Chapter 19. I/O Redirection

Table of Contents
19.1. Using exec
19.2. Redirecting Code Blocks
19.3. Applications

There are always three default files [1] open, stdin (the keyboard), stdout (the screen), and stderr (error messages output to the screen). These, and any other open files, can be redirected. Redirection simply means capturing output from a file, command, program, script, or even code block within a script (see Example 3-1 and Example 3-2) and sending it as input to another file, command, program, or script.

Each open file gets assigned a file descriptor. [2] The file descriptors for stdin, stdout, and stderr are 0, 1, and 2, respectively. For opening additional files, there remain descriptors 3 to 9. It is sometimes useful to assign one of these additional file descriptors to stdin, stdout, or stderr as a temporary duplicate link. [3] This simplifies restoration to normal after complex redirection and reshuffling (see Example 19-1).

   COMMAND_OUTPUT >
      # Redirect stdout to a file.
      # Creates the file if not present, otherwise overwrites it.

      ls -lR > dir-tree.list
      # Creates a file containing a listing of the directory tree.

   : > filename
      # The > truncates file "filename" to zero length.
      # If file not present, creates zero-length file (same effect as 'touch').
      # The : serves as a dummy placeholder, producing no output.

   > filename    
      # The > truncates file "filename" to zero length.
      # If file not present, creates zero-length file (same effect as 'touch').
      # (Same result as ": >", above, but this does not work with some shells.)

   COMMAND_OUTPUT >>
      # Redirect stdout to a file.
      # Creates the file if not present, otherwise appends to it.


      # Single-line redirection commands (affect only the line they are on):
      # --------------------------------------------------------------------

   1>filename
      # Redirect stdout to file "filename."
   1>>filename
      # Redirect and append stdout to file "filename."
   2>filename
      # Redirect stderr to file "filename."
   2>>filename
      # Redirect and append stderr to file "filename."
   &>filename
      # Redirect both stdout and stderr to file "filename."
      #
      #  Note that   &>>filename
      #+ -- attempting to redirect and *append*
      #+ stdout and stderr to file "filename" --
      #+ fails with the error message,
      #+ syntax error near unexpected token `>'.

   M>N
     # "M" is a file descriptor, which defaults to 1, if not explicitly set.
     # "N" is a filename.
     # File descriptor "M" is redirect to file "N."
   M>&N
     # "M" is a file descriptor, which defaults to 1, if not set.
     # "N" is another file descriptor.

      #==============================================================================

      # Redirecting stdout, one line at a time.
      LOGFILE=script.log

      echo "This statement is sent to the log file, \"$LOGFILE\"." 1>$LOGFILE
      echo "This statement is appended to \"$LOGFILE\"." 1>>$LOGFILE
      echo "This statement is also appended to \"$LOGFILE\"." 1>>$LOGFILE
      echo "This statement is echoed to stdout, and will not appear in \"$LOGFILE\"."
      # These redirection commands automatically "reset" after each line.



      # Redirecting stderr, one line at a time.
      ERRORFILE=script.errors

      bad_command1 2>$ERRORFILE       #  Error message sent to $ERRORFILE.
      bad_command2 2>>$ERRORFILE      #  Error message appended to $ERRORFILE.
      bad_command3                    #  Error message echoed to stderr,
                                      #+ and does not appear in $ERRORFILE.
      # These redirection commands also automatically "reset" after each line.
      #=======================================================================

   2>&1
      # Redirects stderr to stdout.
      # Error messages get sent to same place as standard output.

   i>&j
      # Redirects file descriptor i to j.
      # All output of file pointed to by i gets sent to file pointed to by j.

   >&j
      # Redirects, by default, file descriptor 1 (stdout) to j.
      # All stdout gets sent to file pointed to by j.

   0< FILENAME
    < FILENAME
      # Accept input from a file.
      # Companion command to ">", and often used in combination with it.
      #
      # grep search-word <filename


   [j]<>filename
      #  Open file "filename" for reading and writing,
      #+ and assign file descriptor "j" to it.
      #  If "filename" does not exist, create it.
      #  If file descriptor "j" is not specified, default to fd 0, stdin.
      #
      #  An application of this is writing at a specified place in a file. 
      echo 1234567890 > File    # Write string to "File".
      exec 3<> File             # Open "File" and assign fd 3 to it.
      read -n 4 <&3             # Read only 4 characters.
      echo -n . >&3             # Write a decimal point there.
      exec 3>&-                 # Close fd 3.
      cat File                  # ==> 1234.67890
      #  Random access, by golly.



   |
      # Pipe.
      # General purpose process and command chaining tool.
      # Similar to ">", but more general in effect.
      # Useful for chaining commands, scripts, files, and programs together.
      cat *.txt | sort | uniq > result-file
      # Sorts the output of all the .txt files and deletes duplicate lines,
      # finally saves results to "result-file".

Multiple instances of input and output redirection and/or pipes can be combined in a single command line.
command < input-file > output-file

command1 | command2 | command3 > output-file
See Example 15-31 and Example A-15.

Multiple output streams may be redirected to one file.
ls -yz >> command.log 2>&1
#  Capture result of illegal options "yz" in file "command.log."
#  Because stderr is redirected to the file,
#+ any error messages will also be there.

#  Note, however, that the following does *not* give the same result.
ls -yz 2>&1 >> command.log
#  Outputs an error message and does not write to file.

#  If redirecting both stdout and stderr,
#+ the order of the commands makes a difference.

Closing File Descriptors

n<&-

Close input file descriptor n.

0<&-, <&-

Close stdin.

n>&-

Close output file descriptor n.

1>&-, >&-

Close stdout.

Child processes inherit open file descriptors. This is why pipes work. To prevent an fd from being inherited, close it.
# Redirecting only stderr to a pipe.

exec 3>&1                              # Save current "value" of stdout.
ls -l 2>&1 >&3 3>&- | grep bad 3>&-    # Close fd 3 for 'grep' (but not 'ls').
#              ^^^^   ^^^^
exec 3>&-                              # Now close it for the remainder of the script.

# Thanks, S.C.

For a more detailed introduction to I/O redirection see Appendix E.

Notes

[1]

By convention in UNIX and Linux, data streams and peripherals (device files) are treated as files, in a fashion analogous to ordinary files.

[2]

A file descriptor is simply a number that the operating system assigns to an open file to keep track of it. Consider it a simplified type of file pointer. It is analogous to a file handle in C.

[3]

Using file descriptor 5 might cause problems. When Bash creates a child process, as with exec, the child inherits fd 5 (see Chet Ramey's archived e-mail, SUBJECT: RE: File descriptor 5 is held open). Best leave this particular fd alone.