Commit | Line | Data |
de1df517 |
1 | package encoding::warnings; |
6309fe6d |
2 | $encoding::warnings::VERSION = '0.10'; |
de1df517 |
3 | |
4 | use strict; |
6309fe6d |
5 | use 5.007; |
de1df517 |
6 | |
7 | =head1 NAME |
8 | |
9 | encoding::warnings - Warn on implicit encoding conversions |
10 | |
11 | =head1 VERSION |
12 | |
6309fe6d |
13 | This document describes version 0.10 of encoding::warnings, released |
14 | July 7, 2006. |
de1df517 |
15 | |
16 | =head1 SYNOPSIS |
17 | |
18 | use encoding::warnings; # or 'FATAL' to raise fatal exceptions |
19 | |
20 | utf8::encode($a = chr(20000)); # a byte-string (raw bytes) |
21 | $b = chr(20000); # a unicode-string (wide characters) |
22 | |
23 | # "Bytes implicitly upgraded into wide characters as iso-8859-1" |
24 | $c = $a . $b; |
25 | |
26 | =head1 DESCRIPTION |
27 | |
28 | =head2 Overview of the problem |
29 | |
30 | By default, there is a fundamental asymmetry in Perl's unicode model: |
31 | implicit upgrading from byte-strings to unicode-strings assumes that |
32 | they were encoded in I<ISO 8859-1 (Latin-1)>, but unicode-strings are |
33 | downgraded with UTF-8 encoding. This happens because the first 256 |
34 | codepoints in Unicode happens to agree with Latin-1. |
35 | |
36 | However, this silent upgrading can easily cause problems, if you happen |
37 | to mix unicode strings with non-Latin1 data -- i.e. byte-strings encoded |
38 | in UTF-8 or other encodings. The error will not manifest until the |
39 | combined string is written to output, at which time it would be impossible |
40 | to see where did the silent upgrading occur. |
41 | |
42 | =head2 Detecting the problem |
43 | |
44 | This module simplifies the process of diagnosing such problems. Just put |
45 | this line on top of your main program: |
46 | |
47 | use encoding::warnings; |
48 | |
49 | Afterwards, implicit upgrading of high-bit bytes will raise a warning. |
50 | Ex.: C<Bytes implicitly upgraded into wide characters as iso-8859-1 at |
51 | - line 7>. |
52 | |
53 | However, strings composed purely of ASCII code points (C<0x00>..C<0x7F>) |
54 | will I<not> trigger this warning. |
55 | |
56 | You can also make the warnings fatal by importing this module as: |
57 | |
58 | use encoding::warnings 'FATAL'; |
59 | |
60 | =head2 Solving the problem |
61 | |
62 | Most of the time, this warning occurs when a byte-string is concatenated |
63 | with a unicode-string. There are a number of ways to solve it: |
64 | |
65 | =over 4 |
66 | |
67 | =item * Upgrade both sides to unicode-strings |
68 | |
69 | If your program does not need compatibility for Perl 5.6 and earlier, |
70 | the recommended approach is to apply appropriate IO disciplines, so all |
71 | data in your program become unicode-strings. See L<encoding>, L<open> and |
72 | L<perlfunc/binmode> for how. |
73 | |
74 | =item * Downgrade both sides to byte-strings |
75 | |
76 | The other way works too, especially if you are sure that all your data |
77 | are under the same encoding, or if compatibility with older versions |
78 | of Perl is desired. |
79 | |
80 | You may downgrade strings with C<Encode::encode> and C<utf8::encode>. |
81 | See L<Encode> and L<utf8> for details. |
82 | |
83 | =item * Specify the encoding for implicit byte-string upgrading |
84 | |
85 | If you are confident that all byte-strings will be in a specific |
86 | encoding like UTF-8, I<and> need not support older versions of Perl, |
87 | use the C<encoding> pragma: |
88 | |
89 | use encoding 'utf8'; |
90 | |
91 | Similarly, this will silence warnings from this module, and preserve the |
92 | default behaviour: |
93 | |
94 | use encoding 'iso-8859-1'; |
95 | |
96 | However, note that C<use encoding> actually had three distinct effects: |
97 | |
98 | =over 4 |
99 | |
100 | =item * PerlIO layers for B<STDIN> and B<STDOUT> |
101 | |
102 | This is similar to what L<open> pragma does. |
103 | |
104 | =item * Literal conversions |
105 | |
106 | This turns I<all> literal string in your program into unicode-strings |
107 | (equivalent to a C<use utf8>), by decoding them using the specified |
108 | encoding. |
109 | |
110 | =item * Implicit upgrading for byte-strings |
111 | |
112 | This will silence warnings from this module, as shown above. |
113 | |
114 | =back |
115 | |
116 | Because literal conversions also work on empty strings, it may surprise |
117 | some people: |
118 | |
119 | use encoding 'big5'; |
120 | |
121 | my $byte_string = pack("C*", 0xA4, 0x40); |
122 | print length $a; # 2 here. |
123 | $a .= ""; # concatenating with a unicode string... |
124 | print length $a; # 1 here! |
125 | |
126 | In other words, do not C<use encoding> unless you are certain that the |
127 | program will not deal with any raw, 8-bit binary data at all. |
128 | |
129 | However, the C<Filter =E<gt> 1> flavor of C<use encoding> will I<not> |
130 | affect implicit upgrading for byte-strings, and is thus incapable of |
131 | silencing warnings from this module. See L<encoding> for more details. |
132 | |
133 | =back |
134 | |
135 | =head1 CAVEATS |
136 | |
6309fe6d |
137 | For Perl 5.9.4 or later, this module's effect is lexical. |
138 | |
139 | For Perl versions prior to 5.9.4, this module affects the whole script, |
140 | instead of inside its lexical block. |
de1df517 |
141 | |
142 | =cut |
143 | |
144 | # Constants. |
145 | sub ASCII () { 0 } |
146 | sub LATIN1 () { 1 } |
147 | sub FATAL () { 2 } |
148 | |
149 | # Install a ${^ENCODING} handler if no other one are already in place. |
150 | sub import { |
151 | my $class = shift; |
152 | my $fatal = shift || ''; |
153 | |
154 | local $@; |
155 | return if ${^ENCODING} and ref(${^ENCODING}) ne $class; |
156 | return unless eval { require Encode; 1 }; |
157 | |
158 | my $ascii = Encode::find_encoding('us-ascii') or return; |
159 | my $latin1 = Encode::find_encoding('iso-8859-1') or return; |
160 | |
161 | # Have to undef explicitly here |
162 | undef ${^ENCODING}; |
163 | |
164 | # Install a warning handler for decode() |
6309fe6d |
165 | my $decoder = bless( |
de1df517 |
166 | [ |
167 | $ascii, |
168 | $latin1, |
169 | (($fatal eq 'FATAL') ? 'Carp::croak' : 'Carp::carp'), |
170 | ], $class, |
171 | ); |
6309fe6d |
172 | |
173 | ${^ENCODING} = $decoder; |
174 | $^H{$class} = 1; |
175 | } |
176 | |
177 | sub unimport { |
178 | my $class = shift; |
179 | $^H{$class} = undef; |
de1df517 |
180 | } |
181 | |
182 | # Don't worry about source code literals. |
183 | sub cat_decode { |
184 | my $self = shift; |
185 | return $self->[LATIN1]->cat_decode(@_); |
186 | } |
187 | |
188 | # Warn if the data is not purely US-ASCII. |
189 | sub decode { |
190 | my $self = shift; |
191 | |
6309fe6d |
192 | DO_WARN: { |
193 | if ($] >= 5.009004) { |
194 | my $hints = (caller(0))[10]; |
195 | $hints->{ref($self)} or last DO_WARN; |
196 | } |
197 | |
198 | local $@; |
199 | my $rv = eval { $self->[ASCII]->decode($_[0], Encode::FB_CROAK()) }; |
200 | return $rv unless $@; |
201 | |
202 | require Carp; |
203 | no strict 'refs'; |
204 | $self->[FATAL]->( |
205 | "Bytes implicitly upgraded into wide characters as iso-8859-1" |
206 | ); |
207 | |
208 | } |
de1df517 |
209 | |
de1df517 |
210 | return $self->[LATIN1]->decode(@_); |
211 | } |
212 | |
213 | sub name { 'iso-8859-1' } |
214 | |
215 | 1; |
216 | |
217 | __END__ |
218 | |
219 | =head1 SEE ALSO |
220 | |
221 | L<perlunicode>, L<perluniintro> |
222 | |
223 | L<open>, L<utf8>, L<encoding>, L<Encode> |
224 | |
225 | =head1 AUTHORS |
226 | |
6309fe6d |
227 | Audrey Tang |
de1df517 |
228 | |
229 | =head1 COPYRIGHT |
230 | |
6309fe6d |
231 | Copyright 2004, 2005, 2006 by Audrey Tang E<lt>cpan@audreyt.orgE<gt>. |
de1df517 |
232 | |
233 | This program is free software; you can redistribute it and/or modify it |
234 | under the same terms as Perl itself. |
235 | |
236 | See L<http://www.perl.com/perl/misc/Artistic.html> |
237 | |
238 | =cut |