Comparison of command shells

Background execution allows a shell to run a command without user interaction in the terminal, freeing the command line for additional work with the shell. POSIX shells and other Unix shells allow background execution by using the & character at the end of command.

Completion features assist the user in typing commands at the command line, by looking for and suggesting matching words for incomplete ones. Completion is generally requested by pressing the completion key (often the Tab ↹ key).

Command name completion is the completion of the name of a command. In most shells, a command can be a program in the command path (usually $PATH), a builtin command, a function or alias.

Path completion is the completion of the path to a file, relative or absolute.

Wildcard completion is a generalization of path completion, where an expression matches any number of files, using any supported syntax for file matching.

Variable completion is the completion of the name of a variable name (environment variable or shell variable). Bash, zsh, and fish have completion for all variable names. PowerShell has completions for environment variable names, shell variable names and — from within user-defined functions — parameter names.

Command argument completion is the completion of a specific command's arguments. There are two types of arguments, named and positional: Named arguments, often called options, are identified by their name or letter preceding a value, whereas positional arguments consist only of the value. Some shells allow completion of argument names, but few support completing values.

Bash, zsh and fish offer parameter name completion through a definition external to the command, distributed in a separate completion definition file. For command parameter name/value completions, these shells assume path/filename completion if no completion is defined for the command. Completion can be set up to dynamically suggest completions by calling a shell function.^[43] The fish shell additionally supports parsing of man pages to extract parameter information that can be used to improve completions/suggestions. In PowerShell, all types of commands (cmdlets, functions, script files) inherently expose data about the names, types and valid value ranges/lists for each argument. This metadata is used by PowerShell to automatically support argument name and value completion for built-in commands/functions, user-defined commands/functions as well as for script files. Individual cmdlets can also define dynamic completion of argument values where the completion values are computed dynamically on the running system.

Users of a shell may find themselves typing something similar to what they have typed before. Support for command history means that a user can recall a previous command into the command-line editor and edit it before issuing the potentially modified command.

Shells that support completion may also be able to directly complete the command from the command history given a partial/initial part of the previous command.

Most modern shells support command history. Shells which support command history in general also support completion from history rather than just recalling commands from the history. In addition to the plain command text, PowerShell also records execution start- and end time and execution status in the command history.

Mandatory arguments/parameters are arguments/parameters which must be assigned a value upon invocation of the command, function or script file. A shell that can determine ahead of invocation that there are missing mandatory values, can assist the interactive user by prompting for those values instead of letting the command fail. Having the shell prompt for missing values will allow the author of a script, command or function to mark a parameter as mandatory instead of creating script code to either prompt for the missing values (after determining that it is being run interactively) or fail with a message.

Shells featuring automatic suggestions display optional command-line completions as the user types. The PowerShell and fish shells natively support this feature; pressing the Tab ↹ key inserts the completion.

Implementations of this feature can differ between shells; for example, PowerShell^[44] and zsh^[45] use an external module to provide completions, and fish derives its completions from the user's command history.^[46]

Shells may record a history of directories the user has been in and allow for fast switching to any recorded location. This is referred to as a "directory stack". The concept had been realized as early as 1978^[47] in the release of the C shell (csh).

Command line interpreters 4DOS and its graphical successor Take Command Console also feature a directory stack.

A directory name can be used directly as a command which implicitly changes the current location to the directory.

This must be distinguished from an unrelated load drive feature supported by Concurrent DOS, Multiuser DOS, System Manager and REAL/32, where the drive letter L: will be implicitly updated to point to the load path of a loaded application, thereby allowing applications to refer to files residing in their load directory under a standardized drive letter instead of under an absolute path.^[48]

When a command line does not match a command or arguments directly, spell checking can automatically correct common typing mistakes (such as case sensitivity, missing letters). There are two approaches to this; the shell can either suggest probable corrections upon command invocation, or this can happen earlier as part of a completion or autosuggestion.

The tcsh and zsh shells feature optional spell checking/correction, upon command invocation.

Fish does the autocorrection upon completion and autosuggestion. The feature is therefore not in the way when typing out the whole command and pressing enter, whereas extensive use of the tab and right-arrow keys makes the shell mostly case insensitive.

The PSReadLine^[31] PowerShell module (which is shipped with version 5.0) provides the option to specify a CommandValidationHandler ScriptBlock which runs before submitting the command. This allows for custom correcting of commonly mistyped commands, and verification before actually running the command.

A shell script (or job) can report progress of long running tasks to the interactive user.

Unix/Linux systems may offer other tools support using progress indicators from scripts or as standalone-commands, such as the program "pv".^[49] These are not integrated features of the shells, however.

JP Software command-line processors provide user-configurable colorization of file and directory names in directory listings based on their file extension and/or attributes through an optionally defined %COLORDIR% environment variable.

For the Unix/Linux shells, this is a feature of the ls command and the terminal.

The command line processors in DOS Plus, Multiuser DOS, REAL/32 and in all versions of DR-DOS support a number of optional environment variables to define escape sequences allowing to control text highlighting, reversion or colorization for display or print purposes in commands like TYPE. All mentioned command line processors support %$ON% and %$OFF%. If defined, these sequences will be emitted before and after filenames. A typical sequence for %$ON% would be \033[1m in conjunction with ANSI.SYS, \033p for an ASCII terminal or \016 for an IBM or ESC/P printer. Likewise, typical sequences for %$OFF% would be \033[0m, \033q, \024, respectively. The variables %$HEADER% and %$FOOTER% are only supported by COMMAND.COM in DR-DOS 7.02 and higher to define sequences emitted before and after text blocks in order to control text highlighting, pagination or other formatting options.

For the Unix/Linux shells, this is a feature of the terminal.

A defining feature of the fish shell is built-in syntax highlighting, As the user types, text is colored to represent whether the input is a valid command or not (the executable exists and the user has permissions to run it), and valid file paths are underlined.^[50]

An independent project offers syntax highlighting as an add-on to the Z Shell (zsh).^[51] This is not part of the shell, however.

PowerShell provides customizable syntax highlighting on the command line through the PSReadLine^[31] module. This module can be used with PowerShell v3.0+, and is bundled with v5.0 onwards. It is loaded by default in the command line host "powershell.exe" since v5.0.^[52]

Take Command Console (TCC) offers syntax highlighting in the integrated environment.

4DOS, 4OS2, 4NT / Take Command Console and PowerShell (in PowerShell ISE) looks up context-sensitive help information when F1 is pressed.

Zsh provides various forms of configurable context-sensitive help as part of its run-help widget, _complete_help command, or in the completion of options for some commands.

The fish shell provides brief descriptions of a command's flags during tab completion.

In anticipation of what a given running application may accept as keyboard input, the user of the shell instructs the shell to generate a sequence of simulated keystrokes, which the application will interpret as a keyboard input from an interactive user. By sending keystroke sequences the user may be able to direct the application to perform actions that would be impossible to achieve through input redirection or would otherwise require an interactive user. For example, if an application acts on keystrokes, which cannot be redirected, distinguishes between normal and extended keys, flushes the queue before accepting new input on startup or under certain conditions, or because it does not read through standard input at all. Keystroke stacking typically also provides means to control the timing of simulated keys being sent or to delay new keys until the queue was flushed etc. It also allows to simulate keys which are not present on a keyboard (because the corresponding keys do not physically exist or because a different keyboard layout is being used) and therefore would be impossible to type by a user.

This section may contain information not important or relevant to the article's subject. Please help improve this section. (July 2014) (Learn how and when to remove this message)

Some shell scripts need to query the user for sensitive information such as passwords, private digital keys, PIN codes or other confidential information. Sensitive input should not be echoed back to the screen/input device where it could be gleaned by unauthorized persons. Plaintext memory representation of sensitive information should also be avoided as it could allow the information to be compromised, e.g., through swap files, core dumps etc.^[68]

The shells bash, zsh and PowerShell offer this as a specific feature.^[69][70] Shells which do not offer this as a specific feature may still be able to turn off echoing through some other means. Shells executing on a Unix/Linux operating system can use the stty external command to switch off/on echoing of input characters.^[71] In addition to not echoing back the characters, PowerShell's -AsSecureString option also encrypts the input character-by-character during the input process, ensuring that the string is never represented unencrypted in memory where it could be compromised through memory dumps, scanning, transcription etc.

Some operating systems define an execute permission which can be granted to users/groups for a file when the file system itself supports it.

On Unix systems, the execute permission controls access to invoking the file as a program, and applies both to executables and scripts. As the permission is enforced in the program loader, no obligation is needed from the invoking program, nor the invoked program, in enforcing the execute permission – this also goes for shells and other interpreter programs. The behaviour is mandated by the POSIX C library that is used for interfacing with the kernel. POSIX specifies that the exec family of functions shall fail with EACCESS (permission denied) if the file denies execution permission (see execve – System Interfaces Reference, The Single UNIX Specification, Version 5 from The Open Group).

The execute permission only applies when the script is run directly. If a script is invoked as an argument to the interpreting shell, it will be executed regardless of whether the user holds the execute permission for that script.

Although Windows also specifies an execute permission, none of the Windows-specific shells block script execution if the permission has not been granted.

Several shells can be started or be configured to start in a mode where only a limited set of commands and actions is available to the user. While not a security boundary (the command accessing a resource is blocked rather than the resource) this is nevertheless typically used to restrict users' actions before logging in.

A restricted mode is part of the POSIX specification for shells, and most of the Linux/Unix shells support such a mode where several of the built-in commands are disabled and only external commands from a certain directory can be invoked.^[72][73]

PowerShell supports restricted modes through session configuration files or session configurations. A session configuration file can define visible (available) cmdlets, aliases, functions, path providers and more.^[74]

Scripts that invoke other scripts can be a security risk as they can potentially execute foreign code in the context of the user who launched the initial script. Scripts will usually be designed to exclusively include scripts from known safe locations; but in some instances, e.g. when offering the user a way to configure the environment or loading localized messages, the script may need to include other scripts/files.^[75] One way to address this risk is for the shell to offer a safe subset of commands which can be executed by an included script.