Anatomy of a shell CLI
 | This section needs additional citations for verification. (July 2015) (Learn how and when to remove this template message) |
The general pattern of an OS command line interface is:
Prompt command param1 param2 param3 … paramN
- Prompt — generated by the program to provide context for the client.
- Command — provided by the client. Commands are usually one of three classes:
- Internal commands are recognized and processed by the command line interpreter itself and not dependent upon any external executable file.
- Included commands run separate executable files generally considered part of the operating environment and always included with the OS.
- External commands run executable files that are not part of the basic OS, but added by other parties for specific purposes and applications.
- param1 …paramN — Optional parameters provided by the client. The format and meaning of the parameters depends upon the command issued. In the case of Included or External commands, the values of the parameters are delivered to the program (specified by the Command) as it is launched by the OS. Parameters may be either Arguments or Options.
In this example, the delimiters between command-line elements are whitespace characters and the end-of-line delimiter is the newline delimiter. This is a widely used (but not universal) convention for command-line interfaces.
A CLI can generally be considered as consisting of syntax and semantics. The syntax is the grammar that all commands must follow. In the case of operating systems, DOS and Unix each define their own set of rules that all commands must follow. In the case of embedded systems, each vendor, such as Nortel, Juniper Networks or Cisco Systems, defines their own proprietary set of rules that all commands within their CLI conform to. These rules also dictate how a user navigates through the system of commands. The semantics define what sort of operations are possible, on what sort of data these operations can be performed, and how the grammar represents these operations and data—the symbolic meaning in the syntax.
Two different CLIs may agree on either syntax or semantics, but it is only when they agree on both that they can be considered sufficiently similar to allow users to use both CLIs without needing to learn anything, as well as to enable re-use of scripts.
A simple CLI will display a prompt, accept a "command line" typed by the user terminated by the Enter key, then execute the specified command and provide textual display of results or error messages. Advanced CLIs will validate, interpret and parameter-expand the command line before executing the specified command, and optionally capture or redirect its output.
Unlike a button or menu item in a GUI, a command line is typically self-documenting, stating exactly what the user wants done. In addition, command lines usually include many defaults that can be changed to customize the results. Useful command lines can be saved by assigning a character string or alias to represent the full command, or several commands can be grouped to perform a more complex sequence – for instance, compile the program, install it, and run it — creating a single entity, called a command procedure or script which itself can be treated as a command. These advantages mean that a user must figure out a complex command or series of commands only once, because they can be saved, to be used again.
The commands given to a CLI shell are often in one of the following forms:
- doSomething how toFiles
- doSomething how sourceFile destinationFile
- doSomething how < inputFile > outputFile
- doSomething how | doSomething how | doSomething how > outputFile
where doSomething is, in effect, a verb, how an adverb (for example, should the command be executed "verbosely" or "quietly") and toFiles an object or objects (typically one or more files) on which the command should act. The > in the third example is a redirection operator, telling the command-line interpreter to send the output of the command not to its own standard output (the screen) but to the named file. This will overwrite the file. Using >> will redirect the output and append it to the file. Another redirection operator is the vertical bar (|), which creates a pipeline where the output of one command becomes the input to the next command.
CLI and resource protectionedit
One can modify the set of available commands by modifying which paths appear in the PATH environment variable. Under Unix, commands also need be marked as executable files. The directories in the path variable are searched in the order they are given. By re-ordering the path, one can run e.g. \OS2\MDOS\E.EXE instead of \OS2\E.EXE, when the default is the opposite. Renaming of the executables also works: people often rename their favourite editor to EDIT, for example.
The command line allows one to restrict available commands, such as access to advanced internal commands. The Windows CMD.EXE does this. Often, shareware programs will limit the range of commands, including printing a command 'your administrator has disabled running batch files' from the prompt.
Some CLIs, such as those in network routers, have a hierarchy of modes, with a different set of commands supported in each mode. The set of commands are grouped by association with security, system, interface, etc. In these systems the user might traverse through a series of sub-modes. For example, if the CLI had two modes called interface and system, the user might use the command interface to enter the interface mode. At this point, commands from the system mode may not be accessible until the user exits the interface mode and enters the system mode.
Command promptedit
A command prompt (or just prompt) is a sequence of (one or more) characters used in a command-line interface to indicate readiness to accept commands. It literally prompts the user to take action. A prompt usually ends with one of the characters $, %, #, :, > or - and often includes other information, such as the path of the current working directory and the hostname.
On many Unix and derivative systems, the prompt commonly ends in $ or % if the user is a normal user, but in # if the user is a superuser ("root" in Unix terminology).
End-users can often modify prompts. Depending on the environment, they may include colors, special characters, and other elements (like variables and functions for the current time, user, shell number or working directory) in order, for instance, to make the prompt more informative or visually pleasing, to distinguish sessions on various machines, or to indicate the current level of nesting of commands. On some systems, special tokens in the definition of the prompt can be used to cause external programs to be called by the command-line interpreter while displaying the prompt.
In DOS' COMMAND.COM and in Windows NT's cmd.exe users can modify the prompt by issuing a PROMPT command or by directly changing the value of the corresponding %PROMPT% environment variable. The default of most modern systems, the C:\> style is obtained, for instance, with PROMPT $P$G. The default of older DOS systems, C> is obtained by just PROMPT, although on some systems this produces the newer C:\> style, unless used on floppy drives A: or B:; on those systems PROMPT $N$G can be used to override the automatic default and explicitly switch to the older style.
Many Unix systems feature the $PS1 variable (Prompt String 1), although other variables also may affect the prompt (depending on the shell used). In the bash shell, a prompt of the form:
could be set by issuing the command
In zsh the $RPROMPT variable controls an optional "prompt" on the right-hand side of the display. It is not a real prompt in that the location of text entry does not change. It is used to display information on the same line as the prompt, but right-justified.
In RISC OS the command prompt is a * symbol, and thus (OS)CLI commands are often referred to as "star commands". One can also access the same commands from other command lines (such as the BBC BASIC command line), by preceding the command with a *.
Argumentsedit
A command-line argument or parameter is an item of information provided to a program when it is started. A program can have many command-line arguments that identify sources or destinations of information, or that alter the operation of the program.
When a command processor is active a program is typically invoked by typing its name followed by command-line arguments (if any). For example, in Unix and Unix-like environments, an example of a command-line argument is:
"file.s" is a command-line argument which tells the program rm to remove the file "file.s".
Some programming languages, such as C, C++ and Java, allow a program to interpret the command-line arguments by handling them as string parameters in the main function. Other languages, such as Python, expose operating system specific API (functionality) through sys module, and in particular sys.argv for "command-line arguments".
In Unix-like operating systems, a single hyphen used in place of a file name is a special value specifying that a program should handle data coming from the standard input or send data to the standard output.
Command-line optionedit
A command-line option or simply option (also known as a flag or switch) modifies the operation of a command; the effect is determined by the command's program. Options follow the command name on the command line, separated by spaces. A space before the first option is not always required, such as Dir/? and DIR /? in DOS, which have the same effect of listing the DIR command's available options, whereas dir --help (in many versions of Unix) does require the option to be preceded by at least one space (and is case-sensitive).
The format of options varies widely between operating systems. In most cases the syntax is by convention rather than an operating system requirement; the entire command line is simply a string passed to a program, which can process it in any way the programmer wants, so long as the interpreter can tell where the command name ends and its arguments and options begin.
A few representative samples of command-line options, all relating to listing files in a directory, to illustrate some conventions:
| Operating system | Command | Valid alternative | Notes |
|---|---|---|---|
| OpenVMS | directory/owner |
Dir /Owner |
instruct the directory command to also display the ownership of the files. Note the Directory command name is not case sensitive, and can be abbreviated to as few letters as required to remain unique. |
| Windows | DIR/Q/O:S d* |
dir /q d* /o:s |
display ownership of files whose names begin with "D", sorted by size, smallest first. Note spaces around argument d* are required. |
| Unix-like systems | ls -lS D* |
ls -S -l D* |
display in long format files and directories beginning with "D" (but not "d"), sorted by size (largest first). Note spaces are required around all arguments and options, but some can be run together, e.g. -lS is the same as -l -S. |
| Data General RDOS CLI | list/e/s 04-26-80/b |
List /S/E 4-26-80/B |
list every attribute for files created before 26 April 1980. Note the /B at the end of the date argument is a local switch, that modifies the meaning of that argument, while /S and /E are global switches, i.e. apply to the whole command. |
Abbreviating commandsedit
In Multics, command-line options and subsystem keywords may be abbreviated. This idea appears to derive from the PL/I programming language, with its shortened keywords (e.g., STRG for STRINGRANGE and DCL for DECLARE). For example, in the Multics "forum" subsystem, the -long_subject parameter can be abbreviated -lgsj. It is also common for Multics commands to be abbreviated, typically corresponding to the initial letters of the words that are strung together with underscores to form command names, such as the use of did for delete_iacl_dir.
In some other systems abbreviations are automatic, such as permitting enough of the first characters of a command name to uniquely identify it (such as SU as an abbreviation for SUPERUSER) while others may have some specific abbreviations pre-programmed (e.g. MD for MKDIR in COMMAND.COM) or user-defined via batch scripts and aliases (e.g. alias md mkdir in tcsh).
Option conventions in DOS, Windows, OS/2edit
On DOS, OS/2 and Windows, different programs called from their COMMAND.COM or CMD.EXE (or internal their commands) may use different syntax within the same operating system. For example:
- Options may be indicated by either of the "switch characters":
/,-, or either may be allowed. See below. - They may or may not be case-sensitive.
- Sometimes options and their arguments are run together, sometimes separated by whitespace, and sometimes by a character, typically
:or=; thusProg -fFilename,Prog -f Filename,Prog -f:Filename,Prog -f=Filename. - Some programs allow single-character options to be combined; others do not. The switch
-fAmay mean the same as-f -A, or it may be incorrect, or it may even be a valid but different parameter.
In DOS, OS/2 and Windows, the forward slash (/) is most prevalent, although the hyphen-minus is also sometimes used. In many versions of DOS (MS-DOS/PC DOS 2.xx and higher, all versions of DR-DOS since 5.0, as well as PTS-DOS, Embedded DOS, FreeDOS and RxDOS) the switch character (sometimes abbreviated switchar or switchchar) to be used is defined by a value returned from a system call (INT 21h/AH=37h). The default character returned by this API is /, but can be changed to a hyphen-minus on the above-mentioned systems, except for under Datalight ROM-DOS and MS-DOS/PC DOS 5.0 and higher, which always return / from this call (unless one of many available TSRs to reenable the SwitChar feature is loaded). In some of these systems (MS-DOS/PC DOS 2.xx, DOS Plus 2.1, DR-DOS 7.02 and higher, PTS-DOS, Embedded DOS, FreeDOS and RxDOS), the setting can also be pre-configured by a SWITCHAR directive in CONFIG.SYS. General Software's Embedded DOS provides a SWITCH command for the same purpose, whereas 4DOS allows the setting to be changed via SETDOS /W:n. Under DR-DOS, if the setting has been changed from /, the first directory separator \ in the display of the PROMPT parameter $G will change to a forward slash / (which is also a valid directory separator in DOS, FlexOS, 4680 OS, 4690 OS, OS/2 and Windows) thereby serving as a visual clue to indicate the change. Also, the current setting is reflected also in the built-in help screens. Some versions of DR-DOS COMMAND.COM also support a PROMPT token $/ to display the current setting. COMMAND.COM since DR-DOS 7.02 also provides a pseudo-environment variable named %/% to allow portable batchjobs to be written. Several external DR-DOS commands additionally support an environment variable %SWITCHAR% to override the system setting.
However, many programs are hardwired to use / only, rather than retrieving the switch setting before parsing command-line arguments. A very small number, mainly ports from Unix-like systems, are programmed to accept "-" even if the switch character is not set to it (for example netstat and ping, supplied with Microsoft Windows, will accept the /? option to list available options, and yet the list will specify the "-" convention).
Option conventions in Unix-like systemsedit
| This section does not cite any sources. (April 2015) (Learn how and when to remove this template message) |
In Unix-like systems, the ASCII hyphen-minus begins options; the new (and GNU) convention is to use two hyphens then a word (e.g. --create) to identify the option's use while the old convention (and still available as an option for frequently-used options) is to use one hyphen then one letter (e.g., -c); if one hyphen is followed by two or more letters it may mean two options are being specified, or it may mean the second and subsequent letters are a parameter (such as filename or date) for the first option.
Two hyphen-minus characters without following letters (--) may indicate that the remaining arguments should not be treated as options, which is useful for example if a file name itself begins with a hyphen, or if further arguments are meant for an inner command (e.g., sudo). Double hyphen-minuses are also sometimes used to prefix "long options" where more descriptive option names are used. This is a common feature of GNU software. The getopt function and program, and the getopts command are usually used for parsing command-line options.
Unix command names, arguments and options are case-sensitive (except in a few examples, mainly where popular commands from other operating systems have been ported to Unix).
Option conventions in other systemsedit
FlexOS, 4680 OS and 4690 OS use -.
CP/M typically used .
Conversational Monitor System (CMS) uses a single left parenthesis to separate options at the end of the command from the other arguments. For example, in the following command the options indicate that the target file should be replaced if it exists, and the date and time of the source file should be retained on the copy:
COPY source file a target file b (REPLACE OLDDATE
Data General's CLI under their RDOS, AOS, etc. operating systems, as well as the version of CLI that came with their Business Basic, uses only / as the switch character, is case-insensitive, and allows "local switches" on some arguments to control the way they are interpreted, such as MAC/U LIB/S A B C $LPT/L has the global option "U" to the macro assembler command to append user symbols, but two local switches, one to specify LIB should be skipped on pass 2 and the other to direct listing to the printer, $LPT.
Built-in usage helpedit
One of the criticisms of a CLI is the lack of cues to the user as to the available actions.citation needed In contrast, GUIs usually inform the user of available actions with menus, icons, or other visual cues.citation needed To overcome this limitation, many CLI programs display a brief summary of its valid parameters, typically when invoked with no arguments or one of ?, -?, -h, -H, /?, /h, /H, /Help, -help, or --help.
However, entering a program name without parameters in the hope that it will display usage help can be hazardous, as programs and scripts for which command line arguments are optional will execute without further notice.
Although desirable at least for the help parameter, programs may not support all option lead-in characters exemplified above.
Under DOS, where the default command-line option character can be changed from / to -, programs may query the SwitChar API in order to determine the current setting. So, if a program is not hardwired to support them all, a user may need to know the current setting even to be able to reliably request help.
If the SwitChar has been changed to - and therefore the / character is accepted as alternative path delimiter also at the DOS command line, programs may misinterpret options like /h or /H as paths rather than help parameters. However, if given as first or only parameter, most DOS programs will, by convention, accept it as request for help regardless of the current SwitChar setting.
In some cases, different levels of help can be selected for a program. Some programs supporting this allow to give a verbosity level as an optional argument to the help parameter (as in /H:1, /H:2, etc.) or they give just a short help on help parameters with question mark and a longer help screen for the other help options.
Depending on the program, additional or more specific help on accepted parameters is sometimes available by either providing the parameter in question as an argument to the help parameter or vice versa (as in /H:W or in /W:? (assuming /W would be another parameter supported by the program)).nb
In a similar fashion to the help parameter, but much less common, some programs provide additional information about themselves (like mode, status, version, author, license or contact information) when invoked with an "about" parameter like -!, /!, -about, or --about.
Since the ? and ! characters typically also serve other purposes at the command line, they may not be available in all scenarios, therefore, they should not be the only options to access the corresponding help information.
If more detailed help is necessary than provided by a program's built-in internal help, many systems support a dedicated external "help command" command (or similar), which accepts a command name as calling parameter and will invoke an external help system.
In the DR-DOS family, typing /? or /H at the COMMAND.COM prompt instead of a command itself will display a dynamically generated list of available internal commands; 4DOS and NDOS support the same feature by typing ? at the prompt (which is also accepted by newer versions of DR-DOS COMMAND.COM); internal commands can be individually disabled or reenabled via SETDOS /I. In addition to this, some newer versions of DR-DOS COMMAND.COM also accept a ?% command to display a list of available built-in pseudo-environment variables. Besides their purpose as quick help reference this can be used in batchjobs to query the facilities of the underlying command-line processor.
Command description syntaxedit
Built-in usage help and man pages commonly employ a small syntax to describe the valid command form:nb
- angle brackets for required parameters:
ping <hostname> - square brackets for optional parameters:
mkdir -p <dirname> - ellipses for repeated items:
cp <source1> source2… <dest> - vertical bars for choice of items:
netstat {-t|-u}
Notice that these characters have different meanings than when used directly in the shell. Angle brackets may be omitted when confusing the parameter name with a literal string is not likely.
The space characteredit
In many areas of computing, but particularly in the command line, the space character can cause problems as it has two distinct and incompatible functions: as part of a command or parameter, or as a parameter or name separator. Ambiguity can be prevented either by prohibiting embedded spaces in file and directory names in the first place (for example, by substituting them with underscores _), or by enclosing a name with embedded spaces between quote characters or using an escape character before the space, usually a backslash (\). For example
Long path/Long program name Parameter one Parameter two…
is ambiguous (is "program name" part of the program name, or two parameters?); however
Long_path/Long_program_name Parameter_one Parameter_two…,LongPath/LongProgramName ParameterOne ParameterTwo…,"Long path/Long program name" "Parameter one" "Parameter two"…
and
Long\ path/Long\ program\ name Parameter\ one Parameter\ two…
are not ambiguous. Unix-based operating systems minimize the use of embedded spaces to minimize the need for quotes. In Microsoft Windows, one often has to use quotes because embedded spaces (such as in directory names) are common.
Comments
Post a Comment