Merge pull request #9941 from ahaldane/dragon4

ENH: Use Dragon4 algorithm to print floating values

1 files changed, 24 insertions, 0 deletions

diff --git a/doc/release/1.14.0-notes.rst b/doc/release/1.14.0-notes.rst
index 6feea0e76..c1e632c80 100644
--- a/doc/release/1.14.0-notes.rst
+++ b/doc/release/1.14.0-notes.rst
@@ -16,6 +16,8 @@ New functions
 
 * ``parametrize``: decorator added to numpy.testing
 * ``chebinterpolate``: Interpolate function at Chebyshev points.
+* ``format_float_positional`` and ``format_float_scientific`` : format
+  floating-point scalars unambiguously with control of rounding and padding.
 
 
 Deprecations
@@ -310,6 +312,28 @@ now supported for these arrays:
  * `arr.resize(...)`
  * `pickle.dumps(arr)`
 
+Float printing now uses "dragon4" algorithm for shortest decimal representation
+-------------------------------------------------------------------------------
+All numpy floating-point types (16, 32, 64 and 128 bit) can now be printed to
+give shortest decimal representation of the number, which uniquely identifies
+the value from others of the same type.
+
+New functions ``np.format_float_scientific`` and ``np.format_float_positional``
+are provided to generate these decimal representations, with control over
+rounding, trimming and padding.
+
+The ``str`` and ``repr`` of floating-point scalars now shows the shortest
+unique decimal representation. This means there will be a different number of
+digits compared to numpy 1.13, for instance float64 and float128 will be longer
+and float16 will be shorter.
+
+For arrays of floating-point type, a new formatting option ``floatmode`` has
+been added to ``np.set_printoptions`` and ``np.array2string``, which gives
+control over uniqueness and rounding of printed elements in arrays. The new
+default is ``floatmode='maxprec'`` with ``precision=8``, which will print at most
+8 fractional digits, or fewer if an element can be uniquely represented with
+fewer.
+
 
 Changes
 =======