Development version allow for formatting similar to Python floats, including same format specifiers (i.e. 'f' or 'e'). Though, this introduced an incompatibility in case when the format specifier is missing.
On v1.3.0:
>>> from mpmath import *
>>> x = mpf(pi)
>>> f"{x}"
'3.14159265358979'
>>> f"{x!s}"
'3.14159265358979'
>>> f"{x!r}"
"mpf('3.1415926535897931')"
>>> f"{x!a}"
"mpf('3.1415926535897931')"
>>> f"{x=}"
"x=mpf('3.1415926535897931')"In the master:
>>> from mpmath import *
>>> x = mpf(pi)
>>> f"{x}" # (!)
'3.1415926535897931'
>>> f"{x!s}"
'3.14159265358979'
>>> f"{x!r}"
"mpf('3.1415926535897931')"
>>> f"{x!a}"
"mpf('3.1415926535897931')"
>>> f"{x=}"
"x=mpf('3.1415926535897931')"A quick fix (not sure if it's a good idea):
diff --git a/mpmath/libmp/libmpf.py b/mpmath/libmp/libmpf.py
index 9eb24222..3ca77388 100644
--- a/mpmath/libmp/libmpf.py
+++ b/mpmath/libmp/libmpf.py
@@ -1633,7 +1633,7 @@ def format_digits(num, format_dict, prec):
strip_last_zero = True
if precision < 0:
- precision = repr_dps(prec)
+ precision = prec_to_dps(prec)
if precision == 0:
precision = 1
The problem is that it breaks float(format(x)) == float(format(float(x))) invariant.
Development version allow for formatting similar to Python floats, including same format specifiers (i.e.
'f'or'e'). Though, this introduced an incompatibility in case when the format specifier is missing.On v1.3.0:
In the master:
A quick fix (not sure if it's a good idea):
The problem is that it breaks
float(format(x)) == float(format(float(x)))invariant.