1# split.tcl --
2#
3#	Various ways of splitting a string.
4#
5# Copyright (c) 2000      by Ajuba Solutions.
6# Copyright (c) 2000      by Eric Melski <ericm@ajubasolutions.com>
7# Copyright (c) 2001      by Reinhard Max <max@suse.de>
8# Copyright (c) 2003      by Pat Thoyts <patthoyts@users.sourceforge.net>
9# Copyright (c) 2001-2006 by Andreas Kupries <andreas_kupries@users.sourceforge.net>
10#
11# See the file "license.terms" for information on usage and redistribution
12# of this file, and for a DISCLAIMER OF ALL WARRANTIES.
13#
14# RCS: @(#) $Id: split.tcl,v 1.7 2006/04/21 04:42:28 andreas_kupries Exp $
15
16# ### ### ### ######### ######### #########
17## Requirements
18
19package require Tcl 8.2
20
21namespace eval ::textutil::split {}
22
23########################################################################
24# This one was written by Bob Techentin (RWT in Tcl'ers Wiki):
25# http://www.techentin.net
26# mailto:techentin.robert@mayo.edu
27#
28# Later, he send me an email stated that I can use it anywhere, because
29# no copyright was added, so the code is defacto in the public domain.
30#
31# You can found it in the Tcl'ers Wiki here:
32# http://mini.net/cgi-bin/wikit/460.html
33#
34# Bob wrote:
35# If you need to split string into list using some more complicated rule
36# than builtin split command allows, use following function. It mimics
37# Perl split operator which allows regexp as element separator, but,
38# like builtin split, it expects string to split as first arg and regexp
39# as second (optional) By default, it splits by any amount of whitespace.
40# Note that if you add parenthesis into regexp, parenthesed part of separator
41# would be added into list as additional element. Just like in Perl. -- cary
42#
43# Speed improvement by Reinhard Max:
44# Instead of repeatedly copying around the not yet matched part of the
45# string, I use [regexp]'s -start option to restrict the match to that
46# part. This reduces the complexity from something like O(n^1.5) to
47# O(n). My test case for that was:
48#
49# foreach i {1 10 100 1000 10000} {
50#     set s [string repeat x $i]
51#     puts [time {splitx $s .}]
52# }
53#
54
55if {[package vsatisfies [package provide Tcl] 8.3]} {
56
57    proc ::textutil::split::splitx {str {regexp {[\t \r\n]+}}} {
58        # Bugfix 476988
59        if {[string length $str] == 0} {
60            return {}
61        }
62        if {[string length $regexp] == 0} {
63            return [::split $str ""]
64        }
65        set list  {}
66        set start 0
67        while {[regexp -start $start -indices -- $regexp $str match submatch]} {
68            foreach {subStart subEnd} $submatch break
69            foreach {matchStart matchEnd} $match break
70            incr matchStart -1
71            incr matchEnd
72            lappend list [string range $str $start $matchStart]
73            if {$subStart >= $start} {
74                lappend list [string range $str $subStart $subEnd]
75            }
76            set start $matchEnd
77        }
78        lappend list [string range $str $start end]
79        return $list
80    }
81
82} else {
83    # For tcl <= 8.2 we do not have regexp -start...
84    proc ::textutil::split::splitx [list str [list regexp "\[\t \r\n\]+"]] {
85
86        if {[string length $str] == 0} {
87            return {}
88        }
89        if {[string length $regexp] == 0} {
90            return [::split $str {}]
91        }
92
93        set list  {}
94        while {[regexp -indices -- $regexp $str match submatch]} {
95            lappend list [string range $str 0 [expr {[lindex $match 0] -1}]]
96            if {[lindex $submatch 0] >= 0} {
97                lappend list [string range $str [lindex $submatch 0] \
98                                  [lindex $submatch 1]]
99            }
100            set str [string range $str [expr {[lindex $match 1]+1}] end]
101        }
102        lappend list $str
103        return $list
104    }
105
106}
107
108#
109# splitn --
110#
111# splitn splits the string $str into chunks of length $len.  These
112# chunks are returned as a list.
113#
114# If $str really contains a ByteArray object (as retrieved from binary
115# encoded channels) splitn must honor this by splitting the string
116# into chunks of $len bytes.
117#
118# It is an error to call splitn with a nonpositive $len.
119#
120# If splitn is called with an empty string, it returns the empty list.
121#
122# If the length of $str is not an entire multiple of the chunk length,
123# the last chunk in the generated list will be shorter than $len.
124#
125# The implementation presented here was given by Bryan Oakley, as
126# part of a ``contest'' I staged on c.l.t in July 2004.  I selected
127# this version, as it does not rely on runtime generated code, is
128# very fast for chunk size one, not too bad in all the other cases,
129# and uses [split] or [string range] which have been around for quite
130# some time.
131#
132# -- Robert Suetterlin (robert@mpe.mpg.de)
133#
134proc ::textutil::split::splitn {str {len 1}} {
135
136    if {$len <= 0} {
137        return -code error "len must be > 0"
138    }
139
140    if {$len == 1} {
141        return [split $str {}]
142    }
143
144    set result [list]
145    set max [string length $str]
146    set i 0
147    set j [expr {$len -1}]
148    while {$i < $max} {
149        lappend result [string range $str $i $j]
150        incr i $len
151        incr j $len
152    }
153
154    return $result
155}
156
157# ### ### ### ######### ######### #########
158## Data structures
159
160namespace eval ::textutil::split {
161    namespace export splitx splitn
162}
163
164# ### ### ### ######### ######### #########
165## Ready
166
167package provide textutil::split 0.7
168