendedDayNum, which represents dates pre-1970 and post 2149. * Functions that used to return DayNum now return ExtendedDayNum. * Refactored DateLUTImpl to untie DayNum from the dual role of being a value and an index (due to negative time). Index is now a different type LUTIndex with explicit conversion functions from DatNum, time_t, and ExtendedDayNum. * Updated DateLUTImpl to properly support values close to epoch start (1970-01-01 00:00), including negative ones. * Reduced resolution of DateLUTImpl::Values::time_at_offset_change to multiple of 15-minutes to allow storing 64-bits of time_t in DateLUTImpl::Value while keeping same size. * Minor performance updates to DateLUTImpl when building month LUT by skipping non-start-of-month days. * Fixed extractTimeZoneFromFunctionArguments to work correctly with DateTime64. * New unit-tests and stateless integration tests for both DateTime and DateTime64.
2020-04-17 13:26:44 +00:00
private:
DateTime64::NativeType scale_multiplier = 1;
Transform wrapped_transform = {};
};
}