LiveFire Labs' UNIX Tip, Trick, or Shell Script of the Week

Capturing a Session's Input and Output - The UNIX script Command - Part II

Last week's tip demonstrated how to capture a session’s input and output using the UNIX script command.  If you tried the example from last week on your own system, you probably noticed that script also captured control characters (e.g. ^M) in the file.

This can be seen by viewing the file using the vi text editor:

Script started on Sat 30 Aug 2003 02:06:10 PM CDT
[root@hawk] # date^M
Sat Aug 30 14:06:31 CDT 2003^M
[root@hawk] # uptime^M
2:06pm up 4 min(s), 1 user, load average: 0.07, 0.21, 0.11^M
[root@hawk] # pwd^M
/tmp^M
[root@hawk] # hostname^M
hawk^M
[root@hawk] # who am i^M
root /dev/pts/1 Aug 30 14:06^M
[root@hawk] # exit^M

script done on Sat 30 Aug 2003 02:06:53 PM CDT

This week's tip will present four options for removing the ^M (control character) from the script file.

The first option, which is also the least desirable, is to manually edit the file. This may be fine for a small file like the one used in our example, but what if you had thousands of lines with ^M at the end of them.  Not a very appealing option, is it?

The second option is a more palatable solution, but still not the optimal one. The sed command could be used to strip the last character from each line:

[root@hawk] # sed s/.$//g unixfile > unixfile.sed

Unfortunately, this also removes characters that you may not want removed (e.g. the "T" in "CDT"):

Script started on Sat 30 Aug 2003 02:06:10 PM CD
[root@hawk] # date
Sat Aug 30 14:06:31 CDT 2003
[root@hawk] # uptime
2:06pm up 4 min(s), 1 user, load average: 0.07, 0.21, 0.11
[root@hawk] # pwd
/tmp
[root@hawk] # hostname
hawk
[root@hawk] # who am i
root /dev/pts/1 Aug 30 14:06
[root@hawk] # exit

script done on Sat 30 Aug 2003 02:06:53 PM CD

The third option uses sed again, but strips the specific character instead of the last character on each line:

[root@hawk] # sed s/^M//g unixfile > unixfile.sed2

One very important item to understand about this command is that the "^M" (control character) is not generated by typing the "^" character, and then the "M" character from your keyboard.  Instead, it is accomplished by typing Ctrl-V and then Ctrl-M (the Ctrl key and the V or M key are pressed simultaneously). Typing this sequence will produce the "^M" (control  character), which allows sed to locate and process it as instructed.

The most desirable option in my opinion, is running the dos2unix utility against the script file:

[root@hawk] # dos2unix unixfile unixfile.dos2unix

This command will convert a text file from DOS format to ISO format, but may not be available on your system. Its availability is dependent on what operating system you are using.

New to sed?  Read A Brief Introduction to sed.

Read the PREV article in this series - Capturing a Session's Input and Output - The UNIX script Command - Part I