如何在pandas中使用read_fwf跳过空白行？

Question

I use pandas.read_fwf() function in Python pandas 0.19.2 to read a file fwf.txt that has the following content:

# Column1 Column2
      123     abc

      456     def

#
#

My code is the following:

import pandas as pd
file_path = "fwf.txt"
widths = [len("# Column1"), len(" Column2")]
names = ["Column1", "Column2"]
data = pd.read_fwf(filepath_or_buffer=file_path, widths=widths, 
                   names=names, skip_blank_lines=True, comment="#")

The printed dataframe is like this:

    Column1 Column2
0   123.0   abc
1   NaN     NaN
2   456.0   def
3   NaN     NaN

It looks like the skip_blank_lines=True argument is ignored, as the dataframe contains NaN's.

What should be the valid combination of pandas.read_fwf() arguments that would ensure the skipping of blank lines?

Answer 1

import io
import pandas as pd
file_path = "fwf.txt"
widths = [len("# Column1 "), len("Column2")]
names = ["Column1", "Column2"]

class FileLike(io.TextIOBase):
    def __init__(self, iterable):
        self.iterable = iterable
    def readline(self):
        return next(self.iterable)

with open(file_path, 'r') as f:
    lines = (line for line in f if line.strip())
    data = pd.read_fwf(FileLike(lines), widths=widths, names=names, 
                       comment='#')
    print(data)

prints

   Column1 Column2
0      123     abc
1      456     def

---

with open(file_path, 'r') as f:
    lines = (line for line in f if line.strip())

defines a generator expression (ie an iterable) which yields lines from the file with blank lines removed.

The pd.read_fwf function can accept TextIOBase objects. You can subclass TextIOBase so that its readline method returns lines from an iterable:

class FileLike(io.TextIOBase):
    def __init__(self, iterable):
        self.iterable = iterable
    def readline(self):
        return next(self.iterable)

Putting these two together gives you a way to manipulate/modify lines of a file before passing them to pd.read_fwf .