tclreadline(3) gnu readline for the tcl scripting language


::tclreadline::readline command [options]


The tclreadline package makes the gnu readline available to the scripting language tcl. The primary purpose of the package is to facilitate the interactive script development by the means of word and file name completion as well as history expansion (well known from shells like bash).

Additionally tclreadline can also be used for tcl scripts which want to use a shell like input interface. In this case the ::tclreadline::readline read command has to be called explicitly.

The advantage of tclreadline is that it uses the callback handler mechanism of the gnu readline while it processes tcl events. This way X events from a wish gui will processed as well as events from the tclreadline line interface.

tclreadline is basically a shared library and a few tcl scripts which are accessed with the tcl package require command. Therefore tclreadline should work with all common extensions like blt, itcl, itk, tix ....


If you want to use tclreadline as a line interface for developing tcl scripts, you probably don't have to read this section. In this case the only thing you should do is to modify your .tclshrc according to the section FILES.

For the functionality of the GNU readline you should refer to the readline's documentation.

The following list will give all commands, which are currently implemented in the shared lib (e.g. Additional commands were introduced in a startup script tclreadlineSetup.tcl, which lives in the tclreadline installation directory. (typically something like /usr/local/lib/tclreadline ..) These commands are primarily for internal use and not documented here.

Note that all commands reside in the namespace ::tclreadline::.

::tclreadline::readline add string
adds a string to the completer. If the string contains white spaces, each of the words will be completed consecutively when hitting <Tab>. Example:

    ::tclreadline::readline add "button pathName ?options?"

typing but<Tab> will complete to button. Hitting <Tab> again will complete to "button pathName". ...

::tclreadline::readline complete string
returns 1 if string is a complete tcl command and 0 otherwise.

::tclreadline::readline customcompleter [string]
Register the proc string as custom completer. This proc is called with exactly four arguments each time completion takes place: the word to complete ("text"), the "start" and "end" positions of this word in the line entered so far, and this line ("line"). The custom completion script should return an array of strings which is a list of completions for "text". If there are no completions, it should return an empty string "". The first entry in the returned list is the substitution for "text". The remaining entries are the possible completions. If the custom completion script returns an empty string and builtin completion is enabled (see tclreadline::readline builtincompleter), the builtin completer is called. tclreadline::readline customcompleter simply returns the current custom completer if called w/o string. To turn of custom completion, call tclreadline::readline customcompleter with an empty string.

Example: % puts $b<TAB> will call the custom completer with the four arguments "$b", "5", "8" and "puts $b". The custom completer could return a string like "$bl $black $blue", which will complete "$b" to "$bl" (the longest match) and offer a list of two further matches "$black" and "$blue".

For further reference, see the proc tclreadline::ScriptCompleter in the file tclreadlineSetup.tcl.

::tclreadline::readline builtincompleter [bool]
enable or disable the builtin completer. If the builtin completer is enabled, it will be invoked either if there is no custom completer, or the custom completer returned an empty string. The builtin completer is on by default. tclreadline::readline builtincompleter returns the current custom completer (also, if called w/o the bool argument).

::tclreadline::readline eofchar [script]
set a script which will be called, if readline returns the eof character (this is typically the case if CTRL-D is entered at the very beginning of the line). The default for this script is "puts {}; exit". Setting this to an empty value disables any action on eof. tclreadline::readline eof returns the current eof script.

::tclreadline::readline initialize historyfile
initialize the tclreadline interface and read the history from the historyfile. On success an empty string is returned. This command has to be called before any other tclreadline commands.

::tclreadline::readline read prompt
prints the prompt to stdout and enters the tclreadline event loop. Both readline and X events are processed. Returns the (eventually history-expanded) input string. tclreadline::readline read rises an error, if an error occurs while evaluating a script completer.

::tclreadline::readline write historyfile
writes the history to the historyfile. This command is called automatically from the internal routine ::tclreadline::Exit. If the variable tclreadline::historyLength is non-negative, the historyfile will be truncated to hold only this number lines.

::tclreadline::readline reset-terminal [terminalName]
w/o argument: reset the state of the terminal to what it was before tclreadline was used. With argument: reinitialize readline's idea of the terminal settings using terminalName as the terminal type. The form w/o argument might not work if tclreadline was compiled with an older version of libreadline.

::tclreadline::readline bell
Ring the terminal bell, obeying the setting of bell-style -- audible or visible.

Return the current input.

::tclreadline::readline update
Redraw the current input line.

::tclreadline::Print [yes / no]
turns on or off the default behavior of tclsh to print the result of every command. This is turned on by default, so it will just behave as the tclsh w/o tclreadline. Turning off might be useful, when reading binary data for example. If ::tclreadline::Print is called w/o arguments, it returns the current setting.

::tclreadline::Loop [historyfile]
enter the tclreadline main loop. This command is typically called from the startup resource file (something .tclshrc, depending on the interpreter you use, see the file `sample.tclshrc'). The main loop sets up some completion characteristics as variable -- try something like "puts $b<TAB>" -- and command completion -- try "puts [in<TAB>". If the optional argument historyfile is given, this file will be used for reading and writing the command history instead of the default .tclsh-history. ::tclreadline::Loop will normally not return. If you want to write your own main loop and/or own custom completers, it is probably a good idea to start with tclreadline::Loop (see the file tclreadlineSetup.tcl).

a proc which is called by ::tclreadline::Loop and returns a string which will be displayed as the primary prompt. This prompt will be something like "[info nameofexecutable] \[[pwd]\]" possibly fancy colored. The default proc is defined on entering the ::tclreadline::Loop, if it is not already defined. So: If you define your own proc ::tclreadline::prompt1 before entering ::tclreadline::Loop, this proc is called each time the prompt is to be displayed. Example:
    package require tclreadline
    namespace eval tclreadline {
        proc prompt1 {} {
            return "[clock format [clock seconds]]> "

Note that non-printable control characters as color control characters must be enclosed in literal ctrl-a / ctrl-b to tell readline the length of the printable prompt. See for example the variable `prompt_string' in the file tclreadlineSetup.tcl in your tclreadline installation directory.

a proc which is called by ::tclreadline::Loop and returns a string which will be displayed as the secondary prompt when interactively prompting for continuation of an incomplete command.


tclreadline defines the following variables in the namespace ::tclreadline: (for backwards compatibility the global variables tclreadline_version,
 tclreadline_patchLevel and tclreadline_library are still present).

tclreadline::version (read-only)
holds the version string "2.1.0".

tclreadline::patchLevel (read-only)
holds the patch level string "2.1.0".

tclreadline::library (read-only)
holds the library directory string.

tclreadline::license (read-only)
holds a BSD license statement.

Number of lines, which will be written to the historyfile. This number is -1 by default, which means that the historyfile will not be truncated. See also tclreadline::write.


the .tclshrc file in the HOME directory, which is read on tclsh startup. Alternatively, the name of this initialization file might be .wishrc ... depending on what interpreter you use. These files should typically contain something like

    if {$tcl_interactive} {
        package require tclreadline

which will enter the tclreadline main loop.

the .tclsh-history file in the HOME directory. On startup commands will be read from this file. On exit, the readline history is written to this file. Note that if you abort tclsh with <cntrl-c> no history is written. For the future it is planned to set up a signal handler, which will write the history on <ctrl-c> before exiting.

the .inputrc file in the users HOME directory. This file is used normally for all programs which use the gnu readline (e.g. bash). The `global' readline settings there will be valid also for tclreadline. Additionally the .inputrc might hold conditional settings for the implementation name tclreadline. Example of some lines in your .inputrc:

    $if tclreadline
    "\C-xp": "puts $env(PATH)"

For further documentation please refer to the gnu readline documentation.


citing the readline manual: It's too big and too slow.


Johannes Zellner, <[email protected]>


Magnus Eriksson <[email protected]>, Les Johnson <[email protected]>, Harald Kirsch <[email protected]>, Christian Krone <[email protected]>, Larry W. Virden <[email protected]>, David Engel <[email protected]> <[email protected]>, Matthew Clarke <[email protected]>


David Engel <[email protected]>, <[email protected]>


tclreadline comes with a BSD type license. The read-only variable tclreadline::license holds the complete license statement.