Fix to_compact, infer_base_unit for non-float non_int_type (#1461)

to_compact used math.log10, which converted non-float types to floats;
this caused exceptions when using a non-integer type, like Decimal,
that does not implement multiplication/division with floats, and
generally changed non-integer types to floats.  This addresses the
problem by checking non_int_type.  If non_int_type != float, it
assumes that the type implements a log10 method, as Decimal does.

infer_base_unit created a generic UnitsContainer, not a UnitsContainer
specific to a registry; this allows specification of a registry, or
uses the registry of the passed-in Quantity/UnitLike.

1 files changed, 1 insertions, 0 deletions

diff --git a/CHANGES b/CHANGES
index 2de0a4e..ba0397c 100644
--- a/CHANGES
+++ b/CHANGES
@@ -16,6 +16,7 @@ Pint Changelog
 - Use default numpy `np.printoptions` available since numpy 1.15.
 - Implement `numpy.nanprod` (Issue #1369)
 - Fix default_format ignored for measurement (Issue #1456)
+- Fix `to_compact` and `infer_base_unit` for non-float non_int_type.
 
 ### Breaking Changes