Chapter 8

The Shell (I)

When you type commands, they are interpreted by a command shell. The shell is a program that interprets user commands before passing them along to be executed. Command shells are quite sophisticated and provide a number of useful features. Although there is a lot to learn about shells, we will explore only those features that facilitate use of Humdrum. This chapter is the first of four chapters scattered throughout this book where we will pause and examine some of the more pertinent and valuable features of the shell.

In UNIX environments, many different shells have been developed over the years. The original UNIX shell was the C-shell — a shell whose syntax is similar to the C programming language. A later shell was developed by Stephen Bourne and is known as the Bourne shell. Subsequent improvements by David Korn resulted in the Korn shell. The Bourne shell was improved in light of many features introduced in the Korn shell, and resulted in the Bourne Again Shell — known as Bash. The Korn and Bash shells are the most popular and powerful of the current generation of shells. Although they were originally developed for the UNIX operating system, these shells are also available for DOS, Macintosh, Windows, Windows 98 and many other operating systems.

(Editor’s Note: These days, the BASH shell is the probably the most common, and was the default for many years on Mac OSX and Linux systems. Begining with MacOS 10.15 (Catalina), the default shell in the terminal is the z-shell (zsh). This should still work well with Humdrum Tools, but you can change it back to Bash quite easily (chsh -s /bin/bash), if you’d prefer.)

Shells themselves are advanced programming languages that provide complex control structures. When you type a command, you are already writing a program — although most of your programs are just one line in length.

Shell Special Characters

The shell interprets a number of characters in a special manner. When you type a command, you should know that most shells treat the following characters as having a special meaning: the octothorpe #, the dollar-sign $, the semicolon ;, the ampersand &, the verticule |, the asterisk *, the apostrophe ', the greve `, the greater-than sign >, the less-than sign <, the question-mark ?, the double-quote ", and the backslash ``. We’ll consider the function of each of these characters one at a time.

File Redirection (>)

Some of the special shell characters have already been discussed. The greater-than-sign > is a file redirection operator. It must be followed by a user-specified filename; any output from the preceding command is placed in the specified file. For example, the following command sorts the file inputfile and places the sorted result in the file named outputfile:

sort inputfile > outputfile

If the file outputfile already existed, its contents will be destroyed and over-written with the new output. Be careful not to assign the output to the same file as the input, since this will destroy the original input file.

Sometimes it is useful to add the results of an operation to some already existing file. The double greater-than-sign >> causes the new output to be appended to any data already in the named file. For example, the following command sorts the file inputfile and adds the sorted lines to the end of the file named outputfile. If the outputfile does not already exist, the command will create it.

sort inputfile >> outputfile

Pipe (|)

The vertical bar | is interpreted by the shell as a ‘pipe.’ Pipes are used to join the output of one command to the input of a subsequent command. For example, in the following construction, the output of command1 is routed as the input to command2:

command1 | command2

There is no practical limit to the length of a pipeline. Several pipes can be used to connect successive outputs to ensuing commands:

command1 | command2 | command3 | command4

Shell Wildcard (*)

The asterisk is interpreted by the shell as a “filename wildcard.” When it appears by itself, the asterisk is expanded by the shell to a list of all files in the current directory (in alphabetical order). For example, if the current directory contained just three files: alice, barry and chris — then the following command would be applied to all three files in consecutive order:

command * > people

The file expansion occurs at the moment when the command is invoked. So although the file people is added to the current directory, it is not included as its own input. However, if the above command was executed a second time, then the file expansion would include people — even as the file itself is over-written to receive the output. Including the output file as input is never a good idea.

Comment (#)

The octothorpe character # indicates a shell comment. Any characters following the # (up to the end of the line) are simply ignored by the shell. The following is not a command:

#grep OTL: filename

The comment can begin anywhere in the line. Here the comment begins after the filename:

grep OTL: filename   # (Search for Humdrum titles.)

Escape Character ()

Sometimes we would like to have a special character treated literally. For example, suppose we wanted to search for records containing sharps in a kern file. The following command will not work because the shell will insist on interpreting the octothorpe as beginning a comment:

grep # filename

There are several ways to “turn off” the special meaning of a character. The simplest way is to precede the character by a backslash (\) as in the following command:

grep \# filename

The backslash character itself can be treated literally by preceding it with another backslash. For example, the following command searches for down-stems in a kern file:

grep \\ filename

Escape Quotations (‘…`’)

Another way of escaping the special meaning of shell characters is to place the material in single quotes. For example, we can escape the meaning of the octothorpe # by preceding and following it by single quotes:

grep '#' filename

Single quotes are especially useful for binding spaces. For example, the following command searches for the phrase “Lennon and McCartney” in a file named beatles:

grep 'Lennon and McCartney' beatles

If the single quotes are omitted, the command means something completely different. The following command searches for the string “Lennon” in three files named and, McCartney and beatles:

grep Lennon and McCartney beatles

A common mistake is to fail to match quotation marks in a command. The shell will assume that the command is incomplete until all quotation marks are matched (both single quotes and double quotes). In the following example, we have failed to match the quotation mark. When we press the return key, the shell responds with a change of prompt indicating that it is waiting for us to complete the command.

grep '# inputfile > outputfile  >

Command Delimiter (;)

The semicolon (;) indicates the end of a command. Its presence allows more than one command to be typed on a single line. For example, the following line:

command1 ; command2

is logically identical to:


When both commands appear on the same line, they are still executed sequentially, so the second command doesn’t begin until the first is completed. Although the ability to place two or more commands on a single line may seem redundant, there are a number of circumstances where this feature proves useful.

Background Command (&)

After typing a command, the command begins executing as soon as you type the carriage return or “enter” key. When the command has finished executing, the shell will display a new command prompt. Sometimes a command can take a long time to execute so it will be awhile before the prompt is displayed again. Unfortunately, you must wait for the prompt before you can type a new command. On multitasking systems it is possible for the computer to execute more than one command concurrently. The ampersand & can be used to execute a command as a background process. When a command is ended by an ampersand, the shell creates an independent process to handle the command, and the shell immediately returns with a prompt for a new command from the user. UNIX systems provide sophisticated mechanisms for controlling concurrent processing of commands. For further information concerning these features, refer to a UNIX reference book.

Shell Command Syntax

Shell commands follow a special syntax. There are six possible components to a common command:

  • the command name,
  • one or more options,
  • one or more option parameters,
  • a command argument,
  • one or more input file names,
  • output redirection.

Each of these components is separated by blank space (tabs or spaces). A command begins with the command name — such as uniq, sort, or pitch. A command argument is a special requirement of only some commands. A good example of a command argument is the search pattern given to the grep command. In the following command, grep is the command name, “Lennon” is the command argument and beatles is the input file name:

grep Lennon beatles

For most commands, it is possible to process more than one input file. These files are simply listed at the end of the command. For example, the following grep command searches for the string “McCartney” in the file beatles and in the file wings:

grep McCartney beatles wings

Most commands provide options that modify the behavior of the command in some way. Command options are designated by a leading dash character. The specific option is usually indicated by a single alphabetic letter, such as the b option (spoken: “dash-B” option). In the uniq command, the c option causes a count to be prepended to each output line. In the following command, uniq is the command name, c is the option, and ghana32 is the name of the input file:

uniq -c ghana32

In many cases, the option is followed by a parameter that specifies further information pertaining to the invoked option. In the following command, recode is the command name, f is the option, reassign is the parameter used by the f option, and gagaku is the name of the input file:

recode -f reassign gagaku

Options and their accompanying parameters must be separated by blank space (i.e. one or more spaces and/or tabs). If more than one option is invoked, and none of the invoked options require a parameter, then the option-letters may be combined. For example, the a and b options might be invoked as -ab (or as -ba) — provided neither option requires a parameter.

Whenever an option requires a parameter, the option must be specified alone and followed immediately by the appropriate parameter. For example, in the following command, the command name is trans, the d option is followed by the numerical parameter 3; the parameter for the c option is the number 4 and the input file is named gambia21.

trans -d 3 -c 4 gambia21

Since numerical parameters can sometimes be negative, it can be difficult to discern whether a negative number is a parameter or another option. In the following example, the -3 is a parameter to the d option rather than an option by itself.

trans -d -3 -c 2 gambia21

Output Redirection

Most commands support several input and output modes. Input to a command may come from three sources. In many cases the input will come from one or more existing files. Apart from existing files, input may also come from text typed manually at the terminal, or from the output of preceding commands. When input text is entered manually it must be terminated with an end-of-file character (control-D) on a separate line. (On Microsoft operating systems the end-of-file character is control-Z.) When input is received from preceding commands, the output is sent via a UNIX pipe | as discussed above.

The different ways of providing input to a command are illustrated in the following examples. In the first example, the input (if any) is taken from the terminal (keyboard). In the second example, the input is explicitly taken from a file named input. In the third example, the input is implicitly taken from a file named input. In the fourth example, the input to command2 comes from the output of command1.

command < input
command input
command1 | command2

Outputs produced by commands may similarly be directed to a variety of locations. The default output from most commands is sent to the terminal screen. Alternatively, the output can be sent to another process (i.e. another command) using a pipe (|). Output can also be stored in a file using file redirection operator > or added to the end of a (potentially) existing file using the file-append operator >>. In the first example below, the output is sent to the screen. In the second example, the output is sent to the file outfile; if the file outfile already exists, its contents will be overwritten. In the third example, the output is appended to the end of the file outfile; if the file outfile does not already exist, it will be created. In the fourth example, the output is sent as input to the command command2.

command  command > outfile  command >> outfile  command1 | command2

When two or more commands have their inputs and outputs linked together using the pipe operator |, the entire command line is known as a pipeline. Pipelines occur frequently in Humdrum applications.


A special shell command known as tee can be used to clone a copy of some output, so that two identical output streams are generated. In the first example below, the output is piped to tee which writes one copy of the output to the file outfile and the second copy appears on the screen. In the second example, the output from command1 is split: one copy is piped to command2 for further processing, while an identical copy is stored in the file outfile1; if the file outfile1 already exists, its contents will be overwritten. In the third example, the append option (a) for tee has been invoked — meaning that the output from command will be added to the end of any existing data in the file outfile. If the file outfile does not already exist, it will be created.

command | tee outfile  command1 | tee outfile1 | command2 > outfile2  command | tee -a outfile

The tee command is a useful way of recording or diverting some intermediate data in the middle of a pipeline.


In this chapter we have noted that the shell interprets certain characters in a special way. We learned about the octothorpe (#), the ampersand &, the verticule |, the asterisk *, the apostrophe `, the greater-than sign >, the semicolon ;, and the backslash \. In a later chapter we’ll discuss the remaining special characters: the dollar-sign $, the apostrophe `, the less-than sign <, the question-mark ?, and the double-quote ",

We have also reviewed the syntax for UNIX commands. Commands can include components such as the command name, options, parameters, command arguments, input files and output redirection.