Rand Stats

Physics::Error

zef:p6steve

raku-Physics-Error

some code to handle physical measurement errors (nothing to do with programming errors!)

Build Status

Instructions

Installs automatically withzef --verbose install Physics-Measure uninstall with, zef uninstall Physics::Measure and zef uninstall Physics::Error

Context

In wikipedia, the general topic is https://en.wikipedia.org/wiki/Propagation_of_uncertainty

this module is not intended for use in mission critical applications

Synopsis

Take a look at your keyboard... there's probably a '±' key?

Physics::Error works with the Physics::Measure and Physics::Unit modules to do this:

use Physics::Measure :ALL;

my $x1 = 12.5nm ± 1;                                                    #SI units as raku postfix operators
my $x2 = Length.new(value => 12.5, units => 'nm', error => '4.3%');     #standard raku .new syntax
my $x3 = ♎️ '12.5 ft ±0.5';                                              #libra prefix shorthand

# Error values are included in Measures when output
say ~$x1;                                       #12.5nm ±4% or 12.5nm ±1

# They can be accessed directly via the .error object
say $x1.error.absolute;                         #1
say $x1.error.relative;                         #0.08
say $x1.error.relative.^name;                   #Rat
say $x1.error.percent;                          #8%

All of the main Measure capabilities work as expected...

# Unit conversions and normalization
my $y = Length.new(value => 12.5e2, units => 'nm', error => '4.3%');
say ~$y;                                        #1250nm ±53.75
say ~$y.in('mm');                               #0.00125mm ±0.0000538
say ~$y.norm;                                   #1.25μm ±0.05375

my $t = Time.new(value => 10, units => 'ms', error => 0.2);
say ~( 17 / $t );                               #1700Hz ±34

# Measure math adjusts units and error automagically
say ~( $y / $t );                               #0.000125m/s ±0.000007875e0

# works with add, subtract, multiply, divide, power and root
# add & subtract add absolute error
# multiply & divide add relative error
# power and root multiply relative error by power

my Length $w = Length.new(value => 10, units => 'm', error => '2%');
my $z = $w ** 3;  say ~$z;                      #1000m^3 ±60
$z = $z ** <1/3>; say ~$z;                      #10m ±2.00e-01

# As do Measure cmp operators (to within error limits)
say $w cmp $y;                                  #More

Controls for Error output format and rounding of percentage errors (here with default values). These only act on the .Str output rendering and leave the .error.absolute "truth" untouched.

$Physics::Error::default   = 'absolute';    #default error output [absolute|percent]
$Physics::Error::round-per = 0.001;         #control rounding of percent

Two ways for Measure output precision control. These only act on the .Str output rendering and leave the Measure .value and .error.absolute "truth" untouched.

Automagic

This option uses .error.denorm to right shift the error value and align to the mantissa precision of the measure value. The number of significant digits in the error is then used to round the measure value.

# Here's the mass of the electron Em in action...
my \Em = 9.109_383_701_5e-31kg ±0.000_000_002_8e-31;

say ~Em;                                        #9.1093837015e-31kg ±0.0000000028e-31
say Em.error.absolute;                          #2.8e-40
say Em.error.relative;                          #3.0737534961217373e-10
say Em.error.relative.^name;                    #Num
say Em.error.percent;                           #0%

Manual

Manual precision can be set - this overrides the automagic behaviour.

$Physics::Measure::round-val = 0.01;

my $c = ♎️ '299792458 m/s';
my $ℎ = ♎️ '6.626070015e-34 J.s';

my \λ = 2.5nm; 
is ~λ, '2.5nm',									'~λ';

my \ν = $c / λ;  
is ~ν.norm, '119.92PHz',						'~ν.norm';

my \Ep = $ℎ * ν;  
is ~Ep.norm, '79.46aJ',						    '~Ep.norm';

Physics::Error supports the three use cases for making Measure objects with value, units & error as outlined in the Physics::Measure README.md. The formats are dissected below:

Option 1: Postfix Operator Syntax (SI Units)

my Length $x = 12.5nm ± 10%;
   ------ -- - ------ - ---
      |    | |   |  | |  |
      |    | |   |  | |  > Rat percent error [or '±4.2%' Rat relative error]
      |    | |   |  | |
      |    | |   |  | > ± symbol as custom raku infix operator
      |    | |   |  |
      |    | |   |  > 'nm' Unit constructor as custom raku postfix operator (no ws)
      |    | |   |
      |    | |   > Real number
      |    | |
      |    | > assignment of new Object from postfix rhs
      |    |
      |    > a scalar variable
      |
      > Type (Length is Measure) ... can be omitted

Option 2: Object Constructor Syntax

my Length $x = Length.new(value => 12.5, units => 'nm', error => [0.5|'4.3%']);
say ~$x; #42 ±4.2nanometre

Option 3: Libra Shorthand Syntax

my Length $x = ♎️ '12.5 nm ±0.05';
   ------ --   --  ---- -- -----
      |    |    |   |   |  |  |
      |    |    |   |   |  |  |
      |    |    |   |   |  |  >  Real absolute error [or '±4.2%' Rat relative error]
      |    |    |   |   |  |
      |    |    |   |   |  > '±' symbol as custom raku prefix operator
      |    |    |   |   |
      |    |    |   |   > Str units (of Type Length is Measure)
      |    |    |   |
      |    |    |   > Real number
      |    |    |
      |    |    > parse rhs string, construct object and assign to lhs (♎️ <-- custom <libra> operator)
      |    |
      |    > a scalar variable
      |
      > Type (Length is Measure) ... can be omitted

Help Wanted

Over time I imagine an eco-system of equation parsing / pde plugins and machine calibration matrices - feel free to continue the journey in this direction with a pull request!