LUMIERA.clone/src/lib/time/time.cpp
Ichthyostega 27089550d2 clean-up: relocate frame quantisation operations
The `FixedFrameQuantiser` relied on three functions from the raw-time handling library.
Since this (and NTSC drop-frame) are the only usages, these functions
can be relocated into the implemntation translation unit `lib/time/quantiser.cpp`

On closer inspection, this reveals some room for improvements:
Instead of relying on raw-computation functions written in C,
we could rather revert the dependency and express these computations
in terms of our Time-entities, which are written in C++, are much more
systematic and provide consistency checks and protection against numeric
overflow, all integrated with linear arithmetic and concise notation.

After performing these rearrangements,
most of the functions can be collapsed into ''almost nothing''.

This was taken as opportunity to re-check and improve the remaining
implementation core of the `FixedFrameQuantiser` -- the handling of
extreme corner cases can be much improved, now representing the
"grid-local time" as `Offset`, which doubles the possible value range.

The reworked unit test shows that, with this change, now the limitation
happens prior to quantisation, meaning that we always get a grid-aligned
result, even in the most extreme corner cases.
2025-05-28 23:49:01 +02:00

468 lines
13 KiB
C++
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/*
Time - Lumiera time handling foundation
Copyright (C)
2008, Christian Thaeter <ct@pipapo.org>
2010, Stefan Kangas <skangas@skangas.se>
2011, Hermann Vosseler <Ichthyostega@web.de>
  **Lumiera** is free software; you can redistribute it and/or modify it
  under the terms of the GNU General Public License as published by the
  Free Software Foundation; either version 2 of the License, or (at your
  option) any later version. See the file COPYING for further details.
* *****************************************************************/
/** @file time.cpp
** Lumiera time handling core implementation unit.
** This translation unit generates code for the Lumiera internal time wrapper,
** based on 64bit integral µ-tick values, associated constants, marker classes
** for the derived time entities (TimeVar, Offset, Duration, TimeSpan, FrameRate)
** and for the basic time and frame rate conversion functions.
**
** Client code includes either time.h (for basics and conversion functions)
** or timevalue.hpp (for the time entities), timequant.hpp for grid aligned
** time values or timecode.hpp
**
** @see Time
** @see TimeValue
** @see Grid
** @see TimeValue_test
** @see QuantiserBasics_test
**
*/
#include "lib/error.hpp"
#include "lib/time.h"
#include "lib/time/timevalue.hpp"
#include "lib/rational.hpp"
#include "lib/util-quant.hpp"
#include "lib/format-string.hpp"
#include "lib/util.hpp"
#include <math.h>
#include <limits>
#include <string>
#include <sstream>
#include <boost/rational.hpp>
#include <boost/lexical_cast.hpp>
using std::string;
using util::limited;
using util::floordiv;
using lib::time::FSecs;
using lib::time::FrameRate;
using boost::rational_cast;
using boost::lexical_cast;
#undef MAX
#undef MIN
namespace error = lumiera::error;
namespace lib {
namespace meta {
extern const std::string FAILURE_INDICATOR;
}
namespace time {
const raw_time_64 TimeValue::SCALE = 1'000'000;
/** @note the allowed time range is explicitly limited to help overflow protection */
const Time Time::MAX ( TimeValue::buildRaw_(+std::numeric_limits<raw_time_64>::max() / 30) );
const Time Time::MIN ( TimeValue::buildRaw_(-_raw(Time::MAX) ) );
const Time Time::ZERO;
const Time Time::ANYTIME(Time::MIN);
const Time Time::NEVER (Time::MAX);
const Offset Offset::ZERO (Time::ZERO);
const FSecs FSEC_MAX{std::numeric_limits<int64_t>::max() / lib::time::TimeValue::SCALE};
Literal DIAGNOSTIC_FORMAT{"%s%01d:%02d:%02d.%03d"};
/** scale factor _used locally within this implementation header_.
* TimeValue::SCALE (µ-ticks, i.e. 1e6) is the correct factor or dividend when using
* raw_time_64 for display on a scale with seconds. Since we want to use milliseconds,
* we need to multiply or divide by 1000 to get correct results. */
#define TIME_SCALE_MS (lib::time::TimeValue::SCALE / 1000)
/** convenience constructor to build an
* internal Lumiera Time value from the usual parts
* of an sexagesimal time specification. Arbitrary integral
* values are acceptable and will be summed up accordingly.
* The minute and hour part can be omitted.
* @warning internal Lumiera time values refer to an
* implementation dependent time origin/scale.
* The given value will be used as-is, without
* any further adjustments.
*/
Time::Time ( long millis
, uint secs
, uint mins
, uint hours
)
: TimeValue(lumiera_build_time (millis,secs,mins,hours))
{ }
/** convenience constructor to build an Time value
* from a fraction of seconds, given as rational number.
* An example would be to the time unit of a framerate.
*/
Time::Time (FSecs const& fractionalSeconds)
: TimeValue(lumiera_rational_to_time (fractionalSeconds))
{ }
Offset::Offset (FSecs const& delta_in_secs)
: TimeValue{buildRaw_(symmetricLimit (lumiera_rational_to_time (delta_in_secs)
,Duration::MAX))}
{ }
/** @note recommendation is to use TCode for external representation
* @remarks this is the most prevalent internal diagnostics display
* of any "time-like" value, it is meant to be compact. */
TimeValue::operator string() const
{
raw_time_64 time = t_;
int64_t millis, seconds;
bool negative = (time < 0);
if (negative) time = -time;
time /= TIME_SCALE_MS;
millis = time % 1000;
seconds = time / 1000;
return string (negative ? "-" : "")
+ (seconds>0 or time==0? lexical_cast<string> (seconds)+"s" : "")
+ (millis>0? lexical_cast<string> (millis)+"ms" : "")
;
}
/** display an internal Lumiera Time value
* for diagnostic purposes or internal reporting.
* Format is `-hh:mm:ss.mss`
* @warning internal Lumiera time values refer to an
* implementation dependent time origin/scale.
* @return string rendering of the actual, underlying
* implementation value, as `h:m:s:ms`
*/
Time::operator string() const
{
raw_time_64 time = t_;
int millis, seconds, minutes, hours;
bool negative = (time < 0);
if (negative)
time = -time;
time /= TIME_SCALE_MS;
millis = time % 1000;
time /= 1000;
seconds = time % 60;
time /= 60;
minutes = time % 60;
time /= 60;
hours = time;
return util::_Fmt{string(DIAGNOSTIC_FORMAT)}
% (negative? "-":"")
% hours
% minutes
% seconds
% millis;
}
Offset::operator string() const
{
return (t_< 0? "" : "")
+ TimeValue::operator string();
}
Duration::operator string() const
{
return ""+TimeValue::operator string()+"";
}
TimeSpan::operator string() const
{
return string (start())
+ string (duration());
}
namespace {
template<typename RAT>
string
renderFraction (RAT const& frac, Literal postfx) noexcept
try {
std::ostringstream buffer;
if (1 == frac.denominator() or 0 == frac.numerator())
buffer << frac.numerator() << postfx;
else
buffer << frac <<postfx;
return buffer.str();
}
catch(...)
{ return meta::FAILURE_INDICATOR; }
}
/** visual framerate representation (for diagnostics) */
FrameRate::operator string() const
{
return renderFraction (*this, "FPS");
}
/** @internal backdoor to sneak in a raw time value
* bypassing any normalisation and limiting */
TimeValue
TimeValue::buildRaw_ (raw_time_64 raw)
{
return reinterpret_cast<TimeValue const&> (raw);
}
/** predefined constant for PAL framerate */
const FrameRate FrameRate::PAL (25);
const FrameRate FrameRate::NTSC (30000,1001);
const FrameRate FrameRate::STEP (1);
const FrameRate FrameRate::HALTED (1,std::numeric_limits<int>::max());
/** @return time span of one frame of this rate,
* cast into internal Lumiera time scale */
Duration
FrameRate::duration() const
{
if (0 == *this)
throw error::Logic ("Impossible to quantise to an zero spaced frame grid"
, error::LUMIERA_ERROR_BOTTOM_VALUE);
return Duration (1, *this);
}
/** a rather arbitrary safety limit imposed on internal numbers used to represent a frame rate.
* @remark rational numbers bear the danger to overflow for quite ordinary computations;
* we stay away from the absolute maximum by an additional safety margin of 1/1000.
*/
const uint RATE_LIMIT{std::numeric_limits<uint>::max() / 1024};
/**
* @internal helper to work around the limitations of `uint`.
* @return a fractional number approximating the floating-point spec.
* @todo imposing a quite coarse limitation. If this turns out to be
* a problem: we can do better, use lib::reQuant (rational.hpp)
*/
boost::rational<uint>
__framerate_approximation (double fps)
{
const double UPPER_LIMIT = int64_t(RATE_LIMIT*1024) << 31;
const int64_t HAZARD = util::ilog2(RATE_LIMIT);
double doo = limited (1.0, fabs(fps) * RATE_LIMIT + 0.5, UPPER_LIMIT);
int64_t boo(doo);
util::Rat quantised{boo
,int64_t(RATE_LIMIT)
};
int64_t num = quantised.numerator();
int64_t toxic = util::ilog2(abs(num));
toxic = util::max (0, toxic - HAZARD);
int64_t base = quantised.denominator();
if (toxic)
{
base = util::max (base >> toxic, 1);
num = util::reQuant (num, quantised.denominator(), base);
}
return {limited (1u, num, RATE_LIMIT)
,limited (1u, base, RATE_LIMIT)
};
}
/**
* @internal helper calculate the _count per time span_ approximately,
* to the precision possible to represent as fractional `uint`.
*/
boost::rational<uint>
__framerate_approximation (size_t cnt, Duration timeReference)
{
boost::rational<uint64_t> quot{cnt, _raw(timeReference)};
if (quot.denominator() < RATE_LIMIT
and quot.numerator() < RATE_LIMIT*1024/1e6)
return {uint(quot.numerator()) * uint(Time::SCALE)
,uint(quot.denominator())
};
// precise computation can not be handled numerically...
return __framerate_approximation (rational_cast<double>(quot) * Time::SCALE);
}
/** @internal stretch offset by a possibly fractional factor, and quantise into raw (micro tick) grid */
Offset
Offset::stretchedByRationalFactor (boost::rational<int64_t> factor) const
{
boost::rational<int64_t> distance (this->t_);
distance *= factor;
raw_time_64 microTicks = floordiv (distance.numerator(), distance.denominator());
return Offset{buildRaw_(microTicks)};
}
/** @warning loss of precision on large time values beyond double mantissa length `2^52 ≈ 4.5e15` */
Offset
Offset::stretchedByFloatFactor (double factor) const
{
double distance(this->t_);
distance *= factor;
raw_time_64 microTicks = floor (distance);
return Offset{buildRaw_(microTicks)};
}
/** offset by the given number of frames. */
Offset::Offset (FrameCnt count, FrameRate const& fps)
: TimeValue{buildRaw_(
count? (count<0? -1:+1) * lumiera_framecount_to_time (::abs(count), fps)
:_raw(Duration::NIL))}
{ }
/** constant to indicate "no duration" */
const Duration Duration::NIL {Time::ZERO};
/** maximum possible temporal extension */
const Duration Duration::MAX = []{
auto maxDelta {Time::MAX - Time::MIN};
// bypass limit check, which requires Duration::MAX
return reinterpret_cast<Duration const&> (maxDelta);
}();
const TimeSpan TimeSpan::ALL {Time::MIN, Duration::MAX};
}} // namespace lib::Time
namespace util {
string
StringConv<lib::time::FSecs, void>::invoke (lib::time::FSecs val) noexcept
{
return lib::time::renderFraction (val, "sec");
}
} // namespace util
/* ===== implementation of the C API functions ===== */
/// @todo this utility function could be factored out into a `FSecs` or `RSec` class ///////////////////////TICKET #1262
raw_time_64
lumiera_rational_to_time (FSecs const& fractionalSeconds)
{
// avoid numeric wrap from values not representable as 64bit µ-ticks
if (abs(fractionalSeconds) > lib::time::FSEC_MAX)
return (fractionalSeconds < 0? -1:+1)
* std::numeric_limits<int64_t>::max();
return raw_time_64(util::reQuant (fractionalSeconds.numerator()
,fractionalSeconds.denominator()
,lib::time::TimeValue::SCALE
));
}
raw_time_64
lumiera_framecount_to_time (uint64_t frameCount, FrameRate const& fps)
{
// convert to 64bit
boost::rational<uint64_t> framerate (fps.numerator(), fps.denominator());
return rational_cast<raw_time_64> (lib::time::TimeValue::SCALE * frameCount / framerate);
}
raw_time_64
lumiera_build_time(long millis, uint secs, uint mins, uint hours)
{
raw_time_64 time = millis
+ 1000 * secs
+ 1000 * 60 * mins
+ 1000 * 60 * 60 * hours;
time *= TIME_SCALE_MS;
return time;
}
raw_time_64
lumiera_build_time_fps (uint fps, uint frames, uint secs, uint mins, uint hours)
{
raw_time_64 time = 1000LL * frames/fps
+ 1000 * secs
+ 1000 * 60 * mins
+ 1000 * 60 * 60 * hours;
time *= TIME_SCALE_MS;
return time;
}
int
lumiera_time_hours (raw_time_64 time)
{
return time / TIME_SCALE_MS / 1000 / 60 / 60;
}
int
lumiera_time_minutes (raw_time_64 time)
{
return (time / TIME_SCALE_MS / 1000 / 60) % 60;
}
int
lumiera_time_seconds (raw_time_64 time)
{
return (time / TIME_SCALE_MS / 1000) % 60;
}
int
lumiera_time_millis (raw_time_64 time)
{
return (time / TIME_SCALE_MS) % 1000;
}
int
lumiera_time_frames (raw_time_64 time, uint fps)
{
REQUIRE (fps < uint(std::numeric_limits<int>::max()));
return floordiv<int> (lumiera_time_millis(time) * int(fps), TIME_SCALE_MS);
}
namespace lib {
namespace time { ////////////////////////////////////////////////////////////////////////////////////////////TICKET #1259 : move all calculation functions into a C++ namespace
}} // lib::time