1Sort::Naturally(3) User Contributed Perl Documentation Sort::Naturally(3)
2
6 Sort::Naturally -- sort lexically, but sort numeral parts numerically
7
9 @them = nsort(qw(
10 foo12a foo12z foo13a foo 14 9x foo12 fooa foolio Foolio Foo12a
11 ));
12 print join(' ', @them), "\n";
13 Prints:
15 9x 14 foo fooa foolio Foolio foo12 foo12a Foo12a foo12z foo13a
17 (Or "foo12a" + "Foo12a" and "foolio" + "Foolio" and might be switched,
19 depending on your locale.)
20
22 This module exports two functions, "nsort" and "ncmp"; they are used in
23 implementing my idea of a "natural sorting" algorithm. Under natural
24 sorting, numeric substrings are compared numerically, and other word-
25 characters are compared lexically.
26 This is the way I define natural sorting:
28 · Non-numeric word-character substrings are sorted lexically, case-
30 insensitively: "Foo" comes between "fish" and "fowl".
31 · Numeric substrings are sorted numerically: "100" comes after "20",
33 not before.
34 · \W substrings (neither words-characters nor digits) are ignored.
36 · Our use of \w, \d, \D, and \W is locale-sensitive: Sort::Naturally
38 uses a "use locale" statement.
39 · When comparing two strings, where a numeric substring in one place
41 is not up against a numeric substring in another, the non-numeric
42 always comes first. This is fudged by reading pretending that the
43 lack of a number substring has the value -1, like so:
44 foo => "foo", -1
46 foobar => "foo", -1, "bar"
47 foo13 => "foo", 13,
48 foo13xyz => "foo", 13, "xyz"
49 That's so that "foo" will come before "foo13", which will come
51 before "foobar".
52 · The start of a string is exceptional: leading non-\W (non-word,
54 non-digit) components are are ignored, and numbers come before
55 letters.
56 · I define "numeric substring" just as sequences matching m/\d+/ --
58 scientific notation, commas, decimals, etc., are not seen. If your
59 data has thousands separators in numbers ("20,000 Leagues Under The
60 Sea" or "20.000 lieues sous les mers"), consider stripping them
61 before feeding them to "nsort" or "ncmp".
62 The nsort function
64 This function takes a list of strings, and returns a copy of the list,
65 sorted.
66 This is what most people will want to use:
68 @stuff = nsort(...list...);
70 When nsort needs to compare non-numeric substrings, it uses Perl's "lc"
72 function in scope of a <use locale>. And when nsort needs to lowercase
73 things, it uses Perl's "lc" function in scope of a <use locale>. If
74 you want nsort to use other functions instead, you can specify them in
75 an arrayref as the first argument to nsort:
76 @stuff = nsort( [
78 \&string_comparator, # optional
79 \&lowercaser_function # optional
80 ],
81 ...list...
82 );
83 If you want to specify a string comparator but no lowercaser, then the
85 options list is "[\&comparator, '']" or "[\&comparator]". If you want
86 to specify no string comparator but a lowercaser, then the options list
87 is "['', \&lowercaser]".
88 Any comparator you specify is called as "$comparator->($left, $right)",
90 and, like a normal Perl "cmp" replacement, must return -1, 0, or 1
91 depending on whether the left argument is stringwise less than, equal
92 to, or greater than the right argument.
93 Any lowercaser function you specify is called as "$lowercased =
95 $lowercaser->($original)". The routine must not modify its $_[0].
96 The ncmp function
98 Often, when sorting non-string values like this:
99 @objects_sorted = sort { $a->tag cmp $b->tag } @objects;
101 ...or even in a Schwartzian transform, like this:
103 @strings =
105 map $_->[0]
106 sort { $a->[1] cmp $b->[1] }
107 map { [$_, make_a_sort_key_from($_) ]
108 @_
109 ;
110 ...you wight want something that replaces not "sort", but "cmp".
112 That's what Sort::Naturally's "ncmp" function is for. Call it with the
113 syntax "ncmp($left,$right)" instead of "$left cmp $right", but
114 otherwise it's a fine replacement:
115 @objects_sorted = sort { ncmp($a->tag,$b->tag) } @objects;
117 @strings =
119 map $_->[0]
120 sort { ncmp($a->[1], $b->[1]) }
121 map { [$_, make_a_sort_key_from($_) ]
122 @_
123 ;
124 Just as with "nsort" can take different a string-comparator and/or
126 lowercaser, you can do the same with "ncmp", by passing an arrayref as
127 the first argument:
128 ncmp( [
130 \&string_comparator, # optional
131 \&lowercaser_function # optional
132 ],
133 $left, $right
134 )
135 You might get string comparators from Sort::ArbBiLex.
137
139 · This module is not a substitute for Sort::Versions! If you just
140 need proper version sorting, use that!
141 · If you need something that works sort of like this module's
143 functions, but not quite the same, consider scouting thru this
144 module's source code, and adapting what you see. Besides the
145 functions that actually compile in this module, after the POD,
146 there's several alternate attempts of mine at natural sorting
147 routines, which are not compiled as part of the module, but which
148 you might find useful. They should all be working implementations
149 of slightly different algorithms (all of them based on Martin
150 Pool's "nsort") which I eventually discarded in favor of my
151 algorithm. If you are having to naturally-sort very large data
152 sets, and sorting is getting ridiculously slow, you might consider
153 trying one of those discarded functions -- I have a feeling they
154 might be faster on large data sets. Benchmark them on your data
155 and see. (Unless you need the speed, don't bother. Hint:
156 substitute "sort" for "nsort" in your code, and unless your program
157 speeds up drastically, it's not the sorting that's slowing things
158 down. But if it is "nsort" that's slowing things down, consider
159 just:
160 if(@set >= SOME_VERY_BIG_NUMBER) {
162 no locale; # vroom vroom
163 @sorted = sort(@set); # feh, good enough
164 } elsif(@set >= SOME_BIG_NUMBER) {
165 use locale;
166 @sorted = sort(@set); # feh, good enough
167 } else {
168 # but keep it pretty for normal cases
169 @sorted = nsort(@set);
170 }
171 · If you do adapt the routines in this module, email me; I'd just be
173 interested in hearing about it.
174 · Thanks to the EFNet #perl people for encouraging this module,
176 especially magister and a-mused.
177
179 Copyright 2001, Sean M. Burke "sburke@cpan.org", all rights reserved.
180 This program is free software; you can redistribute it and/or modify it
181 under the same terms as Perl itself.
182 This program is distributed in the hope that it will be useful, but
184 without any warranty; without even the implied warranty of
185 merchantability or fitness for a particular purpose.
186
188 Sean M. Burke "sburke@cpan.org"
189perl v5.30.0 2019-07-26 Sort::Naturally(3)