tcllib 1.16-dfsg-2 / usr / share / tcltk / tcllib1.16 / fileutil / fileutil.tcl

# fileutil.tcl --
#
#	Tcl implementations of standard UNIX utilities.
#
# Copyright (c) 1998-2000 by Ajuba Solutions.
# Copyright (c) 2002      by Phil Ehrens <phil@slug.org> (fileType)
# Copyright (c) 2005-2013 by Andreas Kupries <andreas_kupries@users.sourceforge.net>
#
# See the file "license.terms" for information on usage and redistribution
# of this file, and for a DISCLAIMER OF ALL WARRANTIES.
# 
# RCS: @(#) $Id: fileutil.tcl,v 1.78 2010/06/17 04:46:19 andreas_kupries Exp $

package require Tcl 8.2
package require cmdline
package provide fileutil 1.14.6

namespace eval ::fileutil {
    namespace export \
	    grep find findByPattern cat touch foreachLine \
	    jail stripPwd stripN stripPath tempdir tempfile \
	    install fileType writeFile appendToFile \
	    insertIntoFile removeFromFile replaceInFile \
	    updateInPlace test tempdirReset
}

# ::fileutil::grep --
#
#	Implementation of grep.  Adapted from the Tcler's Wiki.
#
# Arguments:
#	pattern		pattern to search for.
#	files		list of files to search; if NULL, uses stdin.
#
# Results:
#	results		list of matches

proc ::fileutil::grep {pattern {files {}}} {
    set result [list]
    if {[llength $files] == 0} {
	# read from stdin
	set lnum 0
	while {[gets stdin line] >= 0} {
	    incr lnum
	    if {[regexp -- $pattern $line]} {
		lappend result "${lnum}:${line}"
	    }
	}
    } else {
	foreach filename $files {
	    set file [open $filename r]
	    set lnum 0
	    while {[gets $file line] >= 0} {
		incr lnum
		if {[regexp -- $pattern $line]} {
		    lappend result "${filename}:${lnum}:${line}"
		}
	    }
	    close $file
	}
    }
    return $result
}

# ::fileutil::find ==

# Below is the core command, which is portable across Tcl versions and
# platforms. Functionality which is common or platform and/or Tcl
# version dependent, has been factored out/ encapsulated into separate
# (small) commands. Only these commands may have multiple variant
# implementations per the available features of the Tcl core /
# platform.
#
# These commands are
#
# FADD   - Add path result, performs filtering. Portable!
# GLOBF  - Return files in a directory.         Tcl version/platform dependent.
# GLOBD  - Return dirs  in a directory.         Tcl version/platform dependent.
# ACCESS - Check directory for accessibility.   Tcl version/platform dependent.

proc ::fileutil::find {{basedir .} {filtercmd {}}} {
    set result {}
    set filt   [string length $filtercmd]

    if {[file isfile $basedir]} {
	# The base is a file, and therefore only possible result,
	# modulo filtering.

	FADD $basedir

    } elseif {[file isdirectory $basedir]} {

	# For a directory as base we do an iterative recursion through
	# the directory hierarchy starting at the base. We use a queue
	# (Tcl list) of directories we have to check. We access it by
	# index, and stop when we have reached beyond the end of the
	# list. This is faster than removing elements from the be-
	# ginning of the list, as that entails copying down a possibly
	# large list of directories, making it O(n*n). The index is
	# faster, O(n), at the expense of memory. Nothing is deleted
	# from the list until we have processed all directories in the
	# hierarchy.
	#
	# We scan each directory at least twice. First for files, then
	# for directories. The scans may internally make several
	# passes (normal vs hidden files).
	#
	# Looped directory structures due to symbolic links are
	# handled by _fully_ normalizing directory paths and checking
	# if we encountered the normalized form before. The array
	# 'known' is our cache where we record the known normalized
	# paths.

	set pending [list $basedir]
	set at      0
	array set   known {}

	while {$at < [llength $pending]} {
	    # Get next directory not yet processed.
	    set current [lindex $pending $at]
	    incr at

	    # Is the directory accessible? Continue if not.
	    ACCESS $current

	    # Files first, then the sub-directories ...

	    foreach f [GLOBF $current] { FADD $f }

	    foreach f [GLOBD $current] {
		# Ignore current and parent directory, this needs
		# explicit filtering outside of the filter command.
		if {
		    [string equal [file tail $f]  "."] ||
		    [string equal [file tail $f] ".."]
		} continue

		# Extend result, modulo filtering.
		FADD $f

		# Detection of symlink loops via a portable path
		# normalization computing a canonical form of the path
		# followed by a check if that canonical form was
		# encountered before. If ok, record directory for
		# expansion in future iterations.

		set norm [fileutil::fullnormalize $f]
		if {[info exists known($norm)]} continue
		set known($norm) .

		lappend pending $f
	    }
	}
    } else {
	return -code error "$basedir does not exist"
    }

    return $result
}

# Helper command for fileutil::find. Performs the filtering of the
# result per a filter command for the candidates found by the
# traversal core, see above. This is portable.

proc ::fileutil::FADD {filename} {
    upvar 1 result result filt filt filtercmd filtercmd
    if {!$filt} {
	lappend result $filename
	return
    }

    set here [pwd]
    cd [file dirname $filename]

    if {[uplevel 2 [linsert $filtercmd end [file tail $filename]]]} {
	lappend result $filename
    }

    cd $here
    return
}

# The next three helper commands for fileutil::find depend strongly on
# the version of Tcl, and partially on the platform.

# 1. The -directory and -types switches were added to glob in Tcl
#    8.3. This means that we have to emulate them for Tcl 8.2.
#
# 2. In Tcl 8.3 using -types f will return only true files, but not
#    links to files. This changed in 8.4+ where links to files are
#    returned as well. So for 8.3 we have to handle the links
#    separately (-types l) and also filter on our own.
#    Note that Windows file links are hard links which are reported by
#    -types f, but not -types l, so we can optimize that for the two
#    platforms.
#
#    Note further that we have to handle broken links on our own. They
#    are not returned by glob yet we want them in the output.
#
# 3. In Tcl 8.3 we also have a crashing bug in glob (SIGABRT, "stat on
#    a known file") when trying to perform 'glob -types {hidden f}' on
#    a directory without e'x'ecute permissions. We code around by
#    testing if we can cd into the directory (stat might return enough
#    information too (mode), but possibly also not portable).
#
#    For Tcl 8.2 and 8.4+ glob simply delivers an empty result
#    (-nocomplain), without crashing. For them this command is defined
#    so that the bytecode compiler removes it from the bytecode.
#
#    This bug made the ACCESS helper necessary.
#    We code around the problem by testing if we can cd into the
#    directory (stat might return enough information too (mode), but
#    possibly also not portable).

if {[package vsatisfies [package present Tcl] 8.5]} {
    # Tcl 8.5+.
    # We have to check readability of "current" on our own, glob
    # changed to error out instead of returning nothing.

    proc ::fileutil::ACCESS {args} {}

    proc ::fileutil::GLOBF {current} {
	if {![file readable $current]} {
	    return {}
	}
	if {([file type $current] eq "link") &&
	    !([file exists   [file readlink $current]] &&
	      [file readable [file readlink $current]])} {
	    return {}
	}

	set res [lsort -unique [concat \
		     [glob -nocomplain -directory $current -types f          -- *] \
		     [glob -nocomplain -directory $current -types {hidden f} -- *]]]

	# Look for broken links (They are reported as neither file nor directory).
	foreach l [lsort -unique [concat \
		       [glob -nocomplain -directory $current -types l          -- *] \
		       [glob -nocomplain -directory $current -types {hidden l} -- *]]] {
	    if {[file isfile      $l]} continue
	    if {[file isdirectory $l]} continue
	    lappend res $l
	}
	return [lsort -unique $res]
    }

    proc ::fileutil::GLOBD {current} {
	if {![file readable $current]} {
	    return {}
	}
	if {([file type $current] eq "link") &&
	    !([file exists   [file readlink $current]] &&
	      [file readable [file readlink $current]])} {
	    return {}
	}

	lsort -unique [concat \
	    [glob -nocomplain -directory $current -types d          -- *] \
	    [glob -nocomplain -directory $current -types {hidden d} -- *]]
    }

} elseif {[package vsatisfies [package present Tcl] 8.4]} {
    # Tcl 8.4+.
    # (Ad 1) We have -directory, and -types,
    # (Ad 2) Links are returned for -types f/d if they refer to files/dirs.
    # (Ad 3) No bug to code around

    proc ::fileutil::ACCESS {args} {}

    proc ::fileutil::GLOBF {current} {
	set res [lsort -unique [concat \
		    [glob -nocomplain -directory $current -types f          -- *] \
		    [glob -nocomplain -directory $current -types {hidden f} -- *]]]

	# Look for broken links (They are reported as neither file nor directory).
	foreach l [lsort -unique [concat \
		       [glob -nocomplain -directory $current -types l          -- *] \
		       [glob -nocomplain -directory $current -types {hidden l} -- *]]] {
	    if {[file isfile      $l]} continue
	    if {[file isdirectory $l]} continue
	    lappend res $l
	}
	return [lsort -unique $res]
    }

    proc ::fileutil::GLOBD {current} {
	lsort -unique [concat \
	    [glob -nocomplain -directory $current -types d          -- *] \
	    [glob -nocomplain -directory $current -types {hidden d} -- *]]
    }

} elseif {[package vsatisfies [package present Tcl] 8.3]} {
    # 8.3.
    # (Ad 1) We have -directory, and -types,
    # (Ad 2) Links are NOT returned for -types f/d, collect separately.
    #        No symbolic file links on Windows.
    # (Ad 3) Bug to code around.

    proc ::fileutil::ACCESS {current} {
	if {[catch {
	    set h [pwd] ; cd $current ; cd $h
	}]} {return -code continue}
	return
    }

    if {[string equal $::tcl_platform(platform) windows]} {
	proc ::fileutil::GLOBF {current} {
	    concat \
		[glob -nocomplain -directory $current -types f          -- *] \
		[glob -nocomplain -directory $current -types {hidden f} -- *]]
	}
    } else {
	proc ::fileutil::GLOBF {current} {
	    set l [concat \
		       [glob -nocomplain -directory $current -types f          -- *] \
		       [glob -nocomplain -directory $current -types {hidden f} -- *]]

	    foreach x [concat \
			   [glob -nocomplain -directory $current -types l          -- *] \
			   [glob -nocomplain -directory $current -types {hidden l} -- *]] {
		if {[file isdirectory $x]} continue
		# We have now accepted files, links to files, and broken links.
		lappend l $x
	    }

	    return $l
	}
    }

    proc ::fileutil::GLOBD {current} {
	set l [concat \
		   [glob -nocomplain -directory $current -types d          -- *] \
		   [glob -nocomplain -directory $current -types {hidden d} -- *]]

	foreach x [concat \
		       [glob -nocomplain -directory $current -types l          -- *] \
		       [glob -nocomplain -directory $current -types {hidden l} -- *]] {
	    if {![file isdirectory $x]} continue
	    lappend l $x
	}

	return $l
    }
} else {
    # 8.2.
    # (Ad 1,2,3) We do not have -directory, nor -types. Full emulation required.

    proc ::fileutil::ACCESS {args} {}

    if {[string equal $::tcl_platform(platform) windows]} {
	# Hidden files cannot be handled by Tcl 8.2 in glob. We have
	# to punt.

	proc ::fileutil::GLOBF {current} {
	    set current \\[join [split $current {}] \\]
	    set res {}
	    foreach x [glob -nocomplain -- [file join $current *]] {
		if {[file isdirectory $x]} continue
		if {[catch {file type $x}]} continue
		# We have now accepted files, links to files, and
		# broken links. We may also have accepted a directory
		# as well, if the current path was inaccessible. This
		# however will cause 'file type' to throw an error,
		# hence the second check.
		lappend res $x
	    }
	    return $res
	}

	proc ::fileutil::GLOBD {current} {
	    set current \\[join [split $current {}] \\]
	    set res {}
	    foreach x [glob -nocomplain -- [file join $current *]] {
		if {![file isdirectory $x]} continue
		lappend res $x
	    }
	    return $res
	}
    } else {
	# Hidden files on Unix are dot-files. We emulate the switch
	# '-types hidden' by using an explicit pattern.

	proc ::fileutil::GLOBF {current} {
	    set current \\[join [split $current {}] \\]
	    set res {}
	    foreach x [glob -nocomplain -- [file join $current *] [file join $current .*]] {
		if {[file isdirectory $x]} continue
		if {[catch {file type $x}]} continue
		# We have now accepted files, links to files, and
		# broken links. We may also have accepted a directory
		# as well, if the current path was inaccessible. This
		# however will cause 'file type' to throw an error,
		# hence the second check.

		lappend res $x
	    }
	    return $res
	}

	proc ::fileutil::GLOBD {current} {
	    set current \\[join [split $current {}] \\]
	    set res {}
	    foreach x [glob -nocomplain -- $current/* [file join $current .*]] {
		if {![file isdirectory $x]} continue
		lappend res $x
	    }
	    return $res
	}
    }
}

# ::fileutil::findByPattern --
#
#	Specialization of find. Finds files based on their names,
#	which have to match the specified patterns. Options are used
#	to specify which type of patterns (regexp-, glob-style) is
#	used.
#
# Arguments:
#	basedir		Directory to start searching from.
#	args		Options (-glob, -regexp, --) followed by a
#			list of patterns to search for.
#
# Results:
#	files		a list of interesting files.

proc ::fileutil::findByPattern {basedir args} {
    set pos 0
    set cmd ::fileutil::FindGlob
    foreach a $args {
	incr pos
	switch -glob -- $a {
	    --      {break}
	    -regexp {set cmd ::fileutil::FindRegexp}
	    -glob   {set cmd ::fileutil::FindGlob}
	    -*      {return -code error "Unknown option $a"}
	    default {incr pos -1 ; break}
	}
    }

    set args [lrange $args $pos end]

    if {[llength $args] != 1} {
	set pname [lindex [info level 0] 0]
	return -code error \
		"wrong#args for \"$pname\", should be\
		\"$pname basedir ?-regexp|-glob? ?--? patterns\""
    }

    set patterns [lindex $args 0]
    return [find $basedir [list $cmd $patterns]]
}


# ::fileutil::FindRegexp --
#
#	Internal helper. Filter command used by 'findByPattern'
#	to match files based on regular expressions.
#
# Arguments:
#	patterns	List of regular expressions to match against.
#	filename	Name of the file to match against the patterns.
# Results:
#	interesting	A boolean flag. Set to true if the file
#			matches at least one of the patterns.

proc ::fileutil::FindRegexp {patterns filename} {
    foreach p $patterns {
	if {[regexp -- $p $filename]} {
	    return 1
	}
    }
    return 0
}

# ::fileutil::FindGlob --
#
#	Internal helper. Filter command used by 'findByPattern'
#	to match files based on glob expressions.
#
# Arguments:
#	patterns	List of glob expressions to match against.
#	filename	Name of the file to match against the patterns.
# Results:
#	interesting	A boolean flag. Set to true if the file
#			matches at least one of the patterns.

proc ::fileutil::FindGlob {patterns filename} {
    foreach p $patterns {
	if {[string match $p $filename]} {
	    return 1
	}
    }
    return 0
}

# ::fileutil::stripPwd --
#
#	If the specified path references is a path in [pwd] (or [pwd] itself) it
#	is made relative to [pwd]. Otherwise it is left unchanged.
#	In the case of [pwd] itself the result is the string '.'.
#
# Arguments:
#	path		path to modify
#
# Results:
#	path		The (possibly) modified path.

proc ::fileutil::stripPwd {path} {

    # [file split] is used to generate a canonical form for both
    # paths, for easy comparison, and also one which is easy to modify
    # using list commands.

    set pwd [pwd]
    if {[string equal $pwd $path]} {
	return "."
    }

    set pwd   [file split $pwd]
    set npath [file split $path]

    if {[string match ${pwd}* $npath]} {
	set path [eval [linsert [lrange $npath [llength $pwd] end] 0 file join ]]
    }
    return $path
}

# ::fileutil::stripN --
#
#	Removes N elements from the beginning of the path.
#
# Arguments:
#	path		path to modify
#	n		number of elements to strip
#
# Results:
#	path		The modified path

proc ::fileutil::stripN {path n} {
    set path [file split $path]
    if {$n >= [llength $path]} {
	return {}
    } else {
	return [eval [linsert [lrange $path $n end] 0 file join]]
    }
}

# ::fileutil::stripPath --
#
#	If the specified path references/is a path in prefix (or prefix itself) it
#	is made relative to prefix. Otherwise it is left unchanged.
#	In the case of it being prefix itself the result is the string '.'.
#
# Arguments:
#	prefix		prefix to strip from the path.
#	path		path to modify
#
# Results:
#	path		The (possibly) modified path.

if {[string equal $tcl_platform(platform) windows]} {

    # Windows. While paths are stored with letter-case preserved al
    # comparisons have to be done case-insensitive. For reference see
    # SF Tcllib Bug 2499641.

    proc ::fileutil::stripPath {prefix path} {
	# [file split] is used to generate a canonical form for both
	# paths, for easy comparison, and also one which is easy to modify
	# using list commands.

	set prefix [file split $prefix]
	set npath  [file split $path]

	if {[string equal -nocase $prefix $npath]} {
	    return "."
	}

	if {[string match -nocase "${prefix} *" $npath]} {
	    set path [eval [linsert [lrange $npath [llength $prefix] end] 0 file join ]]
	}
	return $path
    }
} else {
    proc ::fileutil::stripPath {prefix path} {
	# [file split] is used to generate a canonical form for both
	# paths, for easy comparison, and also one which is easy to modify
	# using list commands.

	set prefix [file split $prefix]
	set npath  [file split $path]

	if {[string equal $prefix $npath]} {
	    return "."
	}

	if {[string match "${prefix} *" $npath]} {
	    set path [eval [linsert [lrange $npath [llength $prefix] end] 0 file join ]]
	}
	return $path
    }
}

# ::fileutil::jail --
#
#	Ensures that the input path 'filename' stays within the the
#	directory 'jail'. In this way it preventsuser-supplied paths
#	from escaping the jail.
#
# Arguments:
#	jail		The path to the directory the other must
#			not escape from.
#	filename	The path to prevent from escaping.
#
# Results:
#	path		The (possibly) modified path surely within
#			the confines of the jail.

proc fileutil::jail {jail filename} {
    if {![string equal [file pathtype $filename]  "relative"]} {
	# Although the path to check is absolute (or volumerelative on
	# windows) we cannot perform a simple prefix check to see if
	# the path is inside the jail or not. We have to normalize
	# both path and jail and then we can check. If the path is
	# outside we make the original path relative and prefix it
	# with the original jail. We do make the jail pseudo-absolute
	# by prefixing it with the current working directory for that.

	# Normalized jail. Fully resolved sym links, if any. Our main
	# complication is that normalize does not resolve symlinks in the
	# last component of the path given to it, so we add a bogus
	# component, resolve, and then strip it off again. That is why the
	# code is so large and long.

	set njail [eval [list file join] [lrange [file split \
		[Normalize [file join $jail __dummy__]]] 0 end-1]]

	# Normalize filename. Fully resolved sym links, if
	# any. S.a. for an explanation of the complication.

	set nfile [eval [list file join] [lrange [file split \
		[Normalize [file join $filename __dummy__]]] 0 end-1]]

	if {[string match ${njail}* $nfile]} {
	    return $filename
	}

	# Outside the jail, put it inside. ... We normalize the input
	# path lexically for this, to prevent escapes still lurking in
	# the original path. (We cannot use the normalized path,
	# symlinks may have bent it out of shape in unrecognizable ways.

	return [eval [linsert [lrange [file split \
		[lexnormalize $filename]] 1 end] 0 file join [pwd] $jail]]
    } else {
	# The path is relative, consider it as outside
	# implicitly. Normalize it lexically! to prevent escapes, then
	# put the jail in front, use PWD to ensure absoluteness.

	return [eval [linsert [file split [lexnormalize $filename]] 0 \
		file join [pwd] $jail]]
    }
}


# ::fileutil::test --
#
#	Simple API to testing various properties of
#	a path (read, write, file/dir, existence)
#
# Arguments:
#	path	path to test
#	codes	names of the properties to test
#	msgvar	Name of variable to leave an error
#		message in. Optional.
#	label	Label for error message, optional
#
# Results:
#	ok	boolean flag, set if the path passes
#		all tests.

namespace eval ::fileutil {
    variable  test
    array set test {
	read   {readable    {Read access is denied}}
	write  {writable    {Write access is denied}}
	exec   {executable  {Is not executable}}
	exists {exists      {Does not exist}}
	file   {isfile      {Is not a file}}
	dir    {isdirectory {Is not a directory}}
    }
}

proc ::fileutil::test {path codes {msgvar {}} {label {}}} {
    variable test

    if {[string equal $msgvar ""]} {
	set msg ""
    } else {
	upvar 1 $msgvar msg
    }

    if {![string equal $label ""]} {append label { }}

    if {![regexp {^(read|write|exec|exists|file|dir)} $codes]} {
	# Translate single characters into proper codes
	set codes [string map {
	    r read w write e exists x exec f file d dir
	} [split $codes {}]]
    }

    foreach c $codes {
	foreach {cmd text} $test($c) break
	if {![file $cmd $path]} {
	    set msg "$label\"$path\": $text"
	    return 0
	}
    }

    return 1
}

# ::fileutil::cat --
#
#	Tcl implementation of the UNIX "cat" command.  Returns the contents
#	of the specified files.
#
# Arguments:
#	args	names of the files to read, interspersed with options
#		to set encodings, translations, or eofchar.
#
# Results:
#	data	data read from the file.

proc ::fileutil::cat {args} {
    # Syntax: (?options? file)+
    # options = -encoding    ENC
    #         | -translation TRA
    #         | -eofchar     ECH
    #         | --

    if {![llength $args]} {
	# Argument processing stopped with arguments missing.
	return -code error \
		"wrong#args: should be\
		[lindex [info level 0] 0] ?-eofchar|-translation|-encoding arg?+ file ..."
    }

    # We go through the arguments using foreach and keeping track of
    # the index we are at. We do not shift the arguments out to the
    # left. That is inherently quadratic, copying everything down.

    set opts {}
    set mode maybeopt
    set channels {}

    foreach a $args {
	if {[string equal $mode optarg]} {
	    lappend opts $a
	    set mode maybeopt
	    continue
	} elseif {[string equal $mode maybeopt]} {
	    if {[string match -* $a]} {
		switch -exact -- $a {
		    -encoding -
		    -translation -
		    -eofchar {
			lappend opts $a
			set mode optarg
			continue
		    }
		    -- {
			set mode file
			continue
		    }
		    default {
			return -code error \
				"Bad option \"$a\",\
				expected one of\
				-encoding, -eofchar,\
				or -translation"
		    }
		}
	    }
	    # Not an option, but a file. Change mode and fall through.
	    set mode file
	}
	# Process file arguments

	if {[string equal $a -]} {
	    # Stdin reference is special.

	    # Test that the current options are all ok.
	    # For stdin we have to avoid closing it.

	    set old [fconfigure stdin]
	    set fail [catch {
		SetOptions stdin $opts
	    } msg] ; # {}
	    SetOptions stdin $old

	    if {$fail} {
		return -code error $msg
	    }

	    lappend channels [list $a $opts 0]
	} else {
	    if {![file exists $a]} {
		return -code error "Cannot read file \"$a\", does not exist"
	    } elseif {![file isfile $a]} {
		return -code error "Cannot read file \"$a\", is not a file"
	    } elseif {![file readable $a]} {
		return -code error "Cannot read file \"$a\", read access is denied"
	    }

	    # Test that the current options are all ok.
	    set c [open $a r]
	    set fail [catch {
		SetOptions $c $opts
	    } msg] ; # {}
	    close $c
	    if {$fail} {
		return -code error $msg
	    }

	    lappend channels [list $a $opts [file size $a]]
	}

	# We may have more options and files coming after.
	set mode maybeopt
    }

    if {![string equal $mode maybeopt]} {
	# Argument processing stopped with arguments missing.
	return -code error \
		"wrong#args: should be\
		[lindex [info level 0] 0] ?-eofchar|-translation|-encoding arg?+ file ..."
    }

    set data ""
    foreach c $channels {
	foreach {fname opts size} $c break

	if {[string equal $fname -]} {
	    set old [fconfigure stdin]
	    SetOptions stdin $opts
	    append data [read stdin]
	    SetOptions stdin $old
	    continue
	}

	set c [open $fname r]
	SetOptions $c $opts

	if {$size > 0} {
	    # Used the [file size] command to get the size, which
	    # preallocates memory, rather than trying to grow it as
	    # the read progresses.
	    append data [read $c $size]
	} else {
	    # if the file has zero bytes it is either empty, or
	    # something where [file size] reports 0 but the file
	    # actually has data (like the files in the /proc
	    # filesystem on Linux).
	    append data [read $c]
	}
	close $c
    }

    return $data
}

# ::fileutil::writeFile --
#
#	Write the specified data into the named file,
#	creating it if necessary.
#
# Arguments:
#	options...	Options and arguments.
#	filename	Path to the file to write.
#	data		The data to write into the file
#
# Results:
#	None.

proc ::fileutil::writeFile {args} {
    # Syntax: ?options? file data
    # options = -encoding    ENC
    #         | -translation TRA
    #         | -eofchar     ECH
    #         | --

    Spec Writable $args opts fname data

    # Now perform the requested operation.

    file mkdir [file dirname $fname]
    set              c [open $fname w]
    SetOptions      $c $opts
    puts -nonewline $c $data
    close           $c
    return
}

# ::fileutil::appendToFile --
#
#	Append the specified data at the end of the named file,
#	creating it if necessary.
#
# Arguments:
#	options...	Options and arguments.
#	filename	Path to the file to extend.
#	data		The data to extend the file with.
#
# Results:
#	None.

proc ::fileutil::appendToFile {args} {
    # Syntax: ?options? file data
    # options = -encoding    ENC
    #         | -translation TRA
    #         | -eofchar     ECH
    #         | --

    Spec Writable $args opts fname data

    # Now perform the requested operation.

    file mkdir [file dirname $fname]
    set              c [open $fname a]
    SetOptions      $c $opts
    set at    [tell $c]
    puts -nonewline $c $data
    close           $c
    return $at
}

# ::fileutil::insertIntoFile --
#
#	Insert the specified data into the named file,
#	creating it if necessary, at the given locaton.
#
# Arguments:
#	options...	Options and arguments.
#	filename	Path to the file to extend.
#	data		The data to extend the file with.
#
# Results:
#	None.

proc ::fileutil::insertIntoFile {args} {

    # Syntax: ?options? file at data
    # options = -encoding    ENC
    #         | -translation TRA
    #         | -eofchar     ECH
    #         | --

    Spec ReadWritable $args opts fname at data

    set max [file size $fname]
    CheckLocation $at $max insertion

    if {[string length $data] == 0} {
	# Another degenerate case, inserting nothing.
	# Leave the file well enough alone.
	return
    }

    foreach {c o t} [Open2 $fname $opts] break

    # The degenerate cases of both appending and insertion at the
    # beginning of the file allow more optimized implementations of
    # the operation.

    if {$at == 0} {
	puts -nonewline    $o $data
	fcopy           $c $o
    } elseif {$at == $max} {
	fcopy           $c $o
	puts -nonewline    $o $data
    } else {
	fcopy           $c $o -size $at
	puts -nonewline    $o $data
	fcopy           $c $o
    }

    Close2 $fname $t $c $o
    return
}

# ::fileutil::removeFromFile --
#
#	Remove n characters from the named file,
#	starting at the given locaton.
#
# Arguments:
#	options...	Options and arguments.
#	filename	Path to the file to extend.
#	at		Location to start the removal from.
#	n		Number of characters to remove.
#
# Results:
#	None.

proc ::fileutil::removeFromFile {args} {

    # Syntax: ?options? file at n
    # options = -encoding    ENC
    #         | -translation TRA
    #         | -eofchar     ECH
    #         | --

    Spec ReadWritable $args opts fname at n

    set max [file size $fname]
    CheckLocation    $at $max removal
    CheckLength   $n $at $max removal

    if {$n == 0} {
	# Another degenerate case, removing nothing.
	# Leave the file well enough alone.
	return
    }

    foreach {c o t} [Open2 $fname $opts] break

    # The degenerate cases of both removal from the beginning or end
    # of the file allow more optimized implementations of the
    # operation.

    if {$at == 0} {
	seek  $c    $n current
	fcopy $c $o
    } elseif {($at + $n) == $max} {
	fcopy $c $o -size $at
	# Nothing further to copy.
    } else {
	fcopy $c $o -size $at
	seek  $c    $n current
	fcopy $c $o
    }

    Close2 $fname $t $c $o
    return
}

# ::fileutil::replaceInFile --
#
#	Remove n characters from the named file,
#	starting at the given locaton, and replace
#	it with the given data.
#
# Arguments:
#	options...	Options and arguments.
#	filename	Path to the file to extend.
#	at		Location to start the removal from.
#	n		Number of characters to remove.
#	data		The replacement data.
#
# Results:
#	None.

proc ::fileutil::replaceInFile {args} {

    # Syntax: ?options? file at n data
    # options = -encoding    ENC
    #         | -translation TRA
    #         | -eofchar     ECH
    #         | --

    Spec ReadWritable $args opts fname at n data

    set max [file size $fname]
    CheckLocation    $at $max replacement
    CheckLength   $n $at $max replacement

    if {
	($n == 0) &&
	([string length $data] == 0)
    } {
	# Another degenerate case, replacing nothing with
	# nothing. Leave the file well enough alone.
	return
    }

    foreach {c o t} [Open2 $fname $opts] break

    # Check for degenerate cases and handle them separately,
    # i.e. strip the no-op parts out of the general implementation.

    if {$at == 0} {
	if {$n == 0} {
	    # Insertion instead of replacement.

	    puts -nonewline    $o $data
	    fcopy           $c $o

	} elseif {[string length $data] == 0} {
	    # Removal instead of replacement.

	    seek  $c    $n current
	    fcopy $c $o

	} else {
	    # General replacement at front.

	    seek         $c    $n current
	    puts -nonewline $o $data
	    fcopy        $c $o
	}
    } elseif {($at + $n) == $max} {
	if {$n == 0} {
	    # Appending instead of replacement

	    fcopy           $c $o
	    puts -nonewline    $o $data

	} elseif {[string length $data] == 0} {
	    # Truncating instead of replacement

	    fcopy $c $o -size $at
	    # Nothing further to copy.

	} else {
	    # General replacement at end

	    fcopy        $c $o -size $at
	    puts -nonewline $o $data
	}
    } else {
	if {$n == 0} {
	    # General insertion.

	    fcopy           $c $o -size $at
	    puts -nonewline    $o $data
	    fcopy           $c $o

	} elseif {[string length $data] == 0} {
	    # General removal.

	    fcopy $c $o -size $at
	    seek  $c    $n current
	    fcopy $c $o

	} else {
	    # General replacement.

	    fcopy        $c $o -size $at
	    seek         $c    $n current
	    puts -nonewline $o $data
	    fcopy        $c $o
	}
    }

    Close2 $fname $t $c $o
    return
}

# ::fileutil::updateInPlace --
#
#	Run command prefix on the contents of the
#	file and replace them with the result of
#	the command.
#
# Arguments:
#	options...	Options and arguments.
#	filename	Path to the file to extend.
#	cmd		Command prefix to run.
#
# Results:
#	None.

proc ::fileutil::updateInPlace {args} {
    # Syntax: ?options? file cmd
    # options = -encoding    ENC
    #         | -translation TRA
    #         | -eofchar     ECH
    #         | --

    Spec ReadWritable $args opts fname cmd

    # readFile/cat inlined ...

    set             c [open $fname r]
    SetOptions     $c $opts
    set data [read $c]
    close          $c

    # Transformation. Abort and do not modify the target file if an
    # error was raised during this step.

    lappend cmd $data
    set code [catch {uplevel 1 $cmd} res]
    if {$code} {
	return -code $code $res
    }

    # writeFile inlined, with careful preservation of old contents
    # until we are sure that the write was ok.

    if {[catch {
	file rename -force $fname ${fname}.bak

	set              o [open $fname w]
	SetOptions      $o $opts
	puts -nonewline $o $res
	close           $o

	file delete -force ${fname}.bak
    } msg]} {
	if {[file exists ${fname}.bak]} {
	    catch {
		file rename -force ${fname}.bak $fname
	    }
	    return -code error $msg
	}
    }
    return
}

proc ::fileutil::Writable {fname mv} {
    upvar 1 $mv msg
    if {[file exists $fname]} {
	if {![file isfile $fname]} {
	    set msg "Cannot use file \"$fname\", is not a file"
	    return 0
	} elseif {![file writable $fname]} {
	    set msg "Cannot use file \"$fname\", write access is denied"
	    return 0
	}
    }
    return 1
}

proc ::fileutil::ReadWritable {fname mv} {
    upvar 1 $mv msg
    if {![file exists $fname]} {
	set msg "Cannot use file \"$fname\", does not exist"
	return 0
    } elseif {![file isfile $fname]} {
	set msg "Cannot use file \"$fname\", is not a file"
	return 0
    } elseif {![file writable $fname]} {
	set msg "Cannot use file \"$fname\", write access is denied"
	return 0
    } elseif {![file readable $fname]} {
	set msg "Cannot use file \"$fname\", read access is denied"
	return 0
    }
    return 1
}

proc ::fileutil::Spec {check alist ov fv args} {
    upvar 1 $ov opts $fv fname

    set  n [llength $args] ; # Num more args
    incr n                 ; # Count path as well

    set opts {}
    set mode maybeopt

    set at 0
    foreach a $alist {
	if {[string equal $mode optarg]} {
	    lappend opts $a
	    set mode maybeopt
	    incr at
	    continue
	} elseif {[string equal $mode maybeopt]} {
	    if {[string match -* $a]} {
		switch -exact -- $a {
		    -encoding -
		    -translation -
		    -eofchar {
			lappend opts $a
			set mode optarg
			incr at
			continue
		    }
		    -- {
			# Stop processing.
			incr at
			break
		    }
		    default {
			return -code error \
				"Bad option \"$a\",\
				expected one of\
				-encoding, -eofchar,\
				or -translation"
		    }
		}
	    }
	    # Not an option, but a file.
	    # Stop processing.
	    break
	}
    }

    if {([llength $alist] - $at) != $n} {
	# Argument processing stopped with arguments missing, or too
	# many
	return -code error \
		"wrong#args: should be\
		[lindex [info level 1] 0] ?-eofchar|-translation|-encoding arg? file $args"
    }

    set fname [lindex $alist $at]
    incr at
    foreach \
	    var $args \
	    val [lrange $alist $at end] {
	upvar 1 $var A
	set A $val
    }

    # Check given path ...

    if {![eval [linsert $check end $a msg]]} {
	return -code error $msg
    }

    return
}

proc ::fileutil::Open2 {fname opts} {
    set c [open $fname r]
    set t [tempfile]
    set o [open $t     w]

    SetOptions $c $opts
    SetOptions $o $opts

    return [list $c $o $t]
}

proc ::fileutil::Close2 {f temp in out} {
    close $in
    close $out

    file copy   -force $f ${f}.bak
    file rename -force $temp $f
    file delete -force ${f}.bak
    return
}

proc ::fileutil::SetOptions {c opts} {
    if {![llength $opts]} return
    eval [linsert $opts 0 fconfigure $c]
    return
}

proc ::fileutil::CheckLocation {at max label} {
    if {![string is integer -strict $at]} {
	return -code error \
		"Expected integer but got \"$at\""
    } elseif {$at < 0} {
	return -code error \
		"Bad $label point $at, before start of data"
    } elseif {$at > $max} {
	return -code error \
		"Bad $label point $at, behind end of data"
    }
}

proc ::fileutil::CheckLength {n at max label} {
    if {![string is integer -strict $n]} {
	return -code error \
		"Expected integer but got \"$n\""
    } elseif {$n < 0} {
	return -code error \
		"Bad $label size $n"
    } elseif {($at + $n) > $max} {
	return -code error \
		"Bad $label size $n, going behind end of data"
    }
}

# ::fileutil::foreachLine --
#
#	Executes a script for every line in a file.
#
# Arguments:
#	var		name of the variable to contain the lines
#	filename	name of the file to read.
#	cmd		The script to execute.
#
# Results:
#	None.

proc ::fileutil::foreachLine {var filename cmd} {
    upvar 1 $var line
    set fp [open $filename r]

    # -future- Use try/eval from tcllib/control
    catch {
	set code 0
	set result {}
	while {[gets $fp line] >= 0} {
	    set code [catch {uplevel 1 $cmd} result]
	    if {($code != 0) && ($code != 4)} {break}
	}
    }
    close $fp

    if {($code == 0) || ($code == 3) || ($code == 4)} {
        return $result
    }
    if {$code == 1} {
        global errorCode errorInfo
        return \
		-code      $code      \
		-errorcode $errorCode \
		-errorinfo $errorInfo \
		$result
    }
    return -code $code $result
}

# ::fileutil::touch --
#
#	Tcl implementation of the UNIX "touch" command.
#
#	touch [-a] [-m] [-c] [-r ref_file] [-t time] filename ...
#
# Arguments:
#	-a		change the access time only, unless -m also specified
#	-m		change the modification time only, unless -a also specified
#	-c		silently prevent creating a file if it did not previously exist
#	-r ref_file	use the ref_file's time instead of the current time
#	-t time		use the specified time instead of the current time
#			("time" is an integer clock value, like [clock seconds])
#	filename ...	the files to modify
#
# Results
#	None.
#
# Errors:
#	Both of "-r" and "-t" cannot be specified.

if {[package vsatisfies [package provide Tcl] 8.3]} {
    namespace eval ::fileutil {
	namespace export touch
    }

    proc ::fileutil::touch {args} {
        # Don't bother catching errors, just let them propagate up
        
        set options {
            {a          "set the atime only"}
            {m          "set the mtime only"}
            {c          "do not create non-existant files"}
            {r.arg  ""  "use time from ref_file"}
            {t.arg  -1  "use specified time"}
        }
        set usage ": [lindex [info level 0] 0]\
                      \[options] filename ...\noptions:"
        array set params [::cmdline::getoptions args $options $usage]
        
        # process -a and -m options
        set set_atime [set set_mtime "true"]
        if {  $params(a) && ! $params(m)} {set set_mtime "false"}
        if {! $params(a) &&   $params(m)} {set set_atime "false"}
        
        # process -r and -t
        set has_t [expr {$params(t) != -1}]
        set has_r [expr {[string length $params(r)] > 0}]
        if {$has_t && $has_r} {
            return -code error "Cannot specify both -r and -t"
        } elseif {$has_t} {
            set atime [set mtime $params(t)]
        } elseif {$has_r} {
            file stat $params(r) stat
            set atime $stat(atime)
            set mtime $stat(mtime)
        } else {
            set atime [set mtime [clock seconds]]
        }

        # do it
        foreach filename $args {
            if {! [file exists $filename]} {
                if {$params(c)} {continue}
                close [open $filename w]
            }
            if {$set_atime} {file atime $filename $atime}
            if {$set_mtime} {file mtime $filename $mtime}
        }
        return
    }
}

# ::fileutil::fileType --
#
#	Do some simple heuristics to determine file type.
#
#
# Arguments:
#	filename        Name of the file to test.
#
# Results
#	type            Type of the file.  May be a list if multiple tests
#                       are positive (eg, a file could be both a directory 
#                       and a link).  In general, the list proceeds from most
#                       general (eg, binary) to most specific (eg, gif), so
#                       the full type for a GIF file would be 
#                       "binary graphic gif"
#
#                       At present, the following types can be detected:
#
#                       directory
#                       empty
#                       binary
#                       text
#                       script <interpreter>
#                       executable [elf, dos, ne, pe]
#                       binary graphic [gif, jpeg, png, tiff, bitmap, icns]
#                       ps, eps, pdf
#                       html
#                       xml <doctype>
#                       message pgp
#                       compressed [bzip, gzip, zip, tar]
#                       audio [mpeg, wave]
#                       gravity_wave_data_frame
#                       link
#			doctools, doctoc, and docidx documentation files.
#                  

proc ::fileutil::fileType {filename} {
    ;## existence test
    if { ! [ file exists $filename ] } {
        set err "file not found: '$filename'"
        return -code error $err
    }
    ;## directory test
    if { [ file isdirectory $filename ] } {
        set type directory
        if { ! [ catch {file readlink $filename} ] } {
            lappend type link
        }
        return $type
    }
    ;## empty file test
    if { ! [ file size $filename ] } {
        set type empty
        if { ! [ catch {file readlink $filename} ] } {
            lappend type link
        }
        return $type
    }
    set bin_rx {[\x00-\x08\x0b\x0e-\x1f]}

    if { [ catch {
        set fid [ open $filename r ]
        fconfigure $fid -translation binary
        fconfigure $fid -buffersize 1024
        fconfigure $fid -buffering full
        set test [ read $fid 1024 ]
        ::close $fid
    } err ] } {
        catch { ::close $fid }
        return -code error "::fileutil::fileType: $err"
    }

    if { [ regexp $bin_rx $test ] } {
        set type binary
        set binary 1
    } else {
        set type text
        set binary 0
    }

    # SF Tcllib bug [795585]. Allowing whitespace between #!
    # and path of script interpreter

    set metakit 0

    if { [ regexp {^\#\!\s*(\S+)} $test -> terp ] } {
        lappend type script $terp
    } elseif {([regexp "\\\[manpage_begin " $test] &&
	       !([regexp -- {--- !doctools ---} $test] || [regexp -- "!tcl\.tk//DSL doctools//EN//" $test])) ||
                ([regexp -- {--- doctools ---} $test]  || [regexp -- "tcl\.tk//DSL doctools//EN//" $test])} {
	lappend type doctools
    } elseif {([regexp "\\\[toc_begin " $test] &&
	       !([regexp -- {--- !doctoc ---} $test] || [regexp -- "!tcl\.tk//DSL doctoc//EN//" $test])) ||
                ([regexp -- {--- doctoc ---} $test]  || [regexp -- "tcl\.tk//DSL doctoc//EN//" $test])} {
	lappend type doctoc
    } elseif {([regexp "\\\[index_begin " $test] &&
	       !([regexp -- {--- !docidx ---} $test] || [regexp -- "!tcl\.tk//DSL docidx//EN//" $test])) ||
              ([regexp -- {--- docidx ---} $test] || [regexp -- "tcl\.tk//DSL docidx//EN//" $test])} {
	lappend type docidx
    } elseif {[regexp -- "tcl\\.tk//DSL diagram//EN//" $test]} {
	lappend type tkdiagram
    } elseif { $binary && [ regexp {^[\x7F]ELF} $test ] } {
        lappend type executable elf
    } elseif { $binary && [string match "MZ*" $test] } {
        if { [scan [string index $test 24] %c] < 64 } {
            lappend type executable dos
        } else {
            binary scan [string range $test 60 61] s next
            set sig [string range $test $next [expr {$next + 1}]]
            if { $sig == "NE" || $sig == "PE" } {
                lappend type executable [string tolower $sig]
            } else {
                lappend type executable dos
            }
        }
    } elseif { $binary && [string match "BZh91AY\&SY*" $test] } {
        lappend type compressed bzip
    } elseif { $binary && [string match "\x1f\x8b*" $test] } {
        lappend type compressed gzip
    } elseif { $binary && [string range $test 257 262] == "ustar\x00" } {
        lappend type compressed tar
    } elseif { $binary && [string match "\x50\x4b\x03\x04*" $test] } {
        lappend type compressed zip
    } elseif { $binary && [string match "GIF*" $test] } {
        lappend type graphic gif
    } elseif { $binary && [string match "icns*" $test] } {
        lappend type graphic icns bigendian
    } elseif { $binary && [string match "snci*" $test] } {
        lappend type graphic icns smallendian
    } elseif { $binary && [string match "\x89PNG*" $test] } {
        lappend type graphic png
    } elseif { $binary && [string match "\xFF\xD8\xFF*" $test] } {
        binary scan $test x3H2x2a5 marker txt
        if { $marker == "e0" && $txt == "JFIF\x00" } {
            lappend type graphic jpeg jfif
        } elseif { $marker == "e1" && $txt == "Exif\x00" } {
            lappend type graphic jpeg exif
        }
    } elseif { $binary && [string match "MM\x00\**" $test] } {
        lappend type graphic tiff
    } elseif { $binary && [string match "BM*" $test] && [string range $test 6 9] == "\x00\x00\x00\x00" } {
        lappend type graphic bitmap
    } elseif { $binary && [string match "\%PDF\-*" $test] } {
        lappend type pdf
    } elseif { ! $binary && [string match -nocase "*\<html\>*" $test] } {
        lappend type html
    } elseif { [string match "\%\!PS\-*" $test] } {
       lappend type ps
       if { [string match "* EPSF\-*" $test] } {
           lappend type eps
       }
    } elseif { [string match -nocase "*\<\?xml*" $test] } {
        lappend type xml
        if { [ regexp -nocase {\<\!DOCTYPE\s+(\S+)} $test -> doctype ] } {
            lappend type $doctype
        }
    } elseif { [string match {*BEGIN PGP MESSAGE*} $test] } {
        lappend type message pgp
    } elseif { $binary && [string match {IGWD*} $test] } {
        lappend type gravity_wave_data_frame
    } elseif {[string match "JL\x1a\x00*" $test] && ([file size $filename] >= 27)} {
	lappend type metakit smallendian
	set metakit 1
    } elseif {[string match "LJ\x1a\x00*" $test] && ([file size $filename] >= 27)} {
	lappend type metakit bigendian
	set metakit 1
    } elseif { $binary && [string match "RIFF*" $test] && [string range $test 8 11] == "WAVE" } {
        lappend type audio wave
    } elseif { $binary && [string match "ID3*" $test] } {
        lappend type audio mpeg
    } elseif { $binary && [binary scan $test S tmp] && [expr {$tmp & 0xFFE0}] == 65504 } {
        lappend type audio mpeg
    }

    # Additional checks of file contents at the end of the file,
    # possibly pointing into the middle too (attached metakit,
    # attached zip).

    ## Metakit File format: http://www.equi4.com/metakit/metakit-ff.html
    ## Metakit database attached ? ##

    if {!$metakit && ([file size $filename] >= 27)} {
	# The offsets in the footer are in always bigendian format

	if { [ catch {
	    set fid [ open $filename r ]
	    fconfigure $fid -translation binary
	    fconfigure $fid -buffersize 1024
	    fconfigure $fid -buffering full
	    seek $fid -16 end
	    set test [ read $fid 16 ]
	    ::close $fid
	} err ] } {
	    catch { ::close $fid }
	    return -code error "::fileutil::fileType: $err"
	}

	binary scan $test IIII __ hdroffset __ __
	set hdroffset [expr {[file size $filename] - 16 - $hdroffset}]

	# Further checks iff the offset is actually inside the file.

	if {($hdroffset >= 0) && ($hdroffset < [file size $filename])} {
	    # Seek to the specified location and try to match a metakit header
	    # at this location.

	    if { [ catch {
		set         fid [ open $filename r ]
		fconfigure $fid -translation binary
		fconfigure $fid -buffersize 1024
		fconfigure $fid -buffering full
		seek       $fid $hdroffset start
		set test [ read $fid 16 ]
		::close $fid
	    } err ] } {
		catch { ::close $fid }
		return -code error "::fileutil::fileType: $err"
	    }

	    if {[string match "JL\x1a\x00*" $test]} {
		lappend type attached metakit smallendian
		set metakit 1
	    } elseif {[string match "LJ\x1a\x00*" $test]} {
		lappend type attached metakit bigendian
		set metakit 1
	    }
	}
    }

    ## Zip File Format: http://zziplib.sourceforge.net/zzip-parse.html
    ## http://www.pkware.com/products/enterprise/white_papers/appnote.html


    ;## lastly, is it a link?
    if { ! [ catch {file readlink $filename} ] } {
        lappend type link
    }
    return $type
}

# ::fileutil::tempdir --
#
#	Return the correct directory to use for temporary files.
#	Python attempts this sequence, which seems logical:
#
#       1. The directory named by the `TMPDIR' environment variable.
#
#       2. The directory named by the `TEMP' environment variable.
#
#       3. The directory named by the `TMP' environment variable.
#
#       4. A platform-specific location:
#            * On Macintosh, the `Temporary Items' folder.
#
#            * On Windows, the directories `C:\\TEMP', `C:\\TMP',
#              `\\TEMP', and `\\TMP', in that order.
#
#            * On all other platforms, the directories `/tmp',
#              `/var/tmp', and `/usr/tmp', in that order.
#
#       5. As a last resort, the current working directory.
#
#	The code here also does
#
#	0. The directory set by invoking tempdir with an argument.
#	   If this is present it is used exclusively.
#
# Arguments:
#	None.
#
# Side Effects:
#	None.
#
# Results:
#	The directory for temporary files.

proc ::fileutil::tempdir {args} {
    if {[llength $args] > 1} {
	return -code error {wrong#args: should be "::fileutil::tempdir ?path?"}
    } elseif {[llength $args] == 1} {
	variable tempdir    [lindex $args 0]
	variable tempdirSet 1
	return
    }
    return [Normalize [TempDir]]
}

proc ::fileutil::tempdirReset {} {
    variable tempdir    {}
    variable tempdirSet 0
    return
}

proc ::fileutil::TempDir {} {
    global tcl_platform env
    variable tempdir
    variable tempdirSet

    set attempdirs [list]
    set problems   {}

    if {$tempdirSet} {
	lappend attempdirs $tempdir
	lappend problems {User/Application specified tempdir}
    } else {
	foreach tmp {TMPDIR TEMP TMP} {
	    if { [info exists env($tmp)] } {
		lappend attempdirs $env($tmp)
	    } else {
		lappend problems "No environment variable $tmp"
	    }
	}

	switch $tcl_platform(platform) {
	    windows {
		lappend attempdirs "C:\\TEMP" "C:\\TMP" "\\TEMP" "\\TMP"
	    }
	    macintosh {
		lappend attempdirs $env(TRASH_FOLDER)  ;# a better place?
	    }
	    default {
		lappend attempdirs \
		    [file join / tmp] \
		    [file join / var tmp] \
		    [file join / usr tmp]
	    }
	}

	lappend attempdirs [pwd]
    }

    foreach tmp $attempdirs {
	if { [file isdirectory $tmp] && [file writable $tmp] } {
	    return $tmp
	} elseif { ![file isdirectory $tmp] } {
	    lappend problems "Not a directory: $tmp"
	} else {
	    lappend problems "Not writable: $tmp"
	}
    }

    # Fail if nothing worked.
    return -code error "Unable to determine a proper directory for temporary files\n[join $problems \n]"
}

namespace eval ::fileutil {
    variable tempdir    {}
    variable tempdirSet 0
}

# ::fileutil::tempfile --
#
#   generate a temporary file name suitable for writing to
#   the file name will be unique, writable and will be in the 
#   appropriate system specific temp directory
#   Code taken from http://mini.net/tcl/772 attributed to
#    Igor Volobouev and anon.
#
# Arguments:
#   prefix     - a prefix for the filename, p
# Results:
#   returns a file name
#

proc ::fileutil::tempfile {{prefix {}}} {
    return [Normalize [TempFile $prefix]]
}

proc ::fileutil::TempFile {prefix} {
    set tmpdir [tempdir]

    set chars "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"
    set nrand_chars 10
    set maxtries 10
    set access [list RDWR CREAT EXCL TRUNC]
    set permission 0600
    set channel ""
    set checked_dir_writable 0
    set mypid [pid]
    for {set i 0} {$i < $maxtries} {incr i} {
 	set newname $prefix
 	for {set j 0} {$j < $nrand_chars} {incr j} {
 	    append newname [string index $chars \
		    [expr {int(rand()*62)}]]
 	}
	set newname [file join $tmpdir $newname]
 	if {[file exists $newname]} {
 	    after 1
 	} else {
 	    if {[catch {open $newname $access $permission} channel]} {
 		if {!$checked_dir_writable} {
 		    set dirname [file dirname $newname]
 		    if {![file writable $dirname]} {
 			return -code error "Directory $dirname is not writable"
 		    }
 		    set checked_dir_writable 1
 		}
 	    } else {
 		# Success
		close $channel
 		return $newname
 	    }
 	}
    }
    if {[string compare $channel ""]} {
 	return -code error "Failed to open a temporary file: $channel"
    } else {
 	return -code error "Failed to find an unused temporary file name"
    }
}

# ::fileutil::install --
#
#	Tcl version of the 'install' command, which copies files from
#	one places to another and also optionally sets some attributes
#	such as group, owner, and permissions.
#
# Arguments:
#	-m		Change the file permissions to the specified
#                       value.  Valid arguments are those accepted by
#			file attributes -permissions
#
# Results:
#	None.

# TODO - add options for group/owner manipulation.

proc ::fileutil::install {args} {
    set options {
	{m.arg "" "Set permission mode"}
    }
    set usage ": [lindex [info level 0] 0]\
\[options] source destination \noptions:"
    array set params [::cmdline::getoptions args $options $usage]
    # Args should now just be the source and destination.
    if { [llength $args] < 2 } {
	return -code error $usage
    }
    set src [lindex $args 0]
    set dst [lindex $args 1]
    file copy -force $src $dst
    if { $params(m) != "" } {
	set targets [::fileutil::find $dst]
	foreach fl $targets {
	    file attributes $fl -permissions $params(m)
	}
    }
}

# ### ### ### ######### ######### #########

proc ::fileutil::lexnormalize {sp} {
    set spx [file split $sp]

    # Resolution of embedded relative modifiers (., and ..).

    if {
	([lsearch -exact $spx . ] < 0) &&
	([lsearch -exact $spx ..] < 0)
    } {
	# Quick path out if there are no relative modifiers
	return $sp
    }

    set absolute [expr {![string equal [file pathtype $sp] relative]}]
    # A volumerelative path counts as absolute for our purposes.

    set sp $spx
    set np {}
    set noskip 1

    while {[llength $sp]} {
	set ele    [lindex $sp 0]
	set sp     [lrange $sp 1 end]
	set islast [expr {[llength $sp] == 0}]

	if {[string equal $ele ".."]} {
	    if {
		($absolute  && ([llength $np] >  1)) ||
		(!$absolute && ([llength $np] >= 1))
	    } {
		# .. : Remove the previous element added to the
		# new path, if there actually is enough to remove.
		set np [lrange $np 0 end-1]
	    }
	} elseif {[string equal $ele "."]} {
	    # Ignore .'s, they stay at the current location
	    continue
	} else {
	    # A regular element.
	    lappend np $ele
	}
    }
    if {[llength $np] > 0} {
	return [eval [linsert $np 0 file join]]
	# 8.5: return [file join {*}$np]
    }
    return {}
}

# ### ### ### ######### ######### #########
## Forward compatibility. Some routines require path normalization,
## something we have supported by the builtin 'file' only since Tcl
## 8.4. For versions of Tcl before that, to be supported by the
## module, we implement a normalizer in Tcl itself. Slow, but working.

if {[package vcompare [package provide Tcl] 8.4] < 0} {
    # Pre 8.4. We do not have 'file normalize'. We create an
    # approximation for it based on earlier commands.

    # ... Hm. This is lexical normalization. It does not resolve
    # symlinks in the path to their origin.

    proc ::fileutil::Normalize {sp} {
	set sp [file split $sp]

	# Conversion of the incoming path to absolute.
	if {[string equal [file pathtype [lindex $sp 0]] "relative"]} {
	    set sp [file split [eval [list file join [pwd]] $sp]]
	}

	# Resolution of symlink components, and embedded relative
	# modifiers (., and ..).

	set np {}
	set noskip 1
	while {[llength $sp]} {
	    set ele    [lindex $sp 0]
	    set sp     [lrange $sp 1 end]
	    set islast [expr {[llength $sp] == 0}]

	    if {[string equal $ele ".."]} {
		if {[llength $np] > 1} {
		    # .. : Remove the previous element added to the
		    # new path, if there actually is enough to remove.
		    set np [lrange $np 0 end-1]
		}
	    } elseif {[string equal $ele "."]} {
		# Ignore .'s, they stay at the current location
		continue
	    } else {
		# A regular element. If it is not the last component
		# then check if the combination is a symlink, and if
		# yes, resolve it.

		lappend np $ele

		if {!$islast && $noskip} {
		    # The flag 'noskip' is technically not required,
		    # just 'file exists'. However if a path P does not
		    # exist, then all longer paths starting with P can
		    # not exist either, and using the flag to store
		    # this knowledge then saves us a number of
		    # unnecessary stat calls. IOW this a performance
		    # optimization.

		    set p [eval file join $np]
		    set noskip [file exists $p]
		    if {$noskip} {
			if {[string equal link [file type $p]]} {
			    set dst [file readlink $p]

			    # We always push the destination in front of
			    # the source path (in expanded form). So that
			    # we handle .., .'s, and symlinks inside of
			    # this path as well. An absolute path clears
			    # the result, a relative one just removes the
			    # last, now resolved component.

			    set sp [eval [linsert [file split $dst] 0 linsert $sp 0]]

			    if {![string equal relative [file pathtype $dst]]} {
				# Absolute|volrelative destination, clear
				# result, we have to start over.
				set np {}
			    } else {
				# Relative link, just remove the resolved
				# component again.
				set np [lrange $np 0 end-1]
			    }
			}
		    }
		}
	    }
	}
	if {[llength $np] > 0} {
	    return [eval file join $np]
	}
	return {}
    }
} else {
    proc ::fileutil::Normalize {sp} {
	file normalize $sp
    }
}

# ::fileutil::relative --
#
#	Taking two _directory_ paths, a base and a destination, computes the path
#	of the destination relative to the base.
#
# Arguments:
#	base	The path to make the destination relative to.
#	dst	The destination path
#
# Results:
#	The path of the destination, relative to the base.

proc ::fileutil::relative {base dst} {
    # Ensure that the link to directory 'dst' is properly done relative to
    # the directory 'base'.

    if {![string equal [file pathtype $base] [file pathtype $dst]]} {
	return -code error "Unable to compute relation for paths of different pathtypes: [file pathtype $base] vs. [file pathtype $dst], ($base vs. $dst)"
    }

    set base [lexnormalize [file join [pwd] $base]]
    set dst  [lexnormalize [file join [pwd] $dst]]

    set save $dst
    set base [file split $base]
    set dst  [file split $dst]

    while {[string equal [lindex $dst 0] [lindex $base 0]]} {
	set dst  [lrange $dst  1 end]
	set base [lrange $base 1 end]
	if {![llength $dst]} {break}
    }

    set dstlen  [llength $dst]
    set baselen [llength $base]

    if {($dstlen == 0) && ($baselen == 0)} {
	# Cases:
	# (a) base == dst

	set dst .
    } else {
	# Cases:
	# (b) base is: base/sub = sub
	#     dst  is: base     = {}

	# (c) base is: base     = {}
	#     dst  is: base/sub = sub

	while {$baselen > 0} {
	    set dst [linsert $dst 0 ..]
	    incr baselen -1
	}
	# 8.5: set dst [file join {*}$dst]
	set dst [eval [linsert $dst 0 file join]]
    }

    return $dst
}

# ::fileutil::relativeUrl --
#
#	Taking two _file_ paths, a base and a destination, computes the path
#	of the destination relative to the base, from the inside of the base.
#
#	This is how a browser resolves relative links in a file, hence the
#	url in the command name.
#
# Arguments:
#	base	The file path to make the destination relative to.
#	dst	The destination file path
#
# Results:
#	The path of the destination file, relative to the base file.

proc ::fileutil::relativeUrl {base dst} {
    # Like 'relative', but for links from _inside_ a file to a
    # different file.

    if {![string equal [file pathtype $base] [file pathtype $dst]]} {
	return -code error "Unable to compute relation for paths of different pathtypes: [file pathtype $base] vs. [file pathtype $dst], ($base vs. $dst)"
    }

    set base [lexnormalize [file join [pwd] $base]]
    set dst  [lexnormalize [file join [pwd] $dst]]

    set basedir [file dirname $base]
    set dstdir  [file dirname $dst]

    set dstdir  [relative $basedir $dstdir]

    # dstdir == '.' on input => dstdir output has trailing './'. Strip
    # this superfluous segment off.

    if {[string equal $dstdir "."]} {
	return [file tail $dst]
    } elseif {[string equal [file tail $dstdir] "."]} {
	return [file join [file dirname $dstdir] [file tail $dst]]
    } else {
	return [file join $dstdir [file tail $dst]]
    }
}

# ::fileutil::fullnormalize --
#
#	Normalizes a path completely. I.e. a symlink in the last
#	element is resolved as well, not only symlinks in the higher
#	elements.
#
# Arguments:
#	path	The path to normalize
#
# Results:
#	The input path with all symlinks resolved.

proc ::fileutil::fullnormalize {path} {
    # When encountering symlinks in a file copy operation Tcl copies
    # the link, not the contents of the file it references. There are
    # situations there this is not acceptable. For these this command
    # resolves all symbolic links in the path, including in the last
    # element of the path. A "file copy" using the return value of
    # this command copies an actual file, it will not encounter
    # symlinks.

    # BUG / WORKAROUND. Using the / instead of the join seems to work
    # around a bug in the path handling on windows which can break the
    # core 'file normalize' for symbolic links. This was exposed by
    # the find testsuite which could not reproduced outside. I believe
    # that there is some deep path bug in the core triggered under
    # special circumstances. Use of / likely forces a refresh through
    # the string rep and so avoids the problem with the path intrep.

    return [file dirname [Normalize $path/__dummy__]]
    #return [file dirname [Normalize [file join $path __dummy__]]]
}