Floats

Constants

Action	Code	Details
NaN	math.nan
NaN	float('nan')
NaN	numpy.nan
Infinity	math.inf
Infinity	float('inf')
Infinity	numpy.inf
Negative infinity	-math.inf
Negative infinity	float('-inf')
Negative infinity	numpy.NINF
Pi	math.pi
e	math.e
Float epsilon (smallest representable difference)	sys.float_info.epsilon
Min float	sys.float_info.min
Max float	sys.float_info.max

Create

float() throws ValueError if the input cannot be parsed

Action	Code	Details
Float from string	float(x)	Throws ValueError if string cannot be parsed
Float from hex string representation	float.fromhex(x)
Float from string for locale	locale.atof(x)
Float from string for temporary locale	Babel.parse_decimal('1,25', locale='nl_NL.utf8')	No way to do this cleanly and thread-safe in standard Python...
Float from string for temporary locale	loc = locale.getlocale(locale.LC_NUMERIC) locale.setlocale(locale.LC_NUMERIC, 'nl_NL') f = locale.atof(x) locale.setlocale(locale.LC_NUMERIC, loc)
Float from packed struct bytes	struct.unpack('f', x)[0] | e.g., b'x00x00 @'

Test

Action	Code	Details
Whole number	float.is_integer(x)	e.g., True for 5.0
Approximately equal	math.isclose(x, y)	Uses proportional tolerance
Approximately equal with proportional tolerance tol%	math.isclose(x, y, rel_tol=tol)
Approximately equal with absolute tolerance tol	math.isclose(x, y, abs_tol=tol)
NaN	math.isnan(x)	Does not work for complex numbers (?)
Finite	math.isfinite(x)
Infinite	math.isinf(x)
Positive infinity	math.isinf(x) and x > 0
Negative infinity	math.isinf(x) and x < 0

Extract

Action	Code	Details
Hash	hash(x)

Convert

Action	Code	Details
To string	str(x)
To string, with comma as thousands separator	f'{x:,}'
To string according to locale	import locale locale.str(float)	E.g., different decimal symbol
To hex string representation	x.hex()	Format: [sign] ['0x'] integer ['.' fraction] ['p' exponent], e.g., 0x1.400000p+1
To bytes	struct.pack('f', x)