1 # Test the /a, /d, etc regex modifiers
14 # Each case is a valid element of its hash key. Choose, where available, an
15 # ASCII-range, Latin-1 non-ASCII range, and above Latin1 range code point.
17 '\w' => [ ord("A"), 0xE2, 0x16B ], # Below expects these to all be alpha
18 '\d' => [ ord("0"), 0x0662 ],
19 '\s' => [ ord("\t"), 0xA0, 0x1680 ], # Below expects these to be [:blank:]
20 '[:cntrl:]' => [ 0x00, 0x88 ],
21 '[:graph:]' => [ ord("&"), 0xF7, 0x02C7 ], # Below expects these to be
23 '[:lower:]' => [ ord("g"), 0xE3, 0x0127 ],
24 '[:punct:]' => [ ord("!"), 0xBF, 0x055C ],
25 '[:upper:]' => [ ord("G"), 0xC3, 0x0126 ],
26 '[:xdigit:]' => [ ord("4"), 0xFF15 ],
29 $testcases{'[:digit:]'} = $testcases{'\d'};
30 $testcases{'[:alnum:]'} = $testcases{'\w'};
31 $testcases{'[:alpha:]'} = $testcases{'\w'};
32 $testcases{'[:blank:]'} = $testcases{'\s'};
33 $testcases{'[:print:]'} = $testcases{'[:graph:]'};
34 $testcases{'[:space:]'} = $testcases{'\s'};
35 $testcases{'[:word:]'} = $testcases{'\w'};
37 my @charsets = qw(a d u aa);
38 if (! is_miniperl()) {
40 my $current_locale = POSIX::setlocale( &POSIX::LC_ALL, "C") // "";
41 if ($current_locale eq 'C') {
44 # Some locale implementations don't have the 128-255 characters all
45 # mean nothing. Skip the locale tests in that situation
46 for my $i (128 .. 255) {
47 goto bad_locale if chr($i) =~ /[[:print:]]/;
54 # For each possible character set...
55 foreach my $charset (@charsets) {
58 foreach my $upgrade ("", 'utf8::upgrade($a); ') {
60 # reverse gets the, \w, \s, \d first.
61 for my $class (reverse sort keys %testcases) {
63 # The complement of \w is \W; of [:posix:] is [:^posix:]
64 my $complement = $class;
65 if ($complement !~ s/ ( \[: ) /$1^/x) {
66 $complement = uc($class);
70 foreach my $ord (@{$testcases{$class}}) {
71 my $char = display(chr($ord));
73 # > 255 already implies upgraded. Skip the ones that don't
74 # have an explicit upgrade. This shows more clearly in the
75 # output which tests are in utf8, or not.
76 next if $ord > 255 && ! $upgrade;
78 my $reason = ""; # Explanation output with each test
80 my $match = 1; # Calculated whether test regex should
83 # Everything always matches in ASCII, or under /u
84 if ($ord < 128 || $charset eq 'u') {
85 $reason = "\"$char\" is a $class under /$charset";
86 $neg_reason = "\"$char\" is not a $complement under /$charset";
88 elsif ($charset eq "a" || $charset eq "aa") {
90 $reason = "\"$char\" is non-ASCII, which can't be a $class under /a";
91 $neg_reason = "\"$char\" is non-ASCII, which is a $complement under /a";
94 $reason = "\"$char\" is a $class under /$charset";
95 $neg_reason = "\"$char\" is not a $complement under /$charset";
97 elsif ($charset eq 'l') {
99 # We are using the C locale, which is essentially ASCII,
100 # but under utf8, the above-latin1 chars are treated as
102 $reason = "\"$char\" is not a $class in this locale under /l";
103 $neg_reason = "\"$char\" is a $complement in this locale under /l";
107 $reason = "\"$char\" is a $class in utf8 under /d";
108 $neg_reason = "\"$char\" is not a $complement in utf8 under /d";
111 $reason = "\"$char\" is above-ASCII latin1, which requires utf8 to be a $class under /d";
112 $neg_reason = "\"$char\" is above-ASCII latin1, which is a $complement under /d (unless in utf8)";
115 $reason = "; $reason" if $reason;
116 $neg_reason = "; $neg_reason" if $neg_reason;
130 foreach my $bracketed (0, 1) {
135 # Adds an extra char to the character class to make sure
136 # that the class doesn't get optimized away.
137 $lb = ($bracketed) ? '[_' : "";
138 $rb = ($bracketed) ? ']' : "";
140 else { # [:posix:] must be inside outer [ ]
141 next if $class =~ /\[/;
144 my $length = 10; # For regexec.c regrepeat() cases by
145 # matching more than one item
146 # Test both class and its complement, and with one or more
147 # than one item to match.
149 qq[my \$a = "$char"; $upgrade\$a $op qr/ (?$charset: $lb$class$rb ) /x],
150 qq[my \$a = "$char" x $length; $upgrade\$a $op qr/ (?$charset: $lb$class$rb\{$length} ) /x],
152 ok (eval $eval, $eval . $reason);
155 qq[my \$a = "$char"; $upgrade\$a $neg_op qr/ (?$charset: $lb$complement$rb ) /x],
156 qq[my \$a = "$char" x $length; $upgrade\$a $neg_op qr/ (?$charset: $lb$complement$rb\{$length} ) /x],
158 ok (eval $eval, $eval . $neg_reason);
162 next if $class ne '\w';
164 # Test \b, \B at beginning and end of string
166 qq[my \$a = "$char"; $upgrade\$a $op qr/ (?$charset: ^ \\b . ) /x],
167 qq[my \$a = "$char"; $upgrade\$a $op qr/ (?$charset: . \\b \$) /x],
169 ok (eval $eval, $eval . $reason);
172 qq[my \$a = "$char"; $upgrade\$a $neg_op qr/(?$charset: ^ \\B . ) /x],
173 qq[my \$a = "$char"; $upgrade\$a $neg_op qr/(?$charset: . \\B \$ ) /x],
175 ok (eval $eval, $eval . $neg_reason);
178 # Test \b, \B adjacent to a non-word char, both before it and
179 # after. We test with ASCII, Latin1 and Unicode non-word chars
180 foreach my $space_ord (@{$testcases{'\s'}}) {
182 # Useless to try to test non-utf8 when the ord itself
184 next if $space_ord > 255 && ! $upgrade;
186 my $space = display(chr $space_ord);
189 qq[my \$a = "$space$char"; $upgrade\$a $op qr/ (?$charset: . \\b . ) /x],
190 qq[my \$a = "$char$space"; $upgrade\$a $op qr/ (?$charset: . \\b . ) /x],
192 ok (eval $eval, $eval . $reason . "; \"$space\" is not a \\w");
195 qq[my \$a = "$space$char"; $upgrade\$a $neg_op qr/ (?$charset: . \\B . ) /x],
196 qq[my \$a = "$char$space"; $upgrade\$a $neg_op qr/ (?$charset: . \\B . ) /x],
198 ok (eval $eval, $eval . $neg_reason . "; \"$space\" is not a \\w");
202 # Test \b, \B in the middle of two nominally word chars, but
203 # one or both may be considered non-word depending on range
205 foreach my $other_ord (@{$testcases{'\w'}}) {
206 next if $other_ord > 255 && ! $upgrade;
207 my $other = display(chr $other_ord);
209 # Determine if the other char is a word char in current
211 my $other_is_word = 1;
212 my $other_reason = "\"$other\" is a $class under /$charset";
213 my $other_neg_reason = "\"$other\" is not a $complement under /$charset";
216 && (($charset eq "a" || $charset eq "aa")
217 || ($other_ord < 256 && ($charset eq 'l' || ! $upgrade))))
220 $other_reason = "\"$other\" is not a $class under /$charset";
221 $other_neg_reason = "\"$other\" is a $complement under /$charset";
223 my $both_reason = $reason;
224 $both_reason .= "; $other_reason" if $other_ord != $ord;
225 my $both_neg_reason = $neg_reason;
226 $both_neg_reason .= "; $other_neg_reason" if $other_ord != $ord;
228 # If both are the same wordness, then \b will fail; \B
230 if ($match == $other_is_word) {
240 qq[my \$a = "$other$char"; $upgrade\$a $op qr/ (?$charset: $other \\b $char ) /x],
241 qq[my \$a = "$char$other"; $upgrade\$a $op qr/ (?$charset: $char \\b $other ) /x],
243 ok (eval $eval, $eval . $both_reason);
246 qq[my \$a = "$other$char"; $upgrade\$a $neg_op qr/ (?$charset: $other \\B $char ) /x],
247 qq[my \$a = "$char$other"; $upgrade\$a $neg_op qr/ (?$charset: $char \\B $other ) /x],
249 ok (eval $eval, $eval . $both_neg_reason);
252 next if $other_ord == $ord;
254 # These start with the \b or \B. They are included, based
255 # on source code analysis, to force the testing of the FBC
256 # (find_by_class) portions of regexec.c.
258 qq[my \$a = "$other$char"; $upgrade\$a $op qr/ (?$charset: \\b $char ) /x],
259 qq[my \$a = "$char$other"; $upgrade\$a $op qr/ (?$charset: \\b $other ) /x],
261 ok (eval $eval, $eval . $both_reason);
264 qq[my \$a = "$other$char"; $upgrade\$a $neg_op qr/ (?$charset: \\B $char ) /x],
265 qq[my \$a = "$char$other"; $upgrade\$a $neg_op qr/ (?$charset: \\B $other ) /x],
267 ok (eval $eval, $eval . $both_neg_reason);
270 } # End of each test case in a class
271 } # End of \w, \s, ...
272 } # End of utf8 upgraded or not
275 plan(curr_test() - 1);