按列索引的Numpy RecArray截断浮点数

Question

In a simple recarray in python, the output value is getting truncated when indexed by column name:

import numpy #1.10.0
arr = numpy.zeros(1, dtype=[('a', np.float)])
arr[0]['a'] = 0.1234567891234
print arr
print arr['a']

[(0.1234567891234,)]
[ 0.12345679]

Why does this happen? Can I get the full, non-truncated value with column indexing?

Answer 1

The print precision for a numeric array is 8 digits:

In [250]: np.get_printoptions()
Out[250]: 
{'edgeitems': 3,
 'formatter': None,
 'infstr': 'inf',
 'linewidth': 75,
 'nanstr': 'nan',
 'precision': 8,
 'suppress': False,
 'threshold': 1000}

But it doesn't use that value when displaying the recarray or its records. You'd probably also see the longer print with the scalar value:

print arr['a'].item()

==============

In [252]: arr = np.zeros(1, dtype=[('a', np.float)])
     ...: arr[0]['a'] = 0.1234567891234
     ...: 

In [253]: arr
Out[253]: 
array([(0.1234567891234,)], 
      dtype=[('a', '<f8')])

In [254]: arr[0]
Out[254]: (0.1234567891234,)

In [255]: arr['a']
Out[255]: array([ 0.12345679])

In [256]: arr['a'].item()
Out[256]: 0.1234567891234

In [257]: arr['a'][0]
Out[257]: 0.1234567891234

==================

https://github.com/numpy/numpy/issues/5463

array2string handles floats differently for structured array and ndarray touches on this. The formatting of numbers in a structured array record does not follow print options .

Answer 2

As others have mentioned, this is a question of printing precision. You can set a different value like this, using set_printoptions :

import numpy
numpy.set_printoptions(precision=20)
arr = numpy.zeros(1, dtype=[('a', numpy.float)])
arr[0]['a'] = 0.1234567891234
print arr
print arr['a']

> [(0.1234567891234,)]
> [ 0.12345678912339999589]