MetroloPy

tools for dealing with physical quantities: uncertainty propagation and unit conversion

MetroloPy requires Python 3.10 or later and depends on NumPy, SciPy, pandas, matplotlib and ipython. It looks best in a Jupyter Notebook.

Install MetroloPy with pip install metrolopy or conda install -c conda-forge metrolopy. Alternatively, add it to your uv or pixi project.

Physical quantities can then be represented in Python as gummy objects with an uncertainty and (or) a unit:

>>> import metrolopy as uc
>>> a = uc.gummy(1.2345,u=0.0234,unit='cm')
>>> a
1.234(23) cm

>>> b = uc.gummy(3.034,u=0.174,unit='mm')
>>> f = uc.gummy(uc.UniformDist(center=0.9345,half_width=0.096),unit='N')
>>> p = f/(a*b)
>>> p
2.50(21) N/cm2

>>> p.unit = 'kPa'
>>> p.uunit = '%'
>>> p
25.0 kPa ± 8.5%

MetroloPy can do much more including Monte-Carlo uncertainty propagation, generating uncertainty budget tables, and curve fitting. It can also handle expanded uncertainties, degrees of freedom, correlated quantities, and complex valued quantities. Also gummys work with many numpy functions with no wrapping.

See:

• a tutorial (or download the tutorial as Jupyter notebook)
• the documentation
• the issues page on GitHub
• a list of the units built into MetroloPy
• a list of the physical constants built into MetroloPy

new in version 1.1.0

• The continuous and discrete distributions defined in scipy.stats can now be used directly used with gummys.

• The DistFit class for fitting distributions and the DoF class for descibing quantities drawn from the same underlying distribution have been added.

• The legacy numpy.random.RandomState random number generator has been replaced with the newer numpy.random.Generator for Monte-Carlo uncertainty propagation. The scipy.optimize.leastsq function has been replaced with the newer scipy.optimize.least_squares function as the solver for nonlinear least squares fitting. And the depreciated scipy.odr package has been replaced with odrpack for orthogonal distance regression.

• A class properties has been added to gummy to control the separator between the digit groupings when displaying long numbers.

• Lazy loading for gummy module components has been implemented and added lazy_loader as a dependancy.

• Fixed issues that affected the gummy.apply and gummy.napply method when broadcasing over arguments and and applying functions that have array like return values.

• Fixed an number of bugs in the fitting module.

new in version 1.0.0

• The calculation of effective degrees of freedom has been improved. In previous versions, in a multi-step calculation, the effective degree of freedom were calculated at each step based on the degrees of freedom calculated for the previous step (using a modified Welch-Satterthwaite approximation). Now effective degrees of freedom are always calculated directly from the independent variables using the standard Welch-Satterthwaite approximation.

• CODATA 2022 values instead of 2018 values are used in the Constants module.

• The significance value in budget table has been redefined from (sensitivity coefficient * standard uncertainty/combined uncertainty) to the square of that value so that the significance values in a budget sum to one.

• Units can now be raised to a fractional power and many other bug fixes.

new in version 0.6.0

• A constant library has been added with physical constants that can be accessed by name or alias with the constant function. The search_constants function with no argument gives a listing of all built-in constants. Each constant definition includes any correlations with other constants.

• The Quantity class has been added to represent a general numerical value multiplied by a unit and the unit function has been added to retrieve Unit instances from the unit library by name or alias. Unit instances can now be multiplied and divided by other Unit instances to produce composite units, can be multiplied and divided by numbers to produce Quantity instances or multiply or divide Quantity instances. The gummy class is now a subclass of Quantity with a nummy value rather than a subclass of nummy. A QuantityArray class has been introduced to represent an array of values all with the same unit. Multiplying a Unit instance by a list, tuple, or numpy array produces a QuantityArray instance.

• The immy class has been introduced as an ummy valued counterpart of the jummy class for representing complex values with uncertainties. immy and jummy values can now be displayed in a polar representation in addition to a cartesian representation. immy and jummy .r and .phi properties have been added to access the magnitude and argument of the values as a complement to the .real and .imag properties.