remove limits attribute and xmin and xmax kwargs (#212)

* remove the `limits` attribut of `Samples`, `MCMCSamples`, and `NestedSamples`, as well as the `xmin` and `xmax` kwargs passed to the plotting functions

* remove unused import for flake8

* remove `limits` attribute from the various readers since it is no longer used by `Samples`

* add tests for incorrect `burn_in` input

* catch empty file UserWarning

* add tests for xmin, xmax only for fastkde to help with code coverage

* change pytest UserWarning match, apparently different for different python versions

* further xmin, xmax tests, now also individually...

* add test for incorrect `ticks` input to `make_2d_axes`

* remove inaccessible ValueError

* add test for `compress_weights` with `None` inputs

* remove what seems to be unneccessary checks whether `data` are empty arrays, this should be handled by `Samples.plot_2d`

* add docstrings for `q` to plotting functions

* Updated burn in so that there is no requirement of float

* modify docstring for `burn_in` to better reflect the distinction between <1 and >1

Co-authored-by: Will Handley <[email protected]>

anesthetic/boundary.py

@@ -4,7 +4,7 @@
 from scipy.special import erf
 
 
-def cut_and_normalise_gaussian(x, p, sigma, xmin=None, xmax=None):
+def cut_and_normalise_gaussian(x, p, bw, xmin=None, xmax=None):
  """Cut and normalise boundary correction for a Gaussian kernel.
 
  Parameters
@@ -15,7 +15,7 @@ def cut_and_normalise_gaussian(x, p, sigma, xmin=None, xmax=None):
  p: np.array
  probability densities for normalisation correction
 
- sigma: float
+ bw: float
  bandwidth of KDE
 
  xmin, xmax: float
@@ -31,9 +31,9 @@ def cut_and_normalise_gaussian(x, p, sigma, xmin=None, xmax=None):
  correction = np.ones_like(x)
 
  if xmin is not None:
- correction *= 0.5*(1 + erf((x - xmin)/sigma/np.sqrt(2)))
+ correction *= 0.5 * (1 + erf((x-xmin)/bw/np.sqrt(2)))
  correction[x < xmin] = np.inf
  if xmax is not None:
- correction *= 0.5*(1 + erf((xmax - x)/sigma/np.sqrt(2)))
+ correction *= 0.5 * (1 + erf((xmax-x)/bw/np.sqrt(2)))
  correction[x > xmax] = np.inf
  return p/correction

anesthetic/convert.py

@@ -19,7 +19,8 @@ def to_getdist(nested_samples):
  weights = nested_samples.weights
  loglikes = -nested_samples.logL.to_numpy()
  names = nested_samples.columns
- ranges = {name: nested_samples._limits(name) for name in names}
+ ranges = {name: (nested_samples[name].min(), nested_samples[name].max())
+ for name in names}
  return getdist.mcsamples.MCSamples(samples=samples,
  weights=weights,
  loglikes=loglikes,