v2.5.0

Minor release with a few bug fixes additional features and speed-ups.

What's Changed

• Fix/complex corr by @fjosw in #135
• first version of skript to read xsf data by @jkuhl-uni in #133
• fix: corrected small bug in qtop_projection with several replica by @s-kuberski in #136
• Alias gm for gamma_method by @fjosw in #137
• fix: remove _filter_zeroes by @fjosw in #139
• Remove attribute is_merged by @fjosw in #140

Full Changelog: v2.4.0...v2.5.0

v2.4.0

Small minor release including the addition of logarithmic derivatives and various bug fixes and speed ups.

Full Changelog: v2.3.1...v2.4.0

v2.3.1

Bugs related to irregular, gapped Monte Carlo chains fixed.

Full Changelog: v2.3.0...v2.3.1

v2.3.0

Support for numerical differentiation in error propagation of fits added which allows to fit more general fit functions which contain special functions or which are only known in an integral representation.

Full Changelog: v2.2.0...v2.3.0

v2.2.0

The most important changes include:

• Support for pandas DataFrames containing Obs and Corr objects added.
• New hash method for Obs objects.
• A function for the extraction of the gradient flow coupling from openQCD flow measurement files added.

Full Changelog: v2.1.3...v2.2.0

v2.1.3

Fixes errors which appear with multi-dimensional Corr objects which have arrays with None elements as entry.

v2.1.2

Fixes a bug in Corr.matrix_symmetric which appeared when the content of a time slice was an array with at least one None entry. After the symmetrization the content of such a time slice is now None.

v2.1.1

Maintenance release which includes bug fixes for error propagation of correlated fits and the Fit_result class.

v2.1.0

The most important changes include:

• Functionality added to the Corr class:
  • Improvements and bugfixes for the GEVP method.
  • New prune method to reduce the size of correlator matrices before solving the GEVP.
• A bug in the json.gz format was fixed which only affected the replica mean values for data with more than one replicum. This should have no effect on the error propagation but might be relevant when attempting to reconstruct original data or constructing disconnected diagrams or reweighting.
• obs.covariance now allows for eigenvalue smoothing following the method detailed in hep-lat/9412087.
• Correlated least square fits now always perform an uncorrelated fit first to obtain more reliable starting parameters for the correlated fit.
• I/O routines for the ALPHA/Zeuthen dobs format were added.

Full Changelog: v2.0.0...v2.1.0

v2.0.0

Second major version of pyerrors which comes with various new features and non-backward compatible changes. The most important novelties include:

• The ensemble/replica distinction was reworked, replica are now indicated by the vertical bar |.
• We now support error propagation and estimation for gapped or irregular data via the parameter idl.
• The inclusion of constants with errors is now possible via cov_Obs. We do not recommend using pseudo_Obs for this purpose anymore.
• pyerrors now supports error propagation for complex valued observables, so called CObs.
• Most workflows are now centered around the Corr class for correlation functions which provides methods for effective masses, derivatives of correlators, fits, the GEVP, ...
• The default data format for exchange and long term storage is now our new json.gz format. Julia I/O routines for compatibility with the ADerrors.jl implementation are also available at https://github.com/fjosw/ADjson.jl

All details can be found in the new documentation at https://fjosw.github.io/pyerrors/pyerrors.html or in the reworked examples.

Full Changelog: v1.1.0...v2.0.0