Version 1.3 2017-06-13
======================

- Significant rewrite of pade_svd and especially pade_gvar to clean them up.
  Also the relative tolerance rtol is more flexible now.

- vacpol will now raise an exception if the poles in its
  Pade approximant are complex or too close to being positive. This
  behavior can be turned off, but that is likely to be a bad idea.

- vacpol has new method FT(t) which Fourier transforms the Pade back
  into t space.

- vacpol.vector restricts its Pade to order=(1,1) because that is exact.

- Bug fix in fourier_vacpol for periodic=True

Version 1.2.1 2017-05-26
========================

- Bug fix for unusual crashes of pade_svd.

- New option qth for vacpol.

- More tests.

- Default rtol for pade_gvar is now set equal to the geometric
  mean of the ratios f[i].sdev/abs(f[i].mean) for each of the
  Taylor coefficients f[i].

Version 1.2 2017-05-25
======================

- Bug fix so vacpol.taylor() works.

- New option warn in vacpol. Setting warn=True causes a warning to be
  issued if either badpoles() is True or if the order of the Pade is
  reduced.

Version 1.1 2017-05-23
======================

- New class fourier_vacpol that creates a subtracted vacuum polarization
  function from the Fourier transform of G(t). This provides an alternative
  to using moments of G(t).

- New argument qmin for a_mu.

- New method vacpol.taylor which returns Taylor coefficients
  (shorthand for vpol.pseries['taylor'].c).

- Small bug in vacpol.vector --- set n=20 no matter what n set to.


Version 1.0.1  13/03/2017
=========================
Tiny change to make g2tools compatible with the latest lsqfit (v9).

