NAME

SYNOPSIS

INTRODUCTION

Extended Tcl was created by Karl Lehenbauer and Mark Diekhans and is freely redistributable for any use without license or fee.

Available since 1989, Extended Tcl, also known as TclX, not only adds capabilities to Tcl, but has also been the source of many of the capabilities of the baseline Tcl release, including arrays, files, sockets, file events, and date and time handling, among others.

Extended Tcl introduces a set of new commands and a user-extensible library of useful Tcl procedures, any of which can be automatically loaded on the first attempt to execute it.

The command descriptions are separated into several sections:

GENERAL COMMANDS

dirs

This procedure lists the directories in the directory stack.

commandloop ?-async? ?-interactive on | off | tty? ?-prompt1 cmd? ?-prompt2 cmd? ?-endcommand cmd?

Create an interactive command loop reading commands from stdin and writing results to stdout. Command loops are maybe either be blocking or event oriented. This command is useful for Tcl scripts that do not normally converse interactively with a user through a Tcl command interpreter, but which sometimes want to enter this mode, perhaps for debugging or user configuration. The command loop terminates on EOF.

The following options are available:

echo ?str ...?

Writes zero or more strings to standard output, followed by a newline.

infox option

Return information about Extended Tcl, or the current application. The following infox command options are available:

for_array_keys var array_name code

This procedure performs a foreach-style loop for each key in the named array. The break and continue statements work as with foreach.

for_recursive_glob var dirlist globlist code

This procedure performs a foreach-style loop over recursively matched files. All directories in dirlist are recursively searched (breadth-first), comparing each file found against the file glob patterns in globlist. For each matched file, the variable var is set to the file path and code is evaluated. Symbolic links are not followed.

loop var first limit ?increment? body

Loop is a looping command, similar in behavior to the Tcl for statement, except that the loop statement achieves substantially higher performance and is easier to code when the beginning and ending values of a loop are known, and the loop variable is to be incremented by a known, fixed amount every time through the loop.

The var argument is the name of a Tcl variable that will contain the loop index. The loop index is set to the value specified by first. The Tcl interpreter is invoked upon body zero or more times, where var is incremented by increment every time through the loop, or by one if increment is not specified. Increment can be negative in which case the loop will count downwards.

When var reaches limit, the loop terminates without a subsequent execution of body. For instance, if the original loop parameters would cause loop to terminate, say first was one, limit was zero and increment was not specified or was non-negative, body is not executed at all and loop returns.

The first, limit and increment are integer expressions. They are only evaluated once at the beginning of the loop.

If a continue command is invoked within body then any remaining commands in the current execution of body are skipped, as in the for command. If a break command is invoked within body then the loop command will return immediately. Loop returns an empty string.

popd

This procedure pops the top directory entry from the directory stack and make it the current directory.

pushd ?dir?

This procedure pushes the current directory onto the directory stack and cd to the specified directory. If the directory is not specified, then the current directory is pushed, but remains unchanged.

recursive_glob dirlist globlist

This procedure returns a list of recursively matches files. All directories in dirlist are recursively searched (breadth-first), comparing each file found against the file glob patterns in globlist. Symbolic links are not followed.

showproc ?procname ...?

This procedure lists the definition of the named procedures. Loading them if it is not already loaded. If no procedure names are supplied, the definitions of all currently loaded procedures are returned.

try_eval code catch ?finally?

The try_eval command evaluates code in the current context.

If an error occurs during the evaluation and catch is not empty, then catch is evaluated to handler the error. The result of the command, containing the error message, will be stored in a global variable errorResult. The global variables errorResult, errorInfo and errorCode will be imported into the current scope, there is no need to execute a global command. The result of the catch command becomes the result of the try_eval command. If the error that caused the catch to be evaluate is to be continued, the following command should be used:

	error $errorResult $errorCode $errorInfo

If the finally argument is supplied and not empty, it is evaluated after the evaluation of the code and the catch commands. If an error occurs during the evaluation of the finally command, it becomes the result of the try_eval command. Otherwise, the result of either code or catch is preserved, as described above.

DEBUGGING AND DEVELOPMENT COMMANDS

cmdtrace level | on ?noeval? ?notruncate? ?procs? ?fileid? ?command cmd?

Print a trace statement for all commands executed at depth of level or below (1 is the top level). If on is specified, all commands at any level are traced. The following options are available:

The most common use of this command is to enable tracing to a file during the development. If a failure occurs, a trace is then available when needed. Command tracing will slow down the execution of code, so it should be removed when code is debugged. The following command will enable tracing to a file for the remainder of the program:

	cmdtrace on [open cmd.log w]
    

The command option causes a user specified trace command to be called for each command executed. The command will have the following arguments appended to it before evaluation:

The command should be constructed in such a manner that it will work if additional arguments are added in the future. It is suggested that the command be a proc with the final argument being args.

Tracing will be turned off while the command is being executed. The values of the errorInfo and errorCode variables will be saved and restored on return from the command. It is the command's responsibility to preserve all other state.

If an error occurs during the execution of command, an error message is dumped to stderr and the tracing is disabled. The underlying mechanism that this functionality is built on does not support returning an error to the interpreter.

cmdtrace off

Turn off all tracing.

cmdtrace depth

Returns the current maximum trace level, or zero if trace is disabled.

edprocs ?proc...?

This procedure writes the named procedures, or all currently defined procedures, to a temporary file, then calls an editor on it (as specified by the EDITOR environment variable, or vi if none is specified), then sources the file back in if it was changed.

profile ?-commands? ?-eval? on

profile off arrayVar

This command is used to collect a performance profile of a Tcl script. It collects data at the Tcl procedure level. The number of calls to a procedure, and the amount of real and CPU time is collected. Time is also collected for the global context. The procedure data is collected by bucketing it based on the procedure call stack, this allows determination of how much time is spent in a particular procedure in each of it's calling contexts.

The on option enables profile data collection. If the -commands option is specified, data on all commands within a procedure is collected as well a procedures. Multiple occurrences of a command within a procedure are not distinguished, but this data may still be useful for analysis.

The off option turns off profiling and moves the data collected to the array arrayVar. The array is address by a list containing the procedure call stack. Element zero is the top of the stack, the procedure that the data is for. The data in each entry is a list consisting of the procedure call count and the real time and CPU time in milliseconds spent in the procedure (but not any procedures it calls). The list is in the form {count real cpu}.

Normally, the variable scope stack is used in reporting where time is spent. Thus upleveled code is reported in the context that it was executed in, not the context that the uplevel was called in. If the -eval option is specified, the procedure evaluation (call) stack is used instead of the procedure scope stack. Upleveled code is reported in the context of the procedure that did the uplevel.

A Tcl procedure profrep is supplied for reducing the data and producing a report.

On Windows, profile command only reports elapsed real time, CPU time is not available and is reported as zero.

profrep profDataVar sortKey ?outFile? ?userTitle?

This procedure generates a report from data collect from the profile command. ProfDataVar is the name of the array containing the data returned by the profile command. SortKey indicates which data value to sort by. It should be one of "calls", "cpu" or "real". OutFile is the name of file to write the report to. If omitted, stdout is assumed. UserTitle is an optional title line to add to output.

Listed with indentation below each procedure or command is the procedure call stack. The first indented line being the procedure that invoked the reported procedure or command. The next line is the procedure that invoked the procedure above it, and so on. If no indented procedures are shown, the procedure or command was called from the global context. Time actually spent in the global context is listed on a line labeled <global>. Upleveled code is reported in the context that it was executed in, not the context that the uplevel was called in.

saveprocs fileName ?proc...?

This procedure saves the definition of the named procedure, or all currently defined procedures if none is specified, to the named file.

UNIX ACCESS COMMANDS

alarm seconds

Instructs the system to send a SIGALRM signal in the specified number of seconds. This is a floating point number, so fractions of a section may be specified. If seconds is 0.0, any previous alarm request is canceled. Only one alarm at a time may be active; the command returns the number of seconds left in the previous alarm. On systems without the setitimer system call, seconds is rounded up to an integer number of seconds.

The alarm command is not available on Windows.

execl ?-argv0 argv0? prog ?arglist?

Do an execl, replacing the current program (either Extended Tcl or an application with Extended Tcl embedded into it) with prog and passing the arguments in the list arglist.

The -argv0 options specifies that argv0 is to be passed to the program as argv [0] rather than prog.

Note: If you are using execl in a Tk application and it fails, you may not do anything that accesses the X server or you will receive a BadWindow error from the X server. This includes executing the Tk version of the exit command. We suggest using the following command to abort Tk applications after an execl failure:

    kill [id process]
    

On Windows, where the fork command is not available, execl starts a new process and returns the process id.

chroot dirname

Change root directory to dirname, by invoking the POSIX chroot(2) system call. This command only succeeds if running as root.

fork

Fork the current Tcl process. Fork returns zero to the child process and the process number of the child to the parent process. If the fork fails, a Tcl error is generated.

If an execl is not going to be performed before the child process does output, or if a close and dup sequence is going to be performed on stdout or stderr, then a flush should be issued against stdout, stderr and any other open output file before doing the fork. Otherwise characters from the parent process pending in the buffers will be output by both the parent and child processes.

Note: If you are forking in a Tk based application you must execl before doing any window operations in the child or you will receive a BadWindow error from the X server.

The fork command is not available on Windows.

id options

This command provides a means of getting, setting and converting user, group and process ids. The id command has the following options:

kill ?-pgroup ?signal? idlist

Send a signal to the each process in the list idlist, if permitted. Signal, if present, is the signal number or the symbolic name of the signal, see the signal system call manual page. The leading ``SIG'' is optional when the signal is specified by its symbolic name. The default for signo is 15, SIGTERM.

If -pgroup is specified, the numbers in idlist are take as process group ids and the signal is sent to all of the process in that process group. A process group id of 0 specifies the current process group.

On Windows, the kill command is capable of terminating a process, but not of sending an arbitrary signal.

link ?-sym? srcpath destpath

Create a directory entry, destpath, linking it to the existing file, srcpath. If -sym is specified, a symbolic link, rather than a hard link, is created. (The -sym option is only available on systems that support symbolic links.)

The link command is not available on Windows. Use the Tcl 8.4+ file link command instead.

nice ?priorityincr?

Change or return the process priority. If priorityincr is omitted, the current priority is returned. If priorityincr is positive, it is added to the current priority level, up to a system defined maximum (normally 19),

Negative priorityincr values cumulatively increase the program's priority down to a system defined minimum (normally -19); increasing priority with negative niceness values will only work for the superuser.

The new priority is returned.

The nice command is not available on Windows.

readdir ?-hidden? dirPath

Returns a list containing the contents of the directory dirPath. The directory entries "." and ".." are not returned.

On Windows, -hidden maybe specified to include hidden files in the result. This flag is ignored on Unix systems.

signal ?-restart? action siglist ?command?

Warning: If signals are being used as an event source (a trap action), rather than generating an error to terminate a task; one must use the -restart option. This causes a blocked system call, such as read or waitpid to be restarted rather than generate an error. Failure to do this may results in unexpected errors when a signal arrives while in one of these system calls. When available, the -restart option can prevent this problem.

If -restart is specified, restart blocking system calls rather than generating an error. The signal will be handled once the Tcl command that issued the system call completes. The -restart options is not available on all operating systems and its use will generate an error when it is not supported. Use infox have_signal_restart to check for availability.

Specify the action to take when a Unix signal is received by Extended Tcl, or a program that embeds it. Siglist is a list of either the symbolic or numeric Unix signal (the SIG prefix is optional). Action is one of the following actions to be performed on receipt of the signal. To specify all modifiable signals, use `*' (this will not include SIGKILL and SIGSTOP, as they can not be modified).

The signal action will remain enabled after the specified signal has occurred. The exception to this is SIGCHLD on systems without Posix signals. For these systems, SIGCHLD is not be automatically reenabled. After a SIGCHLD signal is received, a call to wait must be performed to retrieve the exit status of the child process before issuing another signal SIGCHLD ... command. For code that is to be portable between both types of systems, use this approach.

Signals are not processed until after the completion of the Tcl command that is executing when the signal is received. If an interactive Tcl shell is running, then the SIGINT will be set to error, non-interactive Tcl sessions leave SIGINT unchanged from when the process started (normally default for foreground processes and ignore for processes in the background).

sleep seconds

Sleep the Extended Tcl process for seconds seconds. Seconds, if specified as a decimal number, is truncated to an integer value.

system cmdstr1 ?cmdstr2...?

Concatenates cmdstr1, cmdstr2 etc with space separators (see the concat command) into a single command and then evaluates the command using the standard system shell. On Unix systems, this is /bin/sh and on Windows its command.com. The exit code of the command is returned.

This command differs from the exec command in that system doesn't return the executed command's standard output as the result string, and system goes through the Unix shell to provide wild card expansion, redirection, etc, as is normal from an sh command line.

sync ?fileId?

If fileId is not specified, or if it is and this system does not support the fsync system call, issues a sync system call to flush all pending disk output. If fileId is specified and the system does support the fsync system call, issues an fsync on the file corresponding to the specified Tcl fileId to force all pending output to that file out to the disk.

If fileId is specified, the file must be writable. A flush will be issued against the fileId before the sync.

The infox have_fsync command can be used to determine if "sync fileId" will do a sync or a fsync.

times

Return a list containing the process and child execution times in the form:

	utime stime cutime cstime
    

Also see the times(2) system call manual page. The values are in milliseconds.

umask ?octalmask?

Sets file-creation mode mask to the octal value of octalmask. If octalmask is omitted, the current mask is returned.

wait ?-nohang? ?-untraced? ?-pgroup? ?pid?

Waits for a process created with the execl command to terminate, either due to an untrapped signal or call to exit system call. If the process id pid is specified, they wait on that process, otherwise wait on any child process to terminate.

If -nohang is specified, then don't block waiting on a process to terminate. If no process is immediately available, return an empty list. If -untraced is specified then the status of child processes that are stopped, and whose status has not yet been reported since they stopped, are also returned. If -pgroup is specified and pid is not specified, then wait on any child process whose process group ID is they same as the calling process. If pid is specified with -pgroup, then it is take as a process group ID, waiting on any process in that process group to terminate.

Wait returns a list containing three elements: The first element is the process id of the process that terminated. If the process exited normally, the second element is `EXIT', and the third contains the numeric exit code. If the process terminated due to a signal, the second element is `SIG', and the third contains the signal name. If the process is currently stopped (on systems that support SIGSTP), the second element is `STOP', followed by the signal name.

Note that it is possible to wait on processes to terminate that were create in the background with the exec command. However, if any other exec command is executed after the process terminates, then the process status will be reaped by the exec command and will not be available to the wait command.

On systems without the waitpid system call, the -nohang, -untraced and -pgroup options are not available. The infox have_waitpid command maybe use to determine if this functionality is available.

FILE COMMANDS

It should be noted that Tcl file I/O is implemented on top of the stdio library. By default, the file is buffered. When communicating to a process through a pipe, a flush command should be issued to force the data out. Alternatively, the fcntl command may be used to set the buffering mode of a file to line-buffered or unbuffered.

bsearch fileId key ?retvar? ?compare_proc?

Search an opened file fileId containing lines of text sorted into ascending order for a match. Key contains the string to match. If retvar is specified, then the line from the file is returned in retvar, and the command returns 1 if key was found, and 0 if it wasn't. If retvar is not specified or is a null name, then the command returns the line that was found, or an empty string if key wasn't found.

By default, the key is matched against the first white-space separated field in each line. The field is treated as an ASCII string. If compare_proc is specified, then it defines the name of a Tcl procedure to evaluate against each line read from the sorted file during the execution of the bsearch command. Compare_proc takes two arguments, the key and a line extracted from the file. The compare routine should return a number less than zero if the key is less than the line, zero if the key matches the line, or greater than zero if the key is greater than the line. The file must be sorted in ascending order according to the same criteria compare_proc uses to compare the key with the line, or erroneous results will occur.

This command does not work on files containing binary data (bytes of zero).

chmod [-fileid] mode filelist

Set permissions of each of the files in the list filelist to mode, where mode is an absolute numeric mode or symbolic permissions as in the UNIX chmod(1) command. To specify a mode as octal, it should be prefixed with a "0" (e.g. 0622).

If the option -fileid is specified, filelist is a list of open file identifiers rather than a list of file names. This option is not available on all Unix systems. Use the infox have_fchmod command to determine if this functionality is available.

The chmod command is not available on Windows.

chown [-fileid] owner | {owner group} filelist

Set owner of each file in the list filelist to owner, which can be a user name or numeric user id. If the first parameter is a list, then the owner is set to the first element of the list and the group is set to the second element. Group can be a group name or numeric group id. If group is {}, then the file group will be set to the login group of the specified user.

If the option -fileid is specified, filelist is a list of open file identifiers rather than a list of file names. This option is not available on all Unix systems. Use the infox have_fchown command to determine if this functionality is available.

The chown command is not available on Windows.

chgrp [-fileid] group filelist

Set the group id of each file in the list filelist to group, which can be either a group name or a numeric group id.

If the option -fileid is specified, filelist is a list of open file identifiers rather than a list of file names. This option is not available on all Unix systems. Use the infox have_fchown command to determine if this functionality is available.

The chgrp command is not available on Windows.

dup fileId ?targetFileId?

Duplicate an open file. A new file id is opened that addresses the same file as fileId.

If targetFileId is specified, the the file is dup to this specified file id. Normally this is stdin, stdout, or stderr. The dup command will handle flushing output and closing this file. The new file will be buffered, if its needs to be unbuffered, use the fcntl command to set it unbuffered.

If fileId is a number rather than a Tcl file id, then the dup command will bind that file to a Tcl file id. This is useful for accessing files that are passed from the parent process. The argument ?targetFileId? is not valid with this operation.

On Windows, only stdin, stdout, or stderr or a non-socket file handle number maybe specified for targetFileId. The dup command does not work on sockets on Windows.

fcntl fileId attribute ?value?

This command either sets or clears a file option or returns its current value. If value is not specified, then the current value of attribute is returned. All values are boolean. Some attributes maybe only be gotten, not modified. The following attributes may be specified:

RDONLY

The file is opened for reading only. (Get only)

WRONLY

The file is opened for writing only. (Get only)

RDWR

The file is opened for reading and writing. (Get only)

READ

If the file is readable. (Get only).

WRITE

If the file is writable. (Get only).

APPEND

The file is opened for append-only writes. All writes will be forced to the end of the file. (Get or set).

NONBLOCK

The file is to be accessed with non-blocking I/O. See the read system call for a description of how it affects the behavior of file reads.

CLOEXEC

Close the file on an process exec. If the execl command or some other mechanism causes the process to do an exec, the file will be closed if this option is set.

NOBUF

The file is not buffered. If set, then there no buffering for the file.

LINEBUF

Output the file will be line buffered. The buffer will be flushed when a newline is written, when the buffer is full, or when input is requested.

KEEPALIVE

Keep a socket connection alive. If SIGPIPE is enabled, then it is sent if connection is broken and data is written to the socket. If SIGPIPE is ignored, an error is returned on the write. This attribute is valid only on sockets. By default, SIGPIPE is ignored in Tcl.

The NONBLOCK, NOBUF and LINEBUF are provided for compatibility with older scripts. Thefconfigure command is preferred method of getting and setting these attributes.

The APPEND and CLOEXEC options are not available on Windows.

flock options fileId ?start? ?length? ?origin?

This command places a lock on all or part of the file specified by fileId. The lock is either advisory or mandatory, depending on the mode bits of the file. The lock is placed beginning at relative byte offset start for length bytes. If start or length is omitted or empty, zero is assumed. If length is zero, then the lock always extents to end of file, even if the file grows. If origin is "start", then the offset is relative to the beginning of the file. If it is "current", it is relative to the current access position in the file. If it is "end", then it is relative to the end-of-file (a negative is before the EOF, positive is after). If origin is omitted, start is assumed.

The following options are recognized:

See your system's fcntl system call documentation for full details of the behavior of file locking. If locking is being done on ranges of a file, it is best to use unbuffered file access (see the fcntl command).

The flock command is not available on Windows 95. It is available on Windows NT.

for_file var filename code

This procedure implements a loop over the contents of a file. For each line in filename, it sets var to the line and executes code.

The break and continue commands work as with foreach.

For example, the command

	for_file line /etc/passwd {echo $line}
    

would echo all the lines in the password file.

funlock fileId ?start? ?length? ?origin?

Remove a locked from a file that was previously placed with the flock command. The arguments are the same as for the flock command, see that command for more details.

The funlock command is not available on Windows 95. It is available on Windows NT.

fstat fileId ?item? | ?stat arrayvar?

Obtain status information about an open file.

The following keys are used to identify data items:

If one of these keys is specified as item, then that data item is returned.

If stat arrayvar is specified, then the information is returned in the array arrayvar. Each of the above keys indexes an element of the array containing the data.

If only fileId is specified, the command returns the data as a keyed list.

The following values may be returned only if explicitly asked for, it will not be returned with the array or keyed list forms:

ftruncate [-fileid] file newsize

Truncate a file to have a length of at most newsize bytes.

If the option -fileid is specified, file is an open file identifier, otherwise it is a file path.

This command is not available or not fully functional if the underlying operating system support is not available. The command infox have_truncate will indicate if this command may truncate by file path. The command infox have_ftruncate will indicate if this command may truncate by file id.

The -fileid option is not available on Windows.

lgets fileId ?varName?

Reads the next Tcl list from the file given by fileId and discards the terminating newline character. This command differs from the gets command, in that it reads Tcl lists rather than lines. If the list contains newlines or binary data, then that newline or bytes of zero will be returned as part of the result. Only a newline not quoted as part of the list indicates the end of the list. There is no corresponding command for outputting lists, as puts will do this correctly.

If varName is specified, then the line is placed in the variable by that name and the return value is a count of the number of characters read (not including the newline). If the end of the file is reached before reading any characters then -1 is returned and varName is set to an empty string. If varName is specified and an error occurs, what ever data was read will be returned in the variable, however the resulting string may not be a valid list.

If varName is not specified then the return value will be the line (minus the newline character) or an empty string if the end of the file is reached before reading any characters. An empty string will also be returned if a line contains no characters except the newline, so eof may have to be used to determine what really happened.

The lgets command maybe used to read and write lists containing binary data, however translation must be set to lf or the data maybe corrupted.

If lgets is currently supported on non-blocking files.

pipe ?fileId_var_r fileId_var_w?

Create a pipe. If fileId_var_r and fileId_var_r are specified, then pipe will set the a variable named fileId_var_r to contain the fileId of the side of the pipe that was opened for reading, and fileId_var_w will contain the fileId of the side of the pipe that was opened for writing.

If the fileId variables are not specified, then a list containing the read and write fileIdw is returned as the result of the command.

read_file ?-nonewline? fileName

read_file fileName numBytes

This procedure reads the file fileName and returns the contents as a string. If -nonewline is specified, then the last character of the file is discarded if it is a newline. The second form specifies exactly how many bytes will be read and returned, unless there are fewer than numBytes bytes left in the file; in this case, all the remaining bytes are returned.

select readfileIds ?writefileIds? ?exceptfileIds? ?timeout?

This command allows an Extended Tcl program to wait on zero or more files being ready for for reading, writing, have an exceptional condition pending, or for a timeout period to expire. readFileIds, writeFileIds, exceptFileIds are each lists of fileIds, as returned from open, to query. An empty list ({}) may be specified if a category is not used.

The files specified by the readFileIds list are checked to see if data is available for reading. The writeFileIds are checked if the specified files are clear for writing. The exceptFileIds are checked to see if an exceptional condition has occurred (typically, an error). The write and exception checking is most useful on devices, however, the read checking is very useful when communicating with multiple processes through pipes. Select considers data pending in the stdio input buffer for read files as being ready for reading, the files do. not have to be unbuffered.

Timeout is a floating point timeout value, in seconds. If an empty list is supplied (or the parameter is omitted), then no timeout is set. If the value is zero, then the select command functions as a poll of the files, returning immediately even if none are ready.

If the timeout period expires with none of the files becoming ready, then the command returns an empty list. Otherwise the command returns a list of three elements, each of those elements is a list of the fileIds that are ready in the read, write and exception classes. If none are ready in a class, then that element will be the null list. For example:

        select {file3 file4 file5} {file6 file7} {} 10.5
could return
        {file3 file4} {file6} {}
or perhaps
        file3 {} {}
    

On Windows, only sockets can be used with the select command. Pipes, as returned by the open command, are not supported.

write_file fileName string ?string...?

This procedure writes the specified strings to the named file.

NETWORK PROGRAMMING SUPPORT

host_info option host

Obtain information about an Internet host. The argument host can be either a host name or an IP address.

The following subcommands are recognized:

FILE SCANNING COMMANDS

File scanning in Tcl centers around the concept of a scan context. A scan context contains one or more match statements, which associate regular expressions to scan for with Tcl code to be executed when the expressions are matched.

scancontext ?option?

This command manages file scan contexts. A scan context is a collection of regular expressions and commands to execute when that regular expression matches a line of the file. A context may also have a single default match, to be applied against lines that do not match any of the regular expressions. Multiple scan contexts may be defined and they may be reused on multiple files. A scan context is identified by a context handle. The scancontext command takes the following forms:

scancontext create

Create a new scan context. The scanmatch command is used to define patterns in the context. A contexthandle is returned, which the Tcl programmer uses to refer to the newly created scan context in calls to the Tcl file scanning commands.

scancontext delete contexthandle

Delete the scan context identified by contexthandle, and free all of the match statements and compiled regular expressions associated with the specified context.

scancontext copyfile contexthandle ?filehandle?

Set or return the file handle that unmatched lines are copied to. (See scanfile). If filehandle is omitted, the copy file handle is returned. If no copy file is associated with the context, {} is returned. If a file handle is specified, it becomes the copy file for this context. If filehandle is {}, then it removes any copy file specification for the context.

scanfile ?-copyfile copyFileId? contexthandle fileId

Scan the file specified by fileId, starting from the current file position. Check all patterns in the scan context specified by contexthandle against it, executing the match commands corresponding to patterns matched.

If the optional -copyfile argument is specified, the next argument is a file ID to which all lines not matched by any pattern (excluding the default pattern) are to be written. If the copy file is specified with this flag, instead of using the scancontext copyfile command, the file is disassociated from the scan context at the end of the scan.

This command does not work on files containing binary data (bytes of zero).

scanmatch ?-nocase? contexthandle ?regexp? commands

Specify Tcl commands, to be evaluated when regexp is matched by a scanfile command. The match is added to the scan context specified by contexthandle. Any number of match statements may be specified for a give context. Regexp is a regular expression (see the regexp command). If -nocase is specified as the first argument, the pattern is matched regardless of alphabetic case.

If regexp is not specified, then a default match is specified for the scan context. The default match will be executed when a line of the file does not match any of the regular expressions in the current scancontext.

The array matchInfo is available to the Tcl code that is executed when an expression matches (or defaults). It contains information about the file being scanned and where within it the expression was matched.

matchInfo is local to the top level of the match command unless declared global at that level by the Tcl global command. If it is to be used as a global, it must be declared global before scanfile is called (since scanfile sets the matchInfo before the match code is executed, a subsequent global will override the local variable). The following array entries are available:

All scanmatch patterns that match a line will be processed in the order in which their specifications were added to the scan context. The remainder of the scanmatch pattern-command pairs may be skipped for a file line if a continue is executed by the Tcl code of a preceding, matched pattern.

If a return is executed in the body of the match command, the scanfile command currently in progress returns, with the value passed to return as its return value.

MATH COMMANDS

The following procedures provide command interfaces to the expr math functions. They take the same arguments as the expr functions and may take expressions as arguments.

abs	acos	asin	atan2	
atan	ceil	cos	cosh	
double	exp	floor	fmod	
hypot	int	log10	log	
pow	round	sin	sinh	
sqrt	tan	tanh	

max num1 ?..numN?

expr max(num1, num2)

Returns the argument that has the highest numeric value. Each argument may be any integer or floating point value.

This functionality is also available as a math function max in the Tcl expr command.

min num1 ?..numN?

expr min(num1, num2)

Returns the argument that has the lowest numeric value. Each argument may be any integer or floating point value.

This functionality is also available as a math function min in the Tcl expr command.

random limit | seed ?seedval?

Generate a pseudorandom integer number greater than or equal to zero and less than limit. If seed is specified, then the command resets the random number generator to a starting point derived from the seedval. This allows one to reproduce pseudorandom number sequences for testing purposes. If seedval is omitted, then the seed is set to a value based on current system state and the current time, providing a reasonably interesting and ever-changing seed.

LIST MANIPULATION COMMANDS

intersect lista listb

Procedure to return the logical intersection of two lists. The returned list will be sorted.

intersect3 lista listb

Procedure to intersects two lists, returning a list containing three lists: The first list returned is everything in lista that wasn't in listb. The second list contains the intersection of the two lists, and the third list contains all the elements that were in listb but weren't in lista. The returned lists will be sorted.

lassign list var ?var...?

Assign successive elements of a list to specified variables. If there are more variable names than fields, the remaining variables are set to the empty string. If there are more elements than variables, a list of the unassigned elements is returned.

For example,

    lassign {dave 100 200 {Dave Foo}} name uid gid longName
    

Assigns name to ``dave'', uid to ``100'', gid to ``200'', and longName to ``Dave Foo''.

lcontain list element

Determine if the element is a list element of list. If the element is contained in the list, 1 is returned, otherwise, 0 is returned.

lempty list

Determine if the specified list is empty. If empty, 1 is returned, otherwise, 0 is returned. This command is an alternative to comparing a list to an empty string, however it checks for a string of all whitespaces, which is an empty list.

lmatch ?mode? list pattern

Search the elements of list, returning a list of all elements matching pattern. If none match, an empty list is returned.

The mode argument indicates how the elements of the list are to be matched against pattern and it must have one of the following values:

If mode is omitted then it defaults to -glob.

Only the -exact comparison will work on binary data.

lrmdups list

Procedure to remove duplicate elements from a list. The returned list will be sorted.

lvarcat var string ?string...?

This command treats each string argument as a list and concatenates them to the end of the contents of var, forming a a single list. The list is stored back into var and also returned as the result. if var does not exist, it is created.

lvarpop var ?indexExpr? ?string?

The lvarpop command pops (deletes) the element indexed by the expression indexExpr from the list contained in the variable var. If index is omitted, then 0 is assumed. If string, is specified, then the deleted element is replaced by string. The replaced or deleted element is returned. Thus ``lvarpop argv 0'' returns the first element of argv, setting argv to contain the remainder of the string.

If the expression indexExpr starts with the string end, then end is replaced with the index of the last element in the list. If the expression starts with len, then len is replaced with the length of the list.

lvarpush var string ?indexExpr?

The lvarpush command pushes (inserts) string as an element in the list contained in the variable var. The element is inserted before position indexExpr in the list. If index is omitted, then 0 is assumed. If var does not exists, it is created.

If the expression indexExpr starts with the string end, then end is replaced with the index of the last element in the list. If the expression starts with len, then len is replaced with the length of the list. Note the a value of end means insert the string before the last element.

union lista listb

Procedure to return the logical union of the two specified lists. Any duplicate elements are removed.

KEYED LISTS

A keyed list is a list in which each element contains a key and value pair. These element pairs are stored as lists themselves, where the key is the first element of the list, and the value is the second. The key-value pairs are referred to as fields. This is an example of a keyed list:

{{NAME {Frank Zappa}} {JOB {musician and composer}}}

If the variable person contained the above list, then keylget person NAME would return {Frank Zappa}. Executing the command:

	keylset person ID 106

would make person contain

{{ID 106} {NAME {Frank Zappa}} {JOB {musician and composer}}

Fields may contain subfields; `.' is the separator character. Subfields are actually fields where the value is another keyed list. Thus the following list has the top level fields ID and NAME, and subfields NAME.FIRST and NAME.LAST:

{ID 106} {NAME {{FIRST Frank} {LAST Zappa}}}

There is no limit to the recursive depth of subfields, allowing one to build complex data structures.

Keyed lists are constructed and accessed via a number of commands. All keyed list management commands take the name of the variable containing the keyed list as an argument (i.e. passed by reference), rather than passing the list directly.

keyldel listvar key

Delete the field specified by key from the keyed list in the variable listvar. This removes both the key and the value from the keyed list.

keylget listvar ?key? ?retvar | {}?

Return the value associated with key from the keyed list in the variable listvar. If retvar is not specified, then the value will be returned as the result of the command. In this case, if key is not found in the list, an error will result.

If retvar is specified and key is in the list, then the value is returned in the variable retvar and the command returns 1 if the key was present within the list. If key isn't in the list, the command will return 0, and retvar will be left unchanged. If {} is specified for retvar, the value is not returned, allowing the Tcl programmer to determine if a key is present in a keyed list without setting a variable as a side-effect.

If key is omitted, then a list of all the keys in the keyed list is returned.

keylkeys listvar ?key?

Return the a list of the keys in the keyed list in the variable listvar. If keys is specified, then it is the name of a key field who's subfield keys are to be retrieve.

keylset listvar key value ?key2 value2 ...?

Set the value associated with key, in the keyed list contained in the variable listvar, to value. If listvar does not exists, it is created. If key is not currently in the list, it will be added. If it already exists, value replaces the existing value. Multiple keywords and values may be specified, if desired.

STRING AND CHARACTER MANIPULATION COMMANDS

ccollate ?-local? string1 string2

This command compares two strings. If returns -1 if string1 is less than string2, 0 if they are equal and 1 if string1 is greater than string2.

If -local is specified, the strings are compared according to the collation environment of the current locale.

This command does not work with binary or UTF data.

cconcat ?string1? ?string2? ?...?

Concatenate the arguments, returning the resulting string. While string concatenation is normally performed by the parser, it is occasionally useful to have a command that returns a string. The is generally useful when a command to evaluate is required. No separators are inserted between the strings.

This command is UTF-aware.

cequal string string

This command compares two strings for equality. It returns 1 if string1 and string2 are the identical and 0 if they are not. This command is a short-cut for string compare and avoids the problems with string expressions being treated unintentionally as numbers.

This command is UTF-aware and will also work on binary data.

cindex string indexExpr

Returns the character indexed by the expression indexExpr (zero based) from string.

If the expression indexExpr starts with the string end, then end is replaced with the index of the last character in the string. If the expression starts with len, then len is replaced with the length of the string.

This command is UTF-aware.

clength string

Returns the length of string in characters. This command is a shortcut for:

    string length string
    

This command is UTF-aware.

crange string firstExpr lastExpr

Returns a range of characters from string starting at the character indexed by the expression firstExpr (zero-based) until the character indexed by the expression lastExpr.

If the expression firstExpr or lastExpr starts with the string end, then end is replaced with the index of the last character in the string. If the expression starts with len, then len is replaced with the length of the string.

This command is UTF-aware.

csubstr string firstExpr lengthExpr

Returns a range of characters from string starting at the character indexed by the expression firstExpr (zero-based) for lengthExpr characters.

If the expression firstExpr or lengthExpr starts with the string end, then end is replaced with the index of the last character in the string. If the expression starts with len, then len is replaced with the length of the string.

This command is UTF-aware.

ctoken strvar separators

Parse a token out of a character string. The string to parse is contained in the variable named strvar. The string separators contains all of the valid separator characters for tokens in the string. All leading separators are skipped and the first token is returned. The variable strvar will be modified to contain the remainder of the string following the token.

This command does not work with binary data.

ctype ?-failindex var? class string

ctype determines whether all characters in string are of the specified class. It returns 1 if they are all of class, and 0 if they are not, or if the string is empty. This command also provides another method (besides format and scan) of converting between an ASCII character and its numeric value. The following ctype commands are available:

If -failindex is specified, then the index into string of the first character that did not match the class is returned in var.

replicate string countExpr

Returns string, replicated the number of times indicated by the expression countExpr.

This command is UTF-aware and will work with binary data.

translit inrange outrange string

Translate characters in string, changing characters occurring in inrange to the corresponding character in outrange. Inrange and outrange may be list of characters or a range in the form `A-M'. For example:

        translit a-z A-Z foobar

This command currently only supports characters in ASCII range; UTF-8 characters
out of this range will generate an error.
    

XPG/3 MESSAGE CATALOG COMMANDS

Tcl programmers can use message catalogs to create applications that are language-independent. Through the use of message catalogs, prompts, messages, menus and so forth can exist for any number of languages, and they can altered, and new languages added, without affecting any Tcl or C source code, greatly easing the maintenance difficulties incurred by supporting multiple languages.

A default text message is passed to the command that fetches entries from message catalogs. This allows the Tcl programmer to create message catalogs containing messages in various languages, but still have a set of default messages available regardless of the presence of any message catalogs, and allow the programs to press on without difficulty when no catalogs are present.

Thus, the normal approach to using message catalogs is to ignore errors on catopen, in which case catgets will return the default message that was specified in the call.

The Tcl message catalog commands normally ignore most errors. If it is desirable to detect errors, a special option is provided. This is normally used only during debugging, to insure that message catalogs are being used. If your Unix implementation does not have XPG/3 message catalog support, stubs will be compiled in that will create a version of catgets that always returns the default string. This allows for easy porting of software to environments that don't have support for message catalogs.

Message catalogs are global to the process, an application with multiple Tcl interpreters within the same process may pass and share message catalog handles.

catopen ?-fail | -nofail? catname

Open the message catalog catname. This may be a relative path name, in which case the NLSPATH environment variable is searched to find an absolute path to the message catalog. A handle in the form msgcat N is returned. Normally, errors are ignored, and in the case of a failed call to catopen, a handle is returned to an unopened message catalog. (This handle may still be passed to catgets and catclose, causing catgets to simply return the default string, as described above. If the -fail option is specified, an error is returned if the open fails. The option -nofail specifies the default behavior of not returning an error when catopen fails to open a specified message catalog. If the handle from a failed catopen is passed to catgets, the default string is returned.

catgets catHandle setnum msgnum defaultstr

Retrieve a message form a message catalog. CatHandle should be a Tcl message catalog handle that was returned by catopen. Setnum is the message set number, and msgnum is the message number. If the message catalog was not opened, or the message set or message number cannot be found, then the default string, defaultstr, is returned.

catclose ?-fail | -nofail? cathandle

Close the message catalog specified by cathandle. Normally, errors are ignored. If -fail is specified, any errors closing the message catalog file are returned. The option -nofail specifies the default behavior of not returning an error. The use of -fail only makes sense if it was also specified in the call to catopen.

mainloop

This procedure sets up a top-level event loop. Events are processed until there are no more active event sources, at which time the process exits. It is used to build event oriented programs using the TclX shell in a style similar to that used with wish. If the global variable tcl_interactive exists and has a true value an interactive command handler is started as well. If the command handler is terminated by an EOF, the process will be exited.

HELP FACILITY

The help facility can be accessed in two ways, as interactive commands in the Extended Tcl shell or as an interactive Tk-based program (if you have built Extended Tcl with Tk).

To run the Tk-based interactive help program:

    tclhelp ?addpaths?

Where addpaths are additional paths to search for help directories. By default, only the auto_path used by tclhelp is search. This will result in help on Tcl, Extended Tcl and Tk.

The following interactive Tcl commands and options are provided with the help package:

help

Help, without arguments, lists of all the help subjects and pages under the current help subject.

help subject

Displays all of help pages and lower level subjects (if any exist) under the subject subject.

help subject/helppage

Display the specified help page. The help output is passed through a simple pager if output exceeds 23 lines, pausing waiting for a return to be entered. If any other character is entered, the output is terminated.

helpcd ?subject?

Change the current subject, which is much like the Unix current directory. If subject is not specified, return to the top-level of the help tree. Help subject path names may also include ``..'' elements.

helppwd

Displays the current help subject.

help help | ?

Displays help on the help facility at any directory level.

apropos pattern

This command locates subjects by searching their one-line descriptions for a pattern. Apropos is useful when you can remember part of the name or description of a command, and want to search through the one-line summaries for matching lines. Full regular expressions may be specified (see the regexp command).

TCL LOADABLE LIBRARIES AND PACKAGES

The package library file is just a regular Unix text file, editable with your favorite text editor, containing packages of Tcl source code. The package library file name must have the suffix .tlib. An index file with the same prefix name and the suffix .tndx resides the same directory as the .tlib file. The .tndx will be automatically created whenever it is out of date or missing (provided there is write access to the directory).

The variable auto_path contains a list of directories that are searched for libraries. The first time an unknown command trap is take, the indexes for the libraries are loaded into memory. If the auto_path variable is changed during execution of a program, it will be re-searched. Only the first package of a given name found during the execution of a program is loaded. This can be overridden with loadlibindex command.

The start of a package is delimited by:

#@package: package_name proc1 ?..procN?

These lines must start in column one. Everything between the #@package: keyword and the next #@package: keyword or a #@packend keyword, or the end of the file, becomes part of the named package. The specified procedures, proc1..procN, are the entry points of the package. When a command named in a package specification is executed and detected as an unknown command, all code in the specified package will be sourced. This package should define all of the procedures named on the package line, define any support procedures required by the package and do any package-specific initialization. Packages declarations maybe continued on subsequent lines using standard Tcl backslash line continuations. The #@packend keyword is useful to make sure only the minimum required section of code is sourced. Thus for example a large comment block at the beginning of the next file won't be loaded.

Care should be taken in defining package_name, as the first package found in the path by with a given name is loaded. This can be useful in developing new version of packages installed on the system.

For example, in a package source file, the presence of the following line:

#@package: directory_stack pushd popd dirs

says that the text lines following that line in the package file up to the next package line or the end of the file is a package named directory_stack and that an attempt to execute either pushd, popd or dirs when the routine is not already defined will cause the directory_stack portion of the package file to be loaded.

PACKAGE LIBRARY MANAGEMENT COMMANDS

auto_commands ?-loaders?

Lists the names of all known loadable procedures and commands procedures. If -loaders is specified, the command that will be executed to load the command will also be returned.

buildpackageindex libfilelist

Build index files for package libraries. The argument libfilelist is a list of package libraries. Each name must end with the suffix .tlib. A corresponding .tndx file will be built. The user must have write access to the directory containing each library.

convert_lib tclIndex packagelib ?ignore?

Convert a Ousterhout style tclIndex index file and associate source files into a package library packagelib. If packagelib does not have a .tlib extension, one will be added. Any files specified in tclIndex that are in the list ignore will be skipped. Files listed in ignore should just be the base file names, not full paths.

loadlibindex libfile.tlib

Load the package library index of the library file libfile (which must have the suffix .tlib). Package library indexes along the auto_path are loaded automatically on the first demand_load; this command is provided to explicitly load libraries that are not in the path. If the index file (with a .tndx suffix) does not exists or is out of date, it will be rebuilt if the user has directory permissions to create it. If a package with the same name as a package in libfile.tlib has already been loaded, its definition will be overridden by the new package. However, if any procedure has actually been used from the previously defined package, the procedures from libfile.tlib will not be loaded.

auto_packages ?-location?

Returns a list of the names of all defined packages. If -location is specified, a list of pairs of package name and the .tlib path name, offset and length of the package within the library.

auto_load_file file

Source a file, as with the source command, except search auto_path for the file.

searchpath path file

Search all directories in the specified path, which is a Tcl list, for the specified file. Returns the full path name of the file, or an empty string if the requested file could not be found.