astropy.io.ascii is able to write ASCII tables out to a file or file-like object using the same class structure and basic user interface as for reading tables.
The write() function provides a way to write a data table as a formatted ASCII table. For example:
>>> import numpy as np
>>> from astropy.io import ascii
>>> x = np.array([1, 2, 3])
>>> y = x ** 2
>>> ascii.write([x, y], 'values.dat', names=['x', 'y'])
The values.dat file will then contain:
x y
1 1
2 4
3 9
Most of the input table Supported formats for reading are also available for writing. This provides a great deal of flexibility in the format for writing. The example below writes the data as a LaTeX table, using the option to send the output to sys.stdout instead of a file:
>>> ascii.write(data, format='latex')
\begin{table}
\begin{tabular}{cc}
x & y \\
1 & 1 \\
2 & 4 \\
3 & 9 \\
\end{tabular}
\end{table}
There is also a faster Cython engine for writing simple formats, which is enabled by default for these formats (see Fast ASCII I/O). To disable this engine, use the parameter fast_writer:
>>> ascii.write(data, 'values.csv', format='csv', fast_writer=False)
The input table argument to write() can be any value that is supported for initializing a Table object. This is documented in detail in the Constructing a table section and includes creating a table with a list of columns, a dictionary of columns, or from numpy arrays (either structured or homogeneous). The sections below show a few examples.
An AstroPy Table object or a NumPy structured array (or record array) can serve as input to the write() function.
>>> from astropy.io import ascii
>>> from astropy.table import Table
>>> data = Table({'a': [1, 2, 3],
... 'b': [4.0, 5.0, 6.0]},
... names=['a', 'b'])
>>> ascii.write(data)
a b
1 4.0
2 5.0
3 6.0
>>> data = np.array([(1, 2., 'Hello'), (2, 3., "World")],
... dtype=('i4,f4,a10'))
>>> ascii.write(data)
f0 f1 f2
1 2.0 Hello
2 3.0 World
The output of astropy.io.ascii.read is a Table or NumPy array data object that can be an input to the write() function.
>>> data = ascii.read('t/daophot.dat', format='daophot')
>>> ascii.write(data, 'space_delimited_table.dat')
A list of Python lists (or any iterable object) can be used as input:
>>> x = [1, 2, 3]
>>> y = [4, 5.2, 6.1]
>>> z = ['hello', 'world', '!!!']
>>> data = [x, y, z]
>>> ascii.write(data)
col0 col1 col2
1 4.0 hello
2 5.2 world
3 6.1 !!!
The data object does not contain information about the column names so Table has chosen them automatically. To specify the names, provide the names keyword argument. This example also shows excluding one of the columns from the output:
>>> ascii.write(data, names=['x', 'y', 'z'], exclude_names=['y'])
x z
1 hello
2 world
3 !!!
A dictionary containing iterable objects can serve as input to write(). Each dict key is taken as the column name while the value must be an iterable object containing the corresponding column values.
Since a Python dictionary is not ordered the output column order will be unpredictable unless the names argument is provided.
>>> data = {'x': [1, 2, 3],
... 'y': [4, 5.2, 6.1],
... 'z': ['hello', 'world', '!!!']}
>>> ascii.write(data, names=['x', 'y', 'z'])
x y z
1 4.0 hello
2 5.2 world
3 6.1 !!!
The write() function accepts a number of parameters that specify the detailed output table format. Each of the Supported formats is handled by a corresponding Writer class that can define different defaults, so the descriptions below sometimes mention “typical” default values. This refers to the Basic writer and other similar Writer classes.
Some output format Writer classes, e.g. Latex or AASTex accept additional keywords, that can customize the output further. See the documentation of these classes for details.
There are two ways to specify the output for the write operation:
For each key (column name) use the given value to convert the column data to a string. If the format value is string-like then it is used as a Python format statement, e.g. ‘%0.2f’ % value. If it is a callable function then that function is called with a single argument containing the column value to be converted. Example:
astropy.io.ascii.write(table, sys.stdout, formats={'XCENTER': '%12.1f',
'YCENTER': lambda x: round(x, 1)},
This can be used to fill missing values in the table or replace values with special meaning.
See the Bad or missing values section for more information on the syntax. The syntax is almost the same as when reading a table. There is a special value astropy.io.ascii.masked that is used a say “output this string for all masked values in a masked table (the default is to use a '--'):
>>> import sys
>>> from astropy.table import Table, Column, MaskedColumn
>>> from astropy.io import ascii
>>> t = Table([(1, 2), (3, 4)], names=('a', 'b'), masked=True)
>>> t['a'].mask = [True, False]
>>> ascii.write(t, sys.stdout)
a b
-- 3
2 4
>>> ascii.write(t, sys.stdout, fill_values=[(ascii.masked, 'N/A')])
a b
N/A 3
2 4
If no fill_values is applied for masked values in astropy.io.ascii, the default set with numpy.ma.masked_print_option.set_display applies (usually that is also '--'):
>>> ascii.write(t, sys.stdout, fill_values=[])
a b
-- 3
2 4
Note that when writing a table all values are converted to strings, before any value is replaced. Because fill_values only replaces cells that are an exact match to the specification, you need to provide the string representation (stripped of whitespace) for each value. For example, in the following commands -99 is formatted with two digits after the comma, so we need to replace -99.00 and not -99:
>>> t = Table([(-99, 2), (3, 4)], names=('a', 'b'))
>>> ascii.write(t, sys.stdout, fill_values = [('-99.00', 'no data')],
... formats={'a': '%4.2f'})
a b
"no data" 3
2.00 4
Similarly, if you replace a value in a column that has a fixed length format, e.g. 'f4.2', then the string you want to replace must have the same number of characters, in the example above fill_values=[(' nan',' N/A')] would work.