Tweak DataFrame formatting to always use same # of digits

[wesm]

[lodagro]

Now that formatting always uses same number of digits, one can wonder what does precision mean?
One gets more or less precision depending on the columns content.

In [1]: import numpy as np

In [2]: import pandas

In [3]: df = pandas.DataFrame({'A': [-1, 1e-2, 123456789, np.pi, np.sqrt(2)]})

In [4]: df
Out[4]: 
   A        
0 -1.0000000
1  0.0100000
2  1.235e+08
3  3.1415927
4  1.4142136

By the way precision seems to be one digit off target when using the default float formatter, ok when using the eng one.

In [5]: df = pandas.DataFrame({'A': [np.pi, np.sqrt(2)]})

In [6]: df
Out[6]: 
   A    
0  3.142
1  1.414

In [7]: pandas.core.common._precision
Out[7]: 4

In [8]: pandas.set_printoptions(precision=3)

In [9]: pandas.core.common._precision
Out[9]: 3

In [10]: df
Out[10]: 
   A   
0  3.14
1  1.41

In [11]: pandas.set_eng_float_format(use_eng_prefix=False, precision=3)

In [12]: df
Out[12]: 
   A        
0  3.142E+00
1  1.414E+00

In [13]: df = pandas.DataFrame({'A': [-1, 1e-2, 123456789, np.pi, np.sqrt(2)]})

In [14]: df
Out[14]: 
   A          
0 -1.000E+00  
1  10.000E-03 
2  123.457E+06
3  3.142E+00  
4  1.414E+00  

In [15]: pandas.core.common._precision
Out[15]: 3

[adamklein]

I took precision to be number of significant digits, in which case to
me it looks like the default float formatter is actually the right
number of digits here, vs the eng one. It will be good to resolve this
ambiguity.

On Jan 8, 2012, at 2:22 PM, Wouter Overmeire
reply@reply.github.com
wrote:

Now that formatting always uses same number of digits, one can wonder what does precision mean?
One gets more or less precision depending on the columns content.

In [1]: import numpy as np

In [2]: import pandas

In [3]: df = pandas.DataFrame({'A': [-1, 1e-2, 123456789, np.pi, np.sqrt(2)]})

In [4]: df
Out[4]:
  A
0 -1.0000000
1  0.0100000
2  1.235e+08
3  3.1415927
4  1.4142136

By the way precision seems to be one digit off target when using the default float formatter, ok when using the eng one.

In [5]: df = pandas.DataFrame({'A': [np.pi, np.sqrt(2)]})

In [6]: df
Out[6]:
  A
0  3.142
1  1.414

In [7]: pandas.core.common._precision
Out[7]: 4

In [8]: pandas.set_printoptions(precision=3)

In [9]: pandas.core.common._precision
Out[9]: 3

In [10]: df
Out[10]:
  A
0  3.14
1  1.41

In [11]: pandas.set_eng_float_format(use_eng_prefix=False, precision=3)

In [12]: df
Out[12]:
  A
0  3.142E+00
1  1.414E+00

In [13]: df = pandas.DataFrame({'A': [-1, 1e-2, 123456789, np.pi, np.sqrt(2)]})

In [14]: df
Out[14]:
  A
0 -1.000E+00
1  10.000E-03
2  123.457E+06
3  3.142E+00
4  1.414E+00

In [15]: pandas.core.common._precision
Out[15]: 3

---

Reply to this email directly or view it on GitHub:
https://github.com/wesm/pandas/issues/395#issuecomment-3403865

[lodagro]

Ok, so it's a matter of definition.

I think definitions are:
precision = the effective number of decimal digits
accuracy = is the effective number of these digits which appear to the right of the decimal point
But i see precision often used when one means accuracy.

Then eng float formatter uses precision argument to indicate the effective number of digits which appear to the right of the decimal point. So for pandas maybe better to use accuracy for eng formatter iso precision (as matplotlib does).

What pandas means by precision is the minimum number of digits, since the actual precision (as defined above) depends on the column content, see my example in previous comment.