Why Does My File Have Funny Characters In It?

One of the most common problems that people come across when moving files between different computer systems is that their line endings seem to be destroyed. In this article, Dave Cross looks at this problem and suggests some simple solutions.

Introduction

If it hasn’t happened to you at some point then I can almost guarantee that it will in the future. You transfer a text file from a DOS/Windows system to a Unix system using FTP and every line has gained extra ^M characters; or you transfer a text file the other way only to find that all the lines have run together and where the line endings were there is a strange black rectangle.

You have become a victim of the non-standardisation of line endings across different operating systems.

What are line endings?

When you edit a text file in your favourite editor, whether it is an HTML page, the source code of a computer program, a poem or a shopping list, you press the Return (or Enter) key at the end of each line. This really changes nothing in the meaning of the file (your browser would still be able to display HTML that all appeared on one line) it merely makes it easier for you (or other humans) to read the file.

When you press the enter key, the computer needs to mark the point in the file where the line break occurs. It does this by inserting a special character (strictly speaking, at least one special character) into the file at this point. When displaying or printing the file for you in the future, the editor knows that the special character means “don’t display this character, but display the character after this on a new line”.

The problem occurs when you try to define which special characters to use. There are two characters in the standard ASCII character set that might be used for this purpose, the line feed (LF) character with ASCII code 10 and the carriage return (CR) character with ASCII code 13. There is really very little real choice between these two characters, but a choice must be made when designing an operating system.

Of course, just because one set of operating system designers choose one particular character, it doesn’t mean that all subsequent operating system designers will agree with their decision. Herein lies our problem. The three most used operating systems today are Unix (in all of its varients – including Linux), Windows and MacOS and just to keep things as complex as possible, they all use different line end characters. Here is a table showing which characters the various systems use.

Line end characters on the three most popular operating systems
OS Line End
Unix LF
Windows/DOS CR/LF
MacOS CR

As you can see, this selection just about covers all possibilities. Unix and MacOS both choose a different one of the characters and Windows is greedy and insists on them both (and notice that they must be in the order given, CR before LF). Armed with this knowledge we can begin to explain what goes wrong when you move a text file from its native operating system.

Moving Text Files

It is important to note that a text file when stored on disk does not contain the text formatted into lines. The lines are stored in one long stream of binary data. It is only when the text is displayed (in an editor or by printing the file) that the special line end characters are interpreted in a manner that will reformat the text into lines.

Normally when you move a text file onto another operating system the binary data is left untouched (that ‘normally’ hints at a solution which we will come to in more detail later). Let’s examine what happens in some common cases and see if we can explain the problems described above.

Windows to Unix

Imagine a Unix editor (like vi or emacs) opening a text file that was created under Windows. The text characters in the first line are all interpreted correctly. At the end of the line, we find a CR. This means nothing special to Unix so the editor attempts to display the character. CR is a non-printable character, which means that it has no obvious visual representation. Under Unix many of these non-printable characters are mapped to control characters when displayed. In the case of CR, its displayable equivalent is Control-M. This is displayed in most editors as ^M.

The next character is a LF. Unix has no problems with this character as it is the standard line end character. The editor therefore moves to the start of the next line on the display and starts to process the next line.

In this way you can see that the file will be displayed with an extra ^M at the end of each line, but with the line endings intact.

Imagine now a Windows editor (like notepad) trying to display a text file that was created under Unix. The text characters in the first line are all interpreted correctly. At the end of the first line the editor finds a LF character. Under Windows this is not a valid line end character as it was not preceded by a CR. The editor therefore just prints the character. Unfortunately, LF is a non-printable character and in most Windows editors, a non-printable character is displayed as a black rectangle.

The editor now moves on to the second line in the file and finds more normal text characters. As there has been no CR/LF combination to indicate the end of the line, the editor just prints the next line straight after the black rectangle.

As the editor moves through the file it displays each line straight after the previous one, separated only by the black rectangle that represent the unprintable LF character.

Macintosh Transfers

Hopefully you can now see how these problems are caused. I’ll leave it to you to think about what will happen when Unix or Windows files are transfered to a Macintosh.

Problems Caused

So how much of a problem does this cause?

It depends really on what you were planning to do with the file. If it was intended for humans to read, then obviously something needs to be done to fix the formatting. On the other hand, if it was intended for machine consumption (like a HTML page or program source code) then you might find that the broken line endings don’t matter. If your program gives unexpected results then it might be worth investigating wherther the input files have come from another system and might be suffering from this problem.

Solutions

There are a number of ways to get round or fix this problem. In increasing order of complexity, they are:

Fixing the file in an editor

If you have a broken file and you want to fix it quickly, it may be possible to fix it using your editor. Certainly on Unix, most editors are powerful enough to fix this problem. For example if you have the file open in vi, you can fix the problem by entering the following command.

   :1,$/^M//g

(You can enter the ^M character by holding down the control key whilst pressing M at the same time.)

I’m not sure if this option is open to you on Windows or MacOS, certainly notepad doesn’t have the capability to fix this problem. If anyone can give any more information on this, then please let me know.

Fixing the file during transfer

Remember earlier when I said that when moving files to another operating system, the binary data in the file was normally unaltered. Well, it is possible to fix the data during transfer, depending on the transfer method you use.

The most common way to move files between different systems is by using File Transfer Protocol (or FTP). FTP is a widely implemented protocol for moving files from one computer to another. You access FTP by using an FTP client program. This might be a command line program that is just called FTP or a GUI-based drag and drop program like WS-FTP or Cute FTP. If you transfer the file using an FTP client you can put the client into ASCII transfer mode. This means that all line endings will be converted to the correct type for the target system during the transfer. A tutorial into using FTP clients is beyond the scope of this article but it is worth noting that many newer FTP clients have the ability to automatically change transfer mode depending on the extension of the file being transfered. You may need to configure the client to understand that more esoteric extensions are, in fact, text files.

Sometimes you don’t need to explicitly transfer files from one system to another. Simply moving the file to another directory on the the same system is enough. This is because the directory is cross-mounted from another system. In this way it is possible to see Unix directories from a Windows machine. The most common method for achieving this is using something like the Network File System (NFS). Obviously, with NFS you can’t change the transfer mode as there isn’t a transfer as such. It is, however, often possible to configure NFS to automatically switch into ASCII mode depending on the extension of the file.

Fixing the file with software

Some Unix varients come complete with filter programs called dos2unix and unix2dos which will handle the conversions for you. They are implemented as filters to make them as flexible as possible. From the command line, they are invoked as follows.

    dos2unix < DOS_file.txt > Unix_file.txt

If your operating system doesn’t have these filters then it is easy enough to create them in your favourite language. Here is an example of dos2unix written in Perl.

  #!/usr/local/bin/perl -w
  use strict;

  while (<STDIN>) {
    s/\r\n/\n/;
    print;
  }

Conclusions

This is certainly becoming one of the the most common problems that I see discussed on Usenet. It has become much more prevalent now that people create many files on Windows machines and then transfer them up to their Internet Service Provider’s system which is probably running Linux or some other Unix varient.

I hope that by explaining the problem and suggesting some solutions I can help to stop too much time from being wasted discussing this simple problem.

10 comments on “Why Does My File Have Funny Characters In It?
  1. Alex Kleider says:

    In your article about “funny characters”: unix2dos and dos2unix, etc
    you suggest the vim command:
    :1,$/^M//g
    On my system it has to be :1,$/^V^M//g
    What shows up on the screen is :1,$/^M//g
    but I’m sure you already knew that and the problem was a typo.

    Thanks for the article- I enjoyed reading it.
    ak

    • Alex Kleider says:

      sorry, I forgot the most important bit:
      :1,$s/^V^M//g
      Without the ‘s’ for substitute, it doesn’t work.nn1nn1nn1

  2. Jay B says:

    This also works.. I know.. just personal preference and thanks for the article:

    :%s/^M//

    • sudharsan says:

      I am facing the issue while transferring file.
      The file having some special characters thats why its not moving.Please find the error snippet
      “Putfile failed asRoot? [1]: scp file.txt to: mv: cannot stat â��/sudha/file.txt��: No such file or directory”
      Why this special characters created “â��” if anyone knows please point me.

      • davorg says:

        It looks like this has nothing to do with the problems described in my article. This looks like your file has UTF8 characters in its name and your software isn’t configured to deal with those correctly.

        I suggest that you try posting a full explanation of the problem to StackOverflow.

  3. Jay B says:

    Sorry, I noticed my cntrl-V did not show up… Same as above do a
    cntrl V, then a cntrl M
    :%s/^V^M//

  4. Acel Dianne says:

    hi! I just recovered my file and this what happens to my file … “1Ž1é1242V2¥23M33­3s4®4Ö4555ˆ5¶5$6ˆ6º6,7†7Ñ78>8I8´8º8É8Ó8â8è8434>4H4Q4Z4¢4é45C5O5c5o5z5„5Š5‘5—5¥5¬5±5Ä5Ï5Õ5Û5á5ë566’6B6H6W6n6¢6¯6ß6å67u7œ7¤7è7
    88 888=8E8K8V88z8œ8²8Ò8N9e9t99…9‹99¹9ö9:::/:4:G:¨:Ç:â:;B;O;‹;¡;À;û;<<%<f<t<‰<©<·y>»>é>ý>7?p?ädemo demo demo

    how can I fix this one? Is there anything to make it more readable?? PLEASE HELP

  5. James Campbell says:

    When copying from *nix to Windows, use Wordpad (instead of Notepad) to open the file. It understands the bare LF and displays lines correctly.

    Make a (trivial) change (e.g. add and then delete a space), save, and you now have Windows formatted line endings.

  6. Vickie says:

    In the first example, the : symbol is used at the beginning of the line. I see this in some text files. What does it mean????
    I have needed this answer for about 15 years! Honestly!

Leave a Reply to Vickie Cancel reply

Your email address will not be published. Required fields are marked *

*