Nway

Added pandas-based API
Improved hashing procedure: faster and less memory use.

Fixed numpy deprecation issue (logical not is ~) (4.1).

Elliptical errors (4.2)

tqdm instead of progressbar2, histogram robustness (4.3)

Elliptical errors in visualisations (4.4)

Elliptical error bug fix (4.5)

Healpix-based hashing instead of a ra/dec grid. Works now accurately
also on the poles and near ra=0.
More accurate great-arc distance formula.

Updated fake catalogue script (3.1)

Added healpy to install dependencies (3.2)

Use faster ra/dec grid when safe in off-pole/off-ra-origin fields (3.3)

Replaced progressbar with progressbar-latest (3.4)

fixed warnings for very recent numpy versions (3.5)

updated manual and made nway-write-header.py more user-friendly (3.7)

added Continuous Integration builds, checking for updates (3.8)

Added correction for associations additional to association with the
primary catalogue source.

Using magnitude histogram ratios rather than fractions (1.9)

Corrected magnitude prior normalisation for the case of NaNs or missing source (1.7)

Column name changes:
bf -> dist_bayesfactor
bfpost -> dist_post
post -> p_single
post_group_no_match -> p_any (Note: p_any = 1 - post_group_no_match)
post_group_this_match -> p_i

match_flag was modified to use p_i instead of post.