Shell scripts

It is often convenient to simply write what one wants done into a file or script, and execute the script as though it were any other operating-system shell command. The interface to more weighty programs is often provided in the form of a script, and users frequently build their own scripts or customize existing ones to suit particular needs. Scripting is arguably the most frequent programming task performed. For many users, it is the only programming they will ever do.

Operating systems such as Unix and DOS (the command-line interface provided in Windows) provide such a scripting mechanism, but the scripting language in both cases is very rudimentary. Often a script is just a sequence or batch of commands that one would type to the shell prompt. It saves the user from having to type every one of the shell commands individually each time they require the same or similar sequence to be performed. Some scripting languages throw in a small amount of programmability in the form of a conditional and a loop, but that is about all. This is enough for smallish tasks, but as one’s scripts become bigger and more demanding, as scripts invariably seem to do, one often feels the need for a fuller fledged programming language. A Scheme with an adequate operating-system interface makes scripting easy and maintainable.

This section will describe how to write scripts in Scheme. Since there is wide variation in the various Scheme dialects on how to accomplish this, we will concentrate on the MzScheme dialect, and document in appendix A the modifications needed for other dialects. We will also concentrate on the Unix operating system for the moment; appendix B will deal with the DOS counterpart.

16.1  Hello, World!, again

We will now create a Scheme script that says hello to the world. Saying hello is of course not a demanding scripting problem for traditional scripting languages. However, understanding how to transcribe it into Scheme will launch us on the path to more ambitious scripts. First, a conventional Unix hello script is a file, with contents that look like:

echo Hello, World! 

It uses the shell command echo. The script can be named hello, made into an executable by doing

chmod +x hello 

and placed in one of the directories named in the PATH environment variable. Thereafter, anytime one types

hello 

at the shell prompt, one promptly gets the insufferable greeting.

A Scheme hello script will perform the same output using Scheme (using the program in sec 1), but we need something in the file to inform the operating system that it needs to construe the commands in the file as Scheme, and not as its default script language. The Scheme script file, also called hello, looks like:

":"; exec mzscheme -r $0 "$@" 
 
(display "Hello, World!") 
(newline)) 

Everything following the first line is straight Scheme. However, the first line is the magic that makes this into a script. When the user types hello at the Unix prompt, Unix will read the file as a regular script. The first thing it sees is the ":", which is a shell no-op. The ; is the shell command separator. The next shell command is the exec. exec tells Unix to abandon the current script and run mzscheme ‑r $0 "$@" instead, where the parameter $0 will be replaced by the name of the script, and the parameter "$@" will be replaced by the list of arguments given by the user to the script. (In this case, there are no such arguments.)

We have now, in effect, transformed the hello shell command into a different shell command, viz,

mzscheme -r /whereveritis/hello 

where /whereveritis/hello is the pathname of hello.

mzscheme calls the MzScheme executable. The ‑r option tells it to load the immediately following argument as a Scheme file after collecting any succeeding arguments into a vector called argv. (In this example, argv will be the null vector.)

Thus, the Scheme script will be run as a Scheme file, and the Scheme forms in the file will have access to the script’s original arguments via the vector argv.

Now, Scheme has to tackle the first line in the script, which as we’ve already seen, was really a well-formed, traditional shell script. The ":" is a self-evaluating string in Scheme and thus harmless. The ‘;’ marks a Scheme comment, and so the exec ... is safely ignored. The rest of the file is of course straight Scheme, and the expressions therein are evaluated in sequence. After all of them have been evaluated, Scheme will exit.

In sum, typing hello at the shell prompt will produce

Hello, World! 

and return you to the shell prompt.

16.2  Scripts with arguments

A Scheme script uses the variable argv to refer to its arguments. For example, the following script echoes all its arguments, each on a line:

":"; exec mzscheme -r $0 "$@"

;Put in argv-count the number of arguments supplied

(define argv-count (vector-length argv))

(let loop ((i 0))
  (unless (>= i argv-count)
    (display (vector-ref argv i))
    (newline)
    (loop (+ i 1))))

Let’s call this script echoall. Calling echoall 1 2 3 will display

1 
2 
3 

Note that the script name ("echoall") is not included in the argument vector.

16.3  Example

Let’s now tackle a more substantial problem. We need to transfer files from one computer to another and the only method we have is to use a 3.5” floppy as a ferry. We need a script split4floppy that will split files larger than 1.44 million bytes into floppy-sized chunks. The script file split4floppy is as follows:

":";exec mzscheme -r $0 "$@"

;floppy-size = number of bytes that will comfortably fit on a
;              3.5" floppy

(define floppy-size 1440000)

;split splits the bigfile f into the smaller, floppy-sized
;subfiles, viz, subfile-prefix.1, subfile-prefix.2, etc.

(define split
  (lambda (f subfile-prefix)
    (call-with-input-file f
      (lambda (i)
        (let loop ((n 1))
          (if (copy-to-floppy-sized-subfile i subfile-prefix n)
              (loop (+ n 1))))))))

;copy-to-floppy-sized-subfile copies the next 1.44 million
;bytes (if there are less than that many bytes left, it
;copies all of them) from the big file to the nth
;subfile.  Returns true if there are bytes left over,
;otherwise returns false.

(define copy-to-floppy-sized-subfile
  (lambda (i subfile-prefix n)
    (let ((nth-subfile (string-append subfile-prefix "."
                                      (number->string n))))
      (if (file-exists? nth-subfile) (delete-file nth-subfile))
      (call-with-output-file nth-subfile
        (lambda (o)
          (let loop ((k 1))
            (let ((c (read-char i)))
              (cond ((eof-object? c) #f)
                    (else
                     (write-char c o)
                     (if (< k floppy-size)
                         (loop (+ k 1))
                         #t))))))))))

;bigfile = script's first arg
;        = the file that needs splitting

(define bigfile (vector-ref argv 0))

;subfile-prefix = script's second arg
;               = the basename of the subfiles

(define subfile-prefix (vector-ref argv 1))

;Call split, making subfile-prefix.{1,2,3,...} from
;bigfile

(split bigfile subfile-prefix)

Script split4floppy is called as follows:

split4floppy largefile chunk 

This splits largefile into subfiles chunk.1, chunk.2, ..., such that each subfile fits on a floppy.

After the chunk.i have been ferried over to the target computer, the file largefile can be retrieved by stringing the chunk.i together. This can be done on Unix with:

cat chunk.1 chunk.2 ... > largefile 

and on DOS with:

copy /b chunk.1+chunk.2+... largefile