Issue Replacing Already Existing Strings with ConfigParser

Question

I am using ConfigParser to save simple settings to a .ini file, and one of these settings is a directory. Whenever I replace a directory string such as D:/Documents/Data, with a shorter directory string such as D:/, the remaining characters are placed two lines under the option. So the .ini file now looks like this:

[Settings]
directory = D:/

Documents/Data

What am I doing wrong? Here is my code:

import ConfigParser

class Settings():
    self.config = ConfigParser.ConfigParser()

    def SetDirectory(self, dir): #dir is the directory string
        self.config.readfp(open('settings.ini'))
        self.config.set('Settings', 'directory', dir)
        with open('settings.ini', 'r+') as configfile: self.config.write(configfile)

Answer 1

The r+ option (in the open in the with) is telling Python to keep the file's previous contents, just overwriting the specific bytes that will be written to it but leaving all others alone. Use w to open a file for complete overwriting, which seems to be what you should be doing here. Overwriting just selected bytes inside an existing file is very rarely what you want to do, particularly for text files, which you're more likely to want to see as sequence of lines of text, rather than bunches of bytes! (It can be useful in very specialized cases, mostly involving large binary files, where the by-byte view may make some sense).

The "by-line organization" with which we like to view text files is not reflected in the underlying filesystem (on any OS that is currently popular, at least -- back in the dark past some file organizations were meant to mimic packs of punched cards, for example, so each line had to be exactly 80 bytes, no more, no less... but that's a far-off ancient memory, at most, for the vast majority of computer programmers and users today;-).

So, "overwriting part of a file in-place" (where the file contains text lines of different lengths) becomes quite a problem. Should you ever need to do that, btw, consider the fileinput module of the standard Python library, which mimics this often-desired but-running-against-the-filesystem's-grain operation quite competently. But, it wouldn't help you much in this case, where simple total overwriting seems to be exactly right;-).