NumPy | set_printoptions method
Start your free 7-days trial now!
Numpy's set_printoptions(~) method customizes how Numpy arrays are printed.
Parameters
1. precision | int or None | optional
The number of decimal places to show. A precision of 3 would mean 3.1415 becomes 3.142. If None is passed and floatmode is not fixed, then the precision will be such that they values are uniquely differentiated. By default, precision=8.
2. thresholdlink | int | optional
If the number of values in the array is larger than threshold, then instead of printing each and every values, the values will be truncated with .... By default, threshold=1000.
3. edgeitemslink | int | optional
If truncation occurs, then the number of values to show in the front and back. By default, edgeitems=3.
4. linewidthlink | int | optional
The number of characters per line. By default, linewidth=75.
5. suppresslink | boolean | optional
Whether to show values in full decimals instead of using scientific notation. This is only applicable for floats whose absolute values are smaller than 1e-4, or the ratio between the largest value and the smallest value in the array is larger than 1000. By default, suppress=False.
6. nanstrlink | string | optional
The string to replace any nan in the array. By default, the nanstr="nan".
7. infstrlink | string | optional
The string to replace any inf in the array. By default, infstr="inf".
8. signlink | string | optional
How to handle the sign of the values:
Value | Description |
|---|---|
"-" | Omits the +. |
"+" | Places a + in front of positive numbers. |
" " | Places a single white space in front of positive numbers. |
By default, sign="-".
9. formatterlink | dict<string,function> | optional
The mapping to apply to different data-types. The dictionary's key-value pair is as follows:
key: the type you wish to apply a mapping on
value: a function that takes as input the value with type
key, and returns a new value.
Here are some of main data-types:
Type | Description |
|---|---|
"bool" | Convert booleans. |
"int" | Convert integers. |
"float" | Convert floats. |
"timedelta" | Convert timedeltas. |
"datatime" | Convert datetimes. |
Here are some special keys that you can provide:
Key | Description |
|---|---|
"all" | Convert all data-types. |
"float_kind" | Convert "float" and "longfloat" |
"str_kind" | Convert "str" and "numpystr" |
By default, formatter=None.
10. floatmodelink | string | optional
How to handle precision for floats:
Value | Description |
|---|---|
"fixed" | Always show the specified |
"unique" | Show minimum number of decimal places so as to uniquely identify the values. This ignores the specified |
"maxprec" | Prints at most the specified |
"maxprec_equal" | Prints at most the specified |
By default, floatmode="maxprec_equal".
Return value
None, since this method just prints on the screen.
Examples
Basic usage
To show 3 fractional digits:
np.set_printoptions(precision=3)np.array([0.000005, 3.1416])
array([5.000e-06, 3.142e+00])
To have unfixed precision, pass a None:
np.set_printoptions(precision=None)np.array([3.14, 3.1416])
array([3.14 , 3.142])
Specifying threshold
By default, threshold=1000, which means that printing arrays that have 1000 values or more will be summarised:
print(np.arange(1500))
[ 0 1 2 ... 1497 1498 1499]
This means that small arrays will not be summarised:
print(np.arange(7))
[0 1 2 3 4 5 6]
We can set a threshold so that even these small arrays will be summarised:
np.set_printoptions(threshold=3)np.arange(7)
array([0, 1, 2, ..., 4, 5, 6])
Specifying edgeitems
By default, edgeitems=3, which means that when values are summarised, 3 values will be shown on the left, and 3 on the right:
print(np.arange(1500))
[ 0 1 2 ... 1497 1498 1499]
We can customise this by setting our own edgeitems:
np.set_printoptions(edgeitems=4)print(np.arange(1500))
[ 0 1 2 3 ... 1496 1497 1498 1499]
Specifying linewidth
By default, linewidth=75, which means that each printed line can have at most 75 characters:
print(np.array([12, 34, 5]))
[12 34 5]
To print at most only 7 characters per line:
np.set_printoptions(linewidth=7)print(np.array([12, 34, 5]))
[12 34 5]
Specifying suppress
To show all decimal places for numbers smaller than 1e-4:
np.set_printoptions(suppress=True)print(np.array([1e-5]))
[0.00001]
The default behaviour of suppress=False gives us the following:
np.set_printoptions(suppress=False)print(np.array([1e-5]))
[1.e-05]
Specifying nanstr
np.set_printoptions(nanstr="Missing!")print(np.array([np.NaN, 3.14]))
[Missing! 3.14]
Specifying infstr
np.set_printoptions(infstr="Infinity!")print(np.array([np.inf, 3.14]))
[Infinity! 3.14]
Specifying sign
To show the + sign for positive numbers:
np.set_printoptions(sign="+")print(np.array([np.inf, 3.14, -2]))
[ +inf +3.14 -2. ]
To add a " " in front of positive numbers:
np.set_printoptions(sign=" ")print(np.array([np.inf, 3.14, -2]))
[ inf 3.14 -2. ]
It's hard to see here, but a single whitespace has been added.
Specifying formatter
To convert all boolean True to 1 and False to "-1".
mapping = { "bool": lambda x: "1" if x else "-1"}
np.set_printoptions(formatter=mapping)print(np.array([True, False, True]))
[1 -1 1]
Here, make sure you return a string in the mapping, otherwise an error will be thrown.
Specifying floatmode
fixed
To print floats with the same decimal places (i.e. 8 by default):
np.set_printoptions(floatmode="fixed")print(np.array([5.05, 5.05001]))
[5.05000000 5.05001000]
unique
To print floats with the minimum number of decimal places so as to uniquely identify the values:
np.set_printoptions(floatmode="unique")print(np.array([5.05, 5.05001]))
[5.05 5.05001]
Note that this ignores the precision parameter.
maxprec
Same as unique, but the floats can have at most precision:
np.set_printoptions(floatmode="maxprec", precision=4)print(np.array([5.05, 5.052999]))
[5.05 5.053]
maxprec_equal
Same as maxprec, but the floats will all have the same precision:
np.set_printoptions(floatmode="maxprec_equal")print(np.array([5.05, 5.05001]))
[5.05000 5.05001]
