- New package extension with
BaseType.jl(#156).
- Correlated variables can be created from a covariance or a correlation matrix (#137).
- Integration with third-party packages makes use of package extensions when using Julia v1.9 or later versions (#143).
- New methods
value(::Missing)anduncertainty(::Missing)(#127, #128). one(::Measurement{T})now returnsone(T), instead of an instance ofMeasurement{T}(#134, #135). This change should not breaking (and hence the major version was not increased), unless you were relying on the previous incorrect return type ofone(::Measurement{T}), which did not follow the precise semantic of theonefunction.
- The counter of the tags of
Measurementobjects is now atomic (#118).
- Fixed norm of
AbstractArray{<:Measurement}(#120).
- Support hashing of
Measurementobjects (#103, #104). - New method
Base.tryparse(::Type{Measurement}, ::AbstractString)(#110, #112).
factorial(a::Measurement)is deprecated in favour ofSpecialFunctions.gamma(a + 1), as a result ofSpecialFunctions.factorial(a::AbstractFloat)being removed inSpecialFunctions.jlv2.0 (#111).
- Work around issue when the extrema of an integral with
QuadGK.jlare numbers with uncertainties and units fromUnitful.jl(#75, #89). - Improved printing of value when it is smaller than the uncertainty (#94).
- When the package
Unitful.jlis loaded, nowMeasurements.valueandMeasurements.uncertaintyaccept anyAbstractQuantity, instead of onlyAbstractQuantity{<:Measurement}(#87, #88).
- New method for
sincospi(::Measurement), whensincospiis defined (#76).
- New methods for
floatmin(::Type{Measurement})andfloatmax(::Type{Measurement})(#72).
- Allow installation of new versions of
RecipesBase.jl(#70).
measurement(missing),measurement(missing, ::Real)andmeasurement(missing, missing)will now returnmissing(#59, #62)
nextfloat(x::Measurement{T})andprevfloat(x::Measurement{T})now return an object of typeMeasurement{T}with the same uncertainty asx, instead of returning an object of typeT. (#64, #65)
- Performance of
@uncertainhas been slightly improved in certain cases (#32). - New method
measurement(::Quantity, ::Quantity), for easier creation of numbers with units (from theUnitful.jlpackage) and uncertainties (#35, #36). measurement(s::String)method now understands decimal uncertainty (#44, #45)- Truncated printing is now always used.
IOContextcan be used to control the behaviour (#40, #43):julia> using Measurements julia> x = pi ± (ℯ/100) 3.142 ± 0.027 julia> println(IOContext(stdout, :error_digits=>0), x) 3.141592653589793 ± 0.02718281828459045 julia> println(IOContext(stdout, :error_digits=>1), x) 3.14 ± 0.03 julia> println(IOContext(stdout, :error_digits=>5), x) 3.141593 ± 0.027183
- Julia 1.0 is required, support for previous v0.7 was dropped.
- Uncertainties of
Measurementobjects are now shown in the REPL with two significant digits (#5, #30). The value of theMeasurementis shown with the same level of rounding. Note: this is only about representation of theMeasurementin the REPL, its value and uncertainty are not actually rounded.
- New method for
rem2pi(#29).
- Fix conversion of a
Measurementobject to anInt. This fixes also power of complex measurements. - Ensure that
Measurement{T}(x)returns aMeasurement{T}(#24). Until Julia 0.6 this was done automatically, now we have to test this explicitly.
- Julia 0.7 is required, support for previous versions was dropped.
- Documentation has been moved to https://juliaphysics.github.io/Measurements.jl/stable/.
- You can easily plot vectors of
Measurementobjects using thePlots.jlpackage. The uncertainties of the numbers will automatically be used as sizes of the errorbars. @uncertainmacro automatically converts all arguments of the function call toMeasurementtype, so that the user does not have to manually cast them.- New method for
measurement(x::Measurement)returningx. - New methods for
Measurements.value(x::Real)returningxandMeasurements.uncertainty(x::Real)returningzero(x). - The package now supports pretty printing in Juno (#12, #13)
- The tag used internally to identify an independent quantity is now a (thread-safe) counter, rather than a (thread-unsafe) random number (#15, #16).
- New
showmethod fortext/latexMIME, used in Jupyter notebooks.
isintegerandiszeronow check also that the uncertainty is zero.- Comparison between
MeasurementandRealwith==now compares also the uncertainty, so that theMeasurementmust have uncertainty equal to zero to be equal to a real number. - Comparison between
MeasurementandIrrationalwith==now gives alwaysfalse, consistently with the rest ofRealtypes.
- New mathematical operations supported:
sinpi,cospi,sinc,cosc,asec,acsc,acot,asech,acsch,acoth,sincos(the last one only on Julia 0.7). - It is now possible to parse a string as a
Measurement{T}with anyT<:AbstractFloat(not onlyMeasurement{Float64}), as long as the parsing method is able to digest the string. Tested withTequal toFloat16,Float32,Float64, andBigFloat.
- Support for Julia 0.4 and 0.5 was dropped.
valueanduncertaintyfunctions are no more exported.Measurements.gradientfunction has been removed. Use the vectorized functionMeasurements.derivative.(x, array)instead.
- There were several performance improvements, both in speed and memory usage.
These improvements are more evident for simple operations like addition and
subtraction (now these operations are 2-3 times faster than in v0.3.0). In
addition, the overhead of operations with arrays of
Measurements and complexMeasurements was greatly reduced by avoiding useless conversions (now summing an array is 100 times faster than in v0.3.0).
quadgkfunction fromQuadGK.jlpackage is extended to supportMeasurementobjects as endpoints of integration (#8). Note that only the case of two real endpoints is supported.- Real
Measurementobjects can be printed with"text/x-tex"and"text/x-latex"MIME types.\pmTeX macro is used to render the±sign. - A new
parsemethod is provided, to parse a string toMeasurement{Float64}. This is the same as callingmeasurementwith a string argument.
- New function
Measurements.uncertainty_componentsto compute the components of the total uncertainty of a derived quantity.
- When both arguments of
stdscoreareMeasurement, the standard score between their difference and zero is now computed, in order to test their consistency. This is most probably what users will expect. As a consequence of this change, the returned value is never aMeasurement.
- New method for
measurement: you can parse a string in order to create aMeasurementobject. - It is now possible to perform calculations involving numbers that have
NaNas nominal value and/or uncertainty. In previous versions it was possible only to define such numbers, but not to perform calculations with them. - Performance of
sumandprodwith arrays ofMeasurements has been slightly improved. Now they are quadratic in the number of input arguments, instead of cubic. - Improve printing of complex measurements. Now each part, real and imaginary, is wrapped in parentheses in order to make it easier to distinguish them.
- New mathematical operations supported:
log2,abs2,fma,muladd,besseli,besselix,besselk,besselkx,widen,big.
- The constructor of
Measurementobjects that you should directly use has been renamed tomeasurement. What you have to do in your programs is to replace all occurrences of the uppercase version with the lowercase one.Measurementconstructor is still exported, but you should never use it (unless you know how it works).±is now an alias ofmeasurement, so you do not need change anything if you used this sign in place of theMeasurementfunction.
measurementfunction is made vectorial, so you can create an array ofMeasurementobjects by feeding it with the array of nominal values and uncertainties.- New functions
valueanduncertaintyto access the nominal value and the uncertainty of measurements.
- Fix derivatives of
div,fld,cld. - Fix
frexpwith arguments whose absolute value is less than 0.5.
@uncertainmacro works with functions with any number of arguments.- New mathematical operations supported:
erfi,dawson,digamma,invdigamma,trigamma,polygamma,beta,lbeta,airy,airyai,airyprime,airyaiprime,airybi,airybiprime. - The complete documentation of the package is available at http://measurementsjl.readthedocs.io
- Support for correlated variables has been added (#3).
- Full support for complex measurements.
- Macro
@uncertainenables propagation of uncertainty in any real function of one or two real arguments, including those functions making use of C/Fortran calls (#4). - New functions
Measurements.derivativeandMeasurements.gradientto get the derivative and the gradient of an expression with respect to one or more independent measurements. - Precompilation enabled.
- New method supported:
typemax.
Measurementis now subtype ofAbstractFloat(#1), but you can feed the constructor with any subtype ofRealthat can be converted to anAbstractFloat.- New
weightedmeanfunction for calculating the weighted mean of measurements using inverse-variance weighting. - New mathematical operations supported:
modf,exp10,isnan,isfinite,isinf,isinteger,copysign,frexp,ldexp,div,cld,fld,mod,rem,mod2pi,eps,flipsign,erfinv,erfcinv,erfcx.
- Function
Constanthas been removed as it was mostly redundant and badly capitalized (#2). - In order to define complex
Measurements you have to usecomplex(Measurement(a, b), Measurement(c, d))so real and imaginary parts of the number have each their uncertainty.
- Fix multiplication by and division of 0. Previously, those operations would
return
NaNas uncertainty, now they give 0.
Initial release.
Measurementtype is a parametric type, subtype ofNumber.- You can define
Measurementobjects withMeasurement(a, b), beingathe value of the measurement, andbits uncertainty as standard deviation. You can also employ the shorter syntaxa ± b. - The function
stdscoreto calculate the standard score is available. - Mathematical operation supported:
+,-,*,/,inv,^,exp2,cos,sin,deg2rad,rad2deg,cosd,sind,cosh,sinh,tan,tand,tanh,acos,acosd,acosh,asin,asind,asinh,atan,atan2,atand,atanh,csc,cscd,csch,sec,secd,sech,cot,cotd,coth,exp,expm1,log,log10,log1p,hypot,sqrt,cbrt,abs,sign,zero,one,erf,erfc,factorial,gamma,lgamma,signbit.