gvar version 11.9.5

Fixes installation problem under Python 3.10.

gvar version 11.9.4

Fixes issue in evalcov_blocks that arises in Windows installations.

gvar version 11.9.3

Minor release: bug fixes in gvar.erf and gvar.PDFHistogram (plotting).

gvar version 11.9.2

Fixes (rare) 0-divide bug in evalcov. New options in correlate.

gvar version 11.9.1

Minor fix in (again) gvar.dataset.avg_data to improve speed for large problems and to work around a bug in a scipy routine when covariance matrices have more than 2**31 elements.

gvar version 11.9

This release fixes lingering problems with gvar.dataset.avg_data(dataset) when dataset is a dictionary whose entries dataset[k] have different sample sizes. It adds a new keyword mismatch that allows three different strategies for dealing with this situation.

This release also addresses segmentation faults caused by replacing a GVar by None in calls to several functions. What happens now depends on the function, but there should be no segmentation faults.

gvar version 11.8

Small update responding to a feature request: gvar.dataset.avg_data(dset) no longer discards data when dset is a dictionary whose entries have different sample sizes. In particular all of the data is used to determine standard deviations.

Also added new option (uniform) to gvar.raniter.

gvar version 11.7

gvar.powerseries now supports multivariate power series. There is also a new numerical module gvar.pade, for Pade approximants.

gvar version 11.6

gvar.regulate is a new tool for regulating singular correlation matrices. It also supports both SVD and Cholesky decompositions of correlation matrices. Based on a recommendation from Giacomo Petrillo.

gvar version 11.5.2

Fixes documentation problems introduced by version 11.5.1. No other changes.