links to this page:    
View this PageEdit this PageUploads to this PageHistory of this PageTop of the SwikiRecent ChangesSearch the SwikiHelp Guide
Overview of CommandShell
Last updated at 11:10 am UTC on 9 March 2018
CommandShell is a Smalltalk implementation of a command processor shell and terminal window. It is intended to behave like to a simple terminal window (like xterm) running a Unix command shell (like /bin/sh). It lacks some elements of Unix shell syntax, and does not provide terminal emulation, but it adds some nice Smalltalk enhancements such as a text editor which works in a command pipeline, and the ability to evalute Smalltalk expressions in a command pipeline with unix commands.

How to use CommandShell

 CommandShell open
or, if you have loaded the change set to put CommandShell in your world menu, just select "squeak shell" from the "open..." menu entry.

To use the command shell window, enter commands on the command line (after the "$ " command prompt), followed by cr to accept the command.

The "help" command gives some clues as to what can be done.

In general, just try any command which you might otherwise enter into a Unix shell or xterm window. Then try some of the Smalltalk features such as mixing Smalltalk expressions with Unix commands, and piping commands into Smalltalk editors.

Basic Concepts


Smalltalk uses objects and messages, with expressions arranged in a natural Noun-Verb-Predicate sentence structure, similar to that of some spoken languages. In contrast, many other computer systems use a style in which users issue a command to the system, in which the "command" (a verb) is followed by various (predicate) modifiers. In the case of a traditional command-line interface, such as a Unix shell, the command is usually the name of a program, and the modifiers are parameters passed to the program to influence its behavior. CommandShell attempts to provide a command line interface for executing commands within Squeak, in a style and environment which is comforable for someone accustomed to the Smalltalk style of expression.

Process Proxies

A command line user interface, such as a Unix shell or a "DOS window" in Windows, revolves around the notion of executing programs from user commands. When Squeak is hosted in another operating system, it is useful to be able to execute an external program from a command line expression. A process proxy is an object which represents the evaluation of such an external command in a command line environment.

From the point of view of Squeak, a process proxy might just as well represent the evaluation of a Smalltalk expression in the context of a command line environment. Thus a process proxy can be an object which represents the execution of an external program or an internal Smalltalk expression. It is created from a command line expression, and provides a representation of certain aspects of the external (or internal) process execution, such as process run state, exit status, and the input, output and error streams for the process.

External process proxies are used to evaluate programs in the external operating system (this requires the OSProcess change set in addition to CommandShell). Internal process proxies are used to evaluate "built in" commands for a command shell, such as the "cd" command to change the working directory for the command shell. Internal proxies are also used for evaluating "doIt" expressions within a command pipeline. In this case, the variables "stdin", "stdout", and "stderr" are pre-defined such that the doIt expression can directly access the input, output and error pipes for the internal process proxy (in other words, the doIt expression "stdout nextPutAll: 'Hello world' !" will write 'Hello world' on the output pipe of the proxy).


Powerful computing systems may be built upon simple design metaphors. In Smalltalk, the concept of objects communicating through messages is generalized and extended to produce the Squeak system. In Unix systems, the metaphor of a pipe with data flowing between two programs is generalized to produce a mechanism for connecting small programs to produce complex systems. The command line shells for Unix support this metaphor by providing syntax for connecting two or more commands together into more complex command pipelines.

CommandShell provides a framework for connecting several process proxies into a command pipeline, using command syntax similar to that of a Unix shell. Since the process proxies are implemented in Smalltalk, the "programs" in a command pipeline may consist of process proxies representing external programs, or of proxies representing the evaluation of internal Smalltalk expressions. The objects which flow through the pipes in a command pipeline are assumed to be characters, such that the external commands can operate on the streams of characters in the usual way, and the internal Smalltalk commands read and write streams of characters.

Just as programs (or internal expressions) are represented by process proxies, the pipes which interconnect two process proxies are represented by pipe objects. The pipe may be either an OSPipe (a proxy representing a Unix system FIFO pipe) or an InternalPipe (a Smalltalk object which behaves similarly to an OSPipe). The combination of process proxies and pipes is a command pipeline, which may be created and evaluated from a command line expression using conventional Unix shell syntax.

Command Line Syntax

A command or command pipeline may be created from a string expression, and evaluated using Unix shell syntax. CommandShell accepts command line strings and does some high level processing to set up process proxy execution. For most of the command parsing, it relies on a ShellSyntax object to do the syntax evaluation. In particular, a real Unix shell interprets command line parameters in the context of the Unix file system, expanding "wildcard" characters and searching for files in the context of a "current working directory" location in the file system tree. A CommandShell, collaborating with an instance of ShellSyntax, implements this syntax parsing and evaluation in Smalltalk.

Command Execution

Beginning with a command line (which may represent a pipeline of several commands), a CommandShell first attempts to treat the entire command line as a Smalltalk expression. The command line string is used to create an instance of PipeableEvaluator (a kind of process process). If the expression is successfully compiled and evaluated in the PipeableEvaluator, the command line is executed just as if it were an external Unix command, with the result of the evaluation written as a string to the output stream of the process proxy.

If the complete command line cannot be treated as a Smalltalk expression, CommandShell assumes that it must be a Unix style command pipeline. If the command line is a comment (starting with '#') it is discarded; otherwise it is broken down into a series of process proxies connected by pipes, with command line parameters and other aspects of the command line parsing handled by a ShellSyntax object.

The process proxies in the command pipeline may represent internal or external commands. CommandShell parses the command line to create the proxies by breaking the command pipeline into segments (separated by the "|" pipe character), with different kinds of process proxy created according to the following priorities:

  1. If the command segment is a simple Smalltalk expression terminated by "!", it is treated as a "doIt" expression, evaluated in a PipeableEvaluator proxy. This kind of simple doIt expression is limited to commands with characters which are unambiguous in a shell command. For example, "|" has special meaning in a command line, and cannot be used in a doIt expression (command line quoting and escape characters have not yet been implemented in the CommandShell syntax).

  2. If the command segment is not a doIt expression, CommandShell checks to see if it matches one of several "shell builtin" commands. These are special commands, implemented in Smalltalk using PipeableEvaluator proxies, which behave analogously to the shell builtin commands in a conventional Unix shell. In CommandShell, commands such as "cd" and "pwd" are implemented as internal builtin commands, rather than relying on Unix external equivalents. In addition, other builtin commands such as "edit" do things which are useful in Squeak and have no real equivalent in Unix (the "edit" builtin opens an editor within Squeak, taking its input from files or from the output of a command pipeline such as "Smalltalk ! | edit" or "who | edit").

  3. If the command segment is not a shell builtin command, it is assumed to be an external command. It this case, the command line segment is fully parsed, then used to create an external OS process proxy to run the external command. The external process proxy responds to a #value message by running the external program, with output and error written to pipes connected to the command pipeline.

  4. Finally, if no external program can be found to execute the command, an error message is provided to the command window, and no command is evaluated.

Once the command proxy pipeline has been created, the individual process proxies are evaluated in such a way that they appear to execute in parallel, with each proxy reading its input from a pipe connected to its predecessor in the pipeline, and writing its output to a pipe connected to its successor in the pipeline. An additional error pipe is shared by all process proxies in the pipeline, such that any error output is accumulated in the shared error pipe stream.

As the command pipeline is created from the command line string, each process proxy sends a #value message to its predecessor cause it to be evaluated (internal process proxies) or to create and run its external program (external process proxies). The chain of process proxies is connected by pipes for the connected input and output streams, and has a single shared pipe to collect error streams from all the process proxies.

A command shell can obtain the output and error of a complex command pipeline simply by evaluating the last process proxy in the chain, and reading up to the end of its output and error pipes. Since a process proxy knows its run state, and a pipe is not "at end" until a process proxy closes one end of the pipe, the pipeline can be constructed in such a way that the command shell is assured that all process proxies have completed their evaluation when the output of the last proxy is read up to the end of the pipe.

As the last process proxy in a command pipeline is evaluated, CommandShell reads its output and error pipes, and displays the text in the View or Morph which it uses as its terminal window. When complete, it issues a new prompt string, and waits for another command line to be entered by the user.

CommandShell exercise