| Filename | /usr/share/perl/5.14/utf8_heavy.pl |
| Statements | Executed 3034 statements in 17.4ms |
| Calls | P | F | Exclusive Time |
Inclusive Time |
Subroutine |
|---|---|---|---|---|---|
| 2 | 2 | 2 | 14.9ms | 16.3ms | utf8::SWASHNEW |
| 2897 | 4 | 1 | 1.27ms | 1.27ms | utf8::CORE:match (opcode) |
| 2 | 1 | 1 | 109µs | 109µs | utf8::CORE:sort (opcode) |
| 1 | 1 | 1 | 15µs | 27µs | utf8::BEGIN@399 |
| 17 | 1 | 1 | 12µs | 12µs | utf8::is_utf8 (xsub) |
| 10 | 5 | 1 | 10µs | 10µs | utf8::CORE:subst (opcode) |
| 1 | 1 | 1 | 10µs | 14µs | utf8::BEGIN@2 |
| 1 | 1 | 1 | 10µs | 22µs | utf8::BEGIN@117 |
| 1 | 1 | 1 | 10µs | 22µs | utf8::BEGIN@460 |
| 1 | 1 | 1 | 8µs | 19µs | utf8::BEGIN@123 |
| 1 | 1 | 1 | 7µs | 13µs | utf8::BEGIN@3 |
| 1 | 1 | 1 | 5µs | 5µs | utf8::BEGIN@135 |
| 1 | 1 | 1 | 4µs | 4µs | utf8::BEGIN@147 |
| 0 | 0 | 0 | 0s | 0s | utf8::DESTROY |
| 0 | 0 | 0 | 0s | 0s | utf8::croak |
| Line | State ments |
Time on line |
Calls | Time in subs |
Code |
|---|---|---|---|---|---|
| 1 | package utf8; | ||||
| 2 | 2 | 25µs | 2 | 17µs | # spent 14µs (10+3) within utf8::BEGIN@2 which was called:
# once (10µs+3µs) by utf8::AUTOLOAD at line 2 # spent 14µs making 1 call to utf8::BEGIN@2
# spent 3µs making 1 call to strict::import |
| 3 | 2 | 458µs | 2 | 18µs | # spent 13µs (7+6) within utf8::BEGIN@3 which was called:
# once (7µs+6µs) by utf8::AUTOLOAD at line 3 # spent 13µs making 1 call to utf8::BEGIN@3
# spent 6µs making 1 call to warnings::import |
| 4 | |||||
| 5 | sub DEBUG () { 0 } | ||||
| 6 | |||||
| 7 | sub DESTROY {} | ||||
| 8 | |||||
| 9 | 1 | 300ns | my %Cache; | ||
| 10 | |||||
| 11 | sub croak { require Carp; Carp::croak(@_) } | ||||
| 12 | |||||
| 13 | ## | ||||
| 14 | ## "SWASH" == "SWATCH HASH". A "swatch" is a swatch of the Unicode landscape. | ||||
| 15 | ## It's a data structure that encodes a set of Unicode characters. | ||||
| 16 | ## | ||||
| 17 | |||||
| 18 | { | ||||
| 19 | # If a floating point number is within this distance from the value of a | ||||
| 20 | # fraction, it is considered to be that fraction, even if many more digits | ||||
| 21 | # are specified that don't exactly match. | ||||
| 22 | 2 | 400ns | my $min_floating_slop; | ||
| 23 | |||||
| 24 | # To guard against this program calling something that in turn ends up | ||||
| 25 | # calling this program with the same inputs, and hence infinitely | ||||
| 26 | # recursing, we keep a stack of the properties that are currently in | ||||
| 27 | # progress, pushed upon entry, popped upon return. | ||||
| 28 | 1 | 700ns | my @recursed; | ||
| 29 | |||||
| 30 | # spent 16.3ms (14.9+1.39) within utf8::SWASHNEW which was called 2 times, avg 8.14ms/call:
# once (12.4ms+754µs) by main::BEGIN@12 at line 18 of utf8.pm
# once (2.45ms+637µs) by main::BEGIN@12 at line 43 of NewBoard.pm | ||||
| 31 | 2 | 2µs | my ($class, $type, $list, $minbits, $none) = @_; | ||
| 32 | 2 | 1µs | local $^D = 0 if $^D; | ||
| 33 | |||||
| 34 | 2 | 500ns | $class = "" unless defined $class; | ||
| 35 | print STDERR __LINE__, ": class=$class, type=$type, list=", | ||||
| 36 | (defined $list) ? $list : ':undef:', | ||||
| 37 | ", minbits=$minbits, none=$none\n" if DEBUG; | ||||
| 38 | |||||
| 39 | ## | ||||
| 40 | ## Get the list of codepoints for the type. | ||||
| 41 | ## Called from swash_init (see utf8.c) or SWASHNEW itself. | ||||
| 42 | ## | ||||
| 43 | ## Callers of swash_init: | ||||
| 44 | ## op.c:pmtrans -- for tr/// and y/// | ||||
| 45 | ## regexec.c:regclass_swash -- for /[]/, \p, and \P | ||||
| 46 | ## utf8.c:is_utf8_common -- for common Unicode properties | ||||
| 47 | ## utf8.c:to_utf8_case -- for lc, uc, ucfirst, etc. and //i | ||||
| 48 | ## | ||||
| 49 | ## Given a $type, our goal is to fill $list with the set of codepoint | ||||
| 50 | ## ranges. If $type is false, $list passed is used. | ||||
| 51 | ## | ||||
| 52 | ## $minbits: | ||||
| 53 | ## For binary properties, $minbits must be 1. | ||||
| 54 | ## For character mappings (case and transliteration), $minbits must | ||||
| 55 | ## be a number except 1. | ||||
| 56 | ## | ||||
| 57 | ## $list (or that filled according to $type): | ||||
| 58 | ## Refer to perlunicode.pod, "User-Defined Character Properties." | ||||
| 59 | ## | ||||
| 60 | ## For binary properties, only characters with the property value | ||||
| 61 | ## of True should be listed. The 3rd column, if any, will be ignored | ||||
| 62 | ## | ||||
| 63 | ## $none is undocumented, so I'm (khw) trying to do some documentation | ||||
| 64 | ## of it now. It appears to be if there is a mapping in an input file | ||||
| 65 | ## that maps to 'XXXX', then that is replaced by $none+1, expressed in | ||||
| 66 | ## hexadecimal. The only place I found it possibly used was in | ||||
| 67 | ## S_pmtrans in op.c. | ||||
| 68 | ## | ||||
| 69 | ## To make the parsing of $type clear, this code takes the a rather | ||||
| 70 | ## unorthodox approach of last'ing out of the block once we have the | ||||
| 71 | ## info we need. Were this to be a subroutine, the 'last' would just | ||||
| 72 | ## be a 'return'. | ||||
| 73 | ## | ||||
| 74 | 2 | 200ns | my $file; ## file to load data from, and also part of the %Cache key. | ||
| 75 | 2 | 400ns | my $ListSorted = 0; | ||
| 76 | |||||
| 77 | # Change this to get a different set of Unicode tables | ||||
| 78 | 2 | 400ns | my $unicore_dir = 'unicore'; | ||
| 79 | |||||
| 80 | 2 | 500ns | if ($type) | ||
| 81 | { | ||||
| 82 | |||||
| 83 | # Verify that this isn't a recursive call for this property. | ||||
| 84 | # Can't use croak, as it may try to recurse here itself. | ||||
| 85 | 2 | 2µs | my $class_type = $class . "::$type"; | ||
| 86 | 2 | 1µs | if (grep { $_ eq $class_type } @recursed) { | ||
| 87 | CORE::die "panic: Infinite recursion in SWASHNEW for '$type'\n"; | ||||
| 88 | } | ||||
| 89 | 2 | 800ns | push @recursed, $class_type; | ||
| 90 | |||||
| 91 | 2 | 8µs | 2 | 3µs | $type =~ s/^\s+//; # spent 3µs making 2 calls to utf8::CORE:subst, avg 1µs/call |
| 92 | 2 | 7µs | 2 | 4µs | $type =~ s/\s+$//; # spent 4µs making 2 calls to utf8::CORE:subst, avg 2µs/call |
| 93 | |||||
| 94 | # regcomp.c surrounds the property name with '__" and '_i' if this | ||||
| 95 | # is to be caseless matching. | ||||
| 96 | 2 | 3µs | 2 | 900ns | my $caseless = $type =~ s/^__(.*)_i$/$1/; # spent 900ns making 2 calls to utf8::CORE:subst, avg 450ns/call |
| 97 | |||||
| 98 | print STDERR __LINE__, ": type=$type, caseless=$caseless\n" if DEBUG; | ||||
| 99 | |||||
| 100 | 2 | 5µs | 2 | 800ns | GETFILE: # spent 800ns making 2 calls to utf8::CORE:subst, avg 400ns/call |
| 101 | { | ||||
| 102 | ## | ||||
| 103 | ## It could be a user-defined property. Look in current | ||||
| 104 | ## package if no package given | ||||
| 105 | ## | ||||
| 106 | |||||
| 107 | 2 | 300ns | my $caller1 = $type =~ s/(.+)::// ? $1 : caller(1); | ||
| 108 | |||||
| 109 | 2 | 5µs | 2 | 1µs | if (defined $caller1 && $type =~ /^I[ns]\w+$/) { # spent 1µs making 2 calls to utf8::CORE:match, avg 550ns/call |
| 110 | my $prop = "${caller1}::$type"; | ||||
| 111 | if (exists &{$prop}) { | ||||
| 112 | # stolen from Scalar::Util::PP::tainted() | ||||
| 113 | my $tainted; | ||||
| 114 | { | ||||
| 115 | local($@, $SIG{__DIE__}, $SIG{__WARN__}); | ||||
| 116 | local $^W = 0; | ||||
| 117 | 2 | 101µs | 2 | 35µs | # spent 22µs (10+12) within utf8::BEGIN@117 which was called:
# once (10µs+12µs) by utf8::AUTOLOAD at line 117 # spent 22µs making 1 call to utf8::BEGIN@117
# spent 12µs making 1 call to warnings::unimport |
| 118 | eval { kill 0 * $prop }; | ||||
| 119 | $tainted = 1 if $@ =~ /^Insecure/; | ||||
| 120 | } | ||||
| 121 | die "Insecure user-defined property \\p{$prop}\n" | ||||
| 122 | if $tainted; | ||||
| 123 | 2 | 73µs | 2 | 31µs | # spent 19µs (8+12) within utf8::BEGIN@123 which was called:
# once (8µs+12µs) by utf8::AUTOLOAD at line 123 # spent 19µs making 1 call to utf8::BEGIN@123
# spent 12µs making 1 call to strict::unimport |
| 124 | $list = &{$prop}($caseless); | ||||
| 125 | last GETFILE; | ||||
| 126 | } | ||||
| 127 | } | ||||
| 128 | |||||
| 129 | # During Perl's compilation, this routine may be called before | ||||
| 130 | # the tables are constructed. If so, we have a chicken/egg | ||||
| 131 | # problem. If we die, the tables never get constructed, so | ||||
| 132 | # keep going, but return an empty table so only what the code | ||||
| 133 | # has compiled in internally (currently ASCII/Latin1 range | ||||
| 134 | # matching) will work. | ||||
| 135 | # spent 5µs within utf8::BEGIN@135 which was called:
# once (5µs+0s) by utf8::AUTOLOAD at line 139 | ||||
| 136 | # Poor man's constant, to avoid a run-time check. | ||||
| 137 | 1 | 5µs | $utf8::{miniperl} | ||
| 138 | = \! defined &DynaLoader::boot_DynaLoader; | ||||
| 139 | 1 | 71µs | 1 | 5µs | } # spent 5µs making 1 call to utf8::BEGIN@135 |
| 140 | 2 | 214µs | if (miniperl) { | ||
| 141 | eval "require '$unicore_dir/Heavy.pl'"; | ||||
| 142 | last GETFILE if $@; | ||||
| 143 | } | ||||
| 144 | else { | ||||
| 145 | 2 | 7.06ms | require "$unicore_dir/Heavy.pl"; | ||
| 146 | } | ||||
| 147 | 1 | 1.27ms | 1 | 4µs | # spent 4µs within utf8::BEGIN@147 which was called:
# once (4µs+0s) by utf8::AUTOLOAD at line 147 # spent 4µs making 1 call to utf8::BEGIN@147 |
| 148 | |||||
| 149 | # All property names are matched caselessly | ||||
| 150 | 2 | 4µs | my $property_and_table = CORE::lc $type; | ||
| 151 | print STDERR __LINE__, ": $property_and_table\n" if DEBUG; | ||||
| 152 | |||||
| 153 | # See if is of the compound form 'property=value', where the | ||||
| 154 | # value indicates the table we should use. | ||||
| 155 | 2 | 5µs | my ($property, $table, @remainder) = | ||
| 156 | split /\s*[:=]\s*/, $property_and_table, -1; | ||||
| 157 | 2 | 600ns | if (@remainder) { | ||
| 158 | pop @recursed if @recursed; | ||||
| 159 | return $type; | ||||
| 160 | } | ||||
| 161 | |||||
| 162 | 2 | 100ns | my $prefix; | ||
| 163 | 2 | 1µs | if (! defined $table) { | ||
| 164 | |||||
| 165 | # Here, is the single form. The property becomes empty, and | ||||
| 166 | # the whole value is the table. | ||||
| 167 | 2 | 500ns | $table = $property; | ||
| 168 | 2 | 800ns | $prefix = $property = ""; | ||
| 169 | } else { | ||||
| 170 | print STDERR __LINE__, ": $property\n" if DEBUG; | ||||
| 171 | |||||
| 172 | # Here it is the compound property=table form. The property | ||||
| 173 | # name is always loosely matched, which means remove any of | ||||
| 174 | # these: | ||||
| 175 | $property =~ s/[_\s-]//g; | ||||
| 176 | |||||
| 177 | # And convert to canonical form. Quit if not valid. | ||||
| 178 | $property = $utf8::loose_property_name_of{$property}; | ||||
| 179 | if (! defined $property) { | ||||
| 180 | pop @recursed if @recursed; | ||||
| 181 | return $type; | ||||
| 182 | } | ||||
| 183 | |||||
| 184 | $prefix = "$property="; | ||||
| 185 | |||||
| 186 | # If the rhs looks like it is a number... | ||||
| 187 | print STDERR __LINE__, ": table=$table\n" if DEBUG; | ||||
| 188 | if ($table =~ qr{ ^ [ \s 0-9 _ + / . -]+ $ }x) { | ||||
| 189 | print STDERR __LINE__, ": table=$table\n" if DEBUG; | ||||
| 190 | |||||
| 191 | # Don't allow leading nor trailing slashes | ||||
| 192 | if ($table =~ / ^ \/ | \/ $ /x) { | ||||
| 193 | pop @recursed if @recursed; | ||||
| 194 | return $type; | ||||
| 195 | } | ||||
| 196 | |||||
| 197 | # Split on slash, in case it is a rational, like \p{1/5} | ||||
| 198 | my @parts = split qr{ \s* / \s* }x, $table, -1; | ||||
| 199 | print __LINE__, ": $type\n" if @parts > 2 && DEBUG; | ||||
| 200 | |||||
| 201 | # Can have maximum of one slash | ||||
| 202 | if (@parts > 2) { | ||||
| 203 | pop @recursed if @recursed; | ||||
| 204 | return $type; | ||||
| 205 | } | ||||
| 206 | |||||
| 207 | foreach my $part (@parts) { | ||||
| 208 | print __LINE__, ": part=$part\n" if DEBUG; | ||||
| 209 | |||||
| 210 | $part =~ s/^\+\s*//; # Remove leading plus | ||||
| 211 | $part =~ s/^-\s*/-/; # Remove blanks after unary | ||||
| 212 | # minus | ||||
| 213 | |||||
| 214 | # Remove underscores between digits. | ||||
| 215 | $part =~ s/( ?<= [0-9] ) _ (?= [0-9] ) //xg; | ||||
| 216 | |||||
| 217 | # No leading zeros (but don't make a single '0' | ||||
| 218 | # into a null string) | ||||
| 219 | $part =~ s/ ^ ( -? ) 0+ /$1/x; | ||||
| 220 | $part .= '0' if $part eq '-' || $part eq ""; | ||||
| 221 | |||||
| 222 | # No trailing zeros after a decimal point | ||||
| 223 | $part =~ s/ ( \. .*? ) 0+ $ /$1/x; | ||||
| 224 | |||||
| 225 | # Begin with a 0 if a leading decimal point | ||||
| 226 | $part =~ s/ ^ ( -? ) \. /${1}0./x; | ||||
| 227 | |||||
| 228 | # Ensure not a trailing decimal point: turn into an | ||||
| 229 | # integer | ||||
| 230 | $part =~ s/ \. $ //x; | ||||
| 231 | |||||
| 232 | print STDERR __LINE__, ": part=$part\n" if DEBUG; | ||||
| 233 | #return $type if $part eq ""; | ||||
| 234 | |||||
| 235 | # Result better look like a number. (This test is | ||||
| 236 | # needed because, for example could have a plus in | ||||
| 237 | # the middle.) | ||||
| 238 | if ($part !~ / ^ -? [0-9]+ ( \. [0-9]+)? $ /x) { | ||||
| 239 | pop @recursed if @recursed; | ||||
| 240 | return $type; | ||||
| 241 | } | ||||
| 242 | } | ||||
| 243 | |||||
| 244 | # If a rational... | ||||
| 245 | if (@parts == 2) { | ||||
| 246 | |||||
| 247 | # If denominator is negative, get rid of it, and ... | ||||
| 248 | if ($parts[1] =~ s/^-//) { | ||||
| 249 | |||||
| 250 | # If numerator is also negative, convert the | ||||
| 251 | # whole thing to positive, or move the minus to | ||||
| 252 | # the numerator | ||||
| 253 | if ($parts[0] !~ s/^-//) { | ||||
| 254 | $parts[0] = '-' . $parts[0]; | ||||
| 255 | } | ||||
| 256 | } | ||||
| 257 | $table = join '/', @parts; | ||||
| 258 | } | ||||
| 259 | elsif ($property ne 'nv' || $parts[0] !~ /\./) { | ||||
| 260 | |||||
| 261 | # Here is not numeric value, or doesn't have a | ||||
| 262 | # decimal point. No further manipulation is | ||||
| 263 | # necessary. (Note the hard-coded property name. | ||||
| 264 | # This could fail if other properties eventually | ||||
| 265 | # had fractions as well; perhaps the cjk ones | ||||
| 266 | # could evolve to do that. This hard-coding could | ||||
| 267 | # be fixed by mktables generating a list of | ||||
| 268 | # properties that could have fractions.) | ||||
| 269 | $table = $parts[0]; | ||||
| 270 | } else { | ||||
| 271 | |||||
| 272 | # Here is a floating point numeric_value. Try to | ||||
| 273 | # convert to rational. First see if is in the list | ||||
| 274 | # of known ones. | ||||
| 275 | if (exists $utf8::nv_floating_to_rational{$parts[0]}) { | ||||
| 276 | $table = $utf8::nv_floating_to_rational{$parts[0]}; | ||||
| 277 | } else { | ||||
| 278 | |||||
| 279 | # Here not in the list. See if is close | ||||
| 280 | # enough to something in the list. First | ||||
| 281 | # determine what 'close enough' means. It has | ||||
| 282 | # to be as tight as what mktables says is the | ||||
| 283 | # maximum slop, and as tight as how many | ||||
| 284 | # digits we were passed. That is, if the user | ||||
| 285 | # said .667, .6667, .66667, etc. we match as | ||||
| 286 | # many digits as they passed until get to | ||||
| 287 | # where it doesn't matter any more due to the | ||||
| 288 | # machine's precision. If they said .6666668, | ||||
| 289 | # we fail. | ||||
| 290 | (my $fraction = $parts[0]) =~ s/^.*\.//; | ||||
| 291 | my $epsilon = 10 ** - (length($fraction)); | ||||
| 292 | if ($epsilon > $utf8::max_floating_slop) { | ||||
| 293 | $epsilon = $utf8::max_floating_slop; | ||||
| 294 | } | ||||
| 295 | |||||
| 296 | # But it can't be tighter than the minimum | ||||
| 297 | # precision for this machine. If haven't | ||||
| 298 | # already calculated that minimum, do so now. | ||||
| 299 | if (! defined $min_floating_slop) { | ||||
| 300 | |||||
| 301 | # Keep going down an order of magnitude | ||||
| 302 | # until find that adding this quantity to | ||||
| 303 | # 1 remains 1; but put an upper limit on | ||||
| 304 | # this so in case this algorithm doesn't | ||||
| 305 | # work properly on some platform, that we | ||||
| 306 | # won't loop forever. | ||||
| 307 | my $count = 0; | ||||
| 308 | $min_floating_slop = 1; | ||||
| 309 | while (1+ $min_floating_slop != 1 | ||||
| 310 | && $count++ < 50) | ||||
| 311 | { | ||||
| 312 | my $next = $min_floating_slop / 10; | ||||
| 313 | last if $next == 0; # If underflows, | ||||
| 314 | # use previous one | ||||
| 315 | $min_floating_slop = $next; | ||||
| 316 | print STDERR __LINE__, ": min_float_slop=$min_floating_slop\n" if DEBUG; | ||||
| 317 | } | ||||
| 318 | |||||
| 319 | # Back off a couple orders of magnitude, | ||||
| 320 | # just to be safe. | ||||
| 321 | $min_floating_slop *= 100; | ||||
| 322 | } | ||||
| 323 | |||||
| 324 | if ($epsilon < $min_floating_slop) { | ||||
| 325 | $epsilon = $min_floating_slop; | ||||
| 326 | } | ||||
| 327 | print STDERR __LINE__, ": fraction=.$fraction; epsilon=$epsilon\n" if DEBUG; | ||||
| 328 | |||||
| 329 | undef $table; | ||||
| 330 | |||||
| 331 | # And for each possible rational in the table, | ||||
| 332 | # see if it is within epsilon of the input. | ||||
| 333 | foreach my $official | ||||
| 334 | (keys %utf8::nv_floating_to_rational) | ||||
| 335 | { | ||||
| 336 | print STDERR __LINE__, ": epsilon=$epsilon, official=$official, diff=", abs($parts[0] - $official), "\n" if DEBUG; | ||||
| 337 | if (abs($parts[0] - $official) < $epsilon) { | ||||
| 338 | $table = | ||||
| 339 | $utf8::nv_floating_to_rational{$official}; | ||||
| 340 | last; | ||||
| 341 | } | ||||
| 342 | } | ||||
| 343 | |||||
| 344 | # Quit if didn't find one. | ||||
| 345 | if (! defined $table) { | ||||
| 346 | pop @recursed if @recursed; | ||||
| 347 | return $type; | ||||
| 348 | } | ||||
| 349 | } | ||||
| 350 | } | ||||
| 351 | print STDERR __LINE__, ": $property=$table\n" if DEBUG; | ||||
| 352 | } | ||||
| 353 | } | ||||
| 354 | |||||
| 355 | # Combine lhs (if any) and rhs to get something that matches | ||||
| 356 | # the syntax of the lookups. | ||||
| 357 | 2 | 600ns | $property_and_table = "$prefix$table"; | ||
| 358 | print STDERR __LINE__, ": $property_and_table\n" if DEBUG; | ||||
| 359 | |||||
| 360 | # First try stricter matching. | ||||
| 361 | 2 | 1µs | $file = $utf8::stricter_to_file_of{$property_and_table}; | ||
| 362 | |||||
| 363 | # If didn't find it, try again with looser matching by editing | ||||
| 364 | # out the applicable characters on the rhs and looking up | ||||
| 365 | # again. | ||||
| 366 | 2 | 600ns | if (! defined $file) { | ||
| 367 | 2 | 8µs | 2 | 1µs | $table =~ s/ [_\s-] //xg; # spent 1µs making 2 calls to utf8::CORE:subst, avg 650ns/call |
| 368 | 2 | 500ns | $property_and_table = "$prefix$table"; | ||
| 369 | print STDERR __LINE__, ": $property_and_table\n" if DEBUG; | ||||
| 370 | 2 | 1µs | $file = $utf8::loose_to_file_of{$property_and_table}; | ||
| 371 | } | ||||
| 372 | |||||
| 373 | # Add the constant and go fetch it in. | ||||
| 374 | 2 | 400ns | if (defined $file) { | ||
| 375 | 2 | 600ns | if ($utf8::why_deprecated{$file}) { | ||
| 376 | warnings::warnif('deprecated', "Use of '$type' in \\p{} or \\P{} is deprecated because: $utf8::why_deprecated{$file};"); | ||||
| 377 | } | ||||
| 378 | |||||
| 379 | 2 | 300ns | if ($caseless | ||
| 380 | && exists $utf8::caseless_equivalent{$property_and_table}) | ||||
| 381 | { | ||||
| 382 | $file = $utf8::caseless_equivalent{$property_and_table}; | ||||
| 383 | } | ||||
| 384 | 2 | 2µs | $file= "$unicore_dir/lib/$file.pl"; | ||
| 385 | 2 | 2µs | last GETFILE; | ||
| 386 | } | ||||
| 387 | print STDERR __LINE__, ": didn't find $property_and_table\n" if DEBUG; | ||||
| 388 | |||||
| 389 | ## | ||||
| 390 | ## See if it's a user-level "To". | ||||
| 391 | ## | ||||
| 392 | |||||
| 393 | my $caller0 = caller(0); | ||||
| 394 | |||||
| 395 | if (defined $caller0 && $type =~ /^To(?:\w+)$/) { | ||||
| 396 | my $map = $caller0 . "::" . $type; | ||||
| 397 | |||||
| 398 | if (exists &{$map}) { | ||||
| 399 | 2 | 332µs | 2 | 39µs | # spent 27µs (15+12) within utf8::BEGIN@399 which was called:
# once (15µs+12µs) by utf8::AUTOLOAD at line 399 # spent 27µs making 1 call to utf8::BEGIN@399
# spent 12µs making 1 call to strict::unimport |
| 400 | |||||
| 401 | $list = &{$map}; | ||||
| 402 | warnings::warnif('deprecated', "User-defined case-mapping '$type' is deprecated"); | ||||
| 403 | last GETFILE; | ||||
| 404 | } | ||||
| 405 | } | ||||
| 406 | |||||
| 407 | ## | ||||
| 408 | ## Last attempt -- see if it's a standard "To" name | ||||
| 409 | ## (e.g. "ToLower") ToTitle is used by ucfirst(). | ||||
| 410 | ## The user-level way to access ToDigit() and ToFold() | ||||
| 411 | ## is to use Unicode::UCD. | ||||
| 412 | ## | ||||
| 413 | if ($type =~ /^To(Digit|Fold|Lower|Title|Upper)$/) { | ||||
| 414 | $file = "$unicore_dir/To/$1.pl"; | ||||
| 415 | ## would like to test to see if $file actually exists.... | ||||
| 416 | last GETFILE; | ||||
| 417 | } | ||||
| 418 | |||||
| 419 | ## | ||||
| 420 | ## If we reach this line, it's because we couldn't figure | ||||
| 421 | ## out what to do with $type. Ouch. | ||||
| 422 | ## | ||||
| 423 | |||||
| 424 | pop @recursed if @recursed; | ||||
| 425 | return $type; | ||||
| 426 | } | ||||
| 427 | |||||
| 428 | 2 | 500ns | if (defined $file) { | ||
| 429 | print STDERR __LINE__, ": found it (file='$file')\n" if DEBUG; | ||||
| 430 | |||||
| 431 | ## | ||||
| 432 | ## If we reach here, it was due to a 'last GETFILE' above | ||||
| 433 | ## (exception: user-defined properties and mappings), so we | ||||
| 434 | ## have a filename, so now we load it if we haven't already. | ||||
| 435 | ## If we have, return the cached results. The cache key is the | ||||
| 436 | ## class and file to load. | ||||
| 437 | ## | ||||
| 438 | 2 | 2µs | my $found = $Cache{$class, $file}; | ||
| 439 | 2 | 300ns | if ($found and ref($found) eq $class) { | ||
| 440 | print STDERR __LINE__, ": Returning cached '$file' for \\p{$type}\n" if DEBUG; | ||||
| 441 | pop @recursed if @recursed; | ||||
| 442 | return $found; | ||||
| 443 | } | ||||
| 444 | |||||
| 445 | 2 | 400ns | local $@; | ||
| 446 | 2 | 3µs | local $!; | ||
| 447 | 4 | 537µs | $list = do $file; die $@ if $@; | ||
| 448 | } | ||||
| 449 | |||||
| 450 | 2 | 1µs | $ListSorted = 1; ## we know that these lists are sorted | ||
| 451 | } | ||||
| 452 | |||||
| 453 | 2 | 300ns | my $extras; | ||
| 454 | 2 | 300ns | my $bits = $minbits; | ||
| 455 | |||||
| 456 | 2 | 79µs | if ($list) { | ||
| 457 | 2 | 1µs | my $taint = substr($list,0,0); # maintain taint | ||
| 458 | 2 | 160µs | my @tmp = split(/^/m, $list); | ||
| 459 | 2 | 400ns | my %seen; | ||
| 460 | 2 | 789µs | 2 | 34µs | # spent 22µs (10+12) within utf8::BEGIN@460 which was called:
# once (10µs+12µs) by utf8::AUTOLOAD at line 460 # spent 22µs making 1 call to utf8::BEGIN@460
# spent 12µs making 1 call to warnings::unimport |
| 461 | 2 | 820µs | 965 | 184µs | $extras = join '', $taint, grep /^[^0-9a-fA-F]/, @tmp; # spent 184µs making 965 calls to utf8::CORE:match, avg 190ns/call |
| 462 | $list = join '', $taint, | ||||
| 463 | map { $_->[1] } | ||||
| 464 | 965 | 1.39ms | 965 | 531µs | sort { $a->[0] <=> $b->[0] } # spent 531µs making 965 calls to utf8::CORE:match, avg 550ns/call |
| 465 | 1930 | 3.08ms | 965 | 556µs | map { /^([0-9a-fA-F]+)/; [ CORE::hex($1), $_ ] } # spent 556µs making 965 calls to utf8::CORE:match, avg 576ns/call |
| 466 | 2 | 806µs | 2 | 109µs | grep { /^([0-9a-fA-F]+)/ and not $seen{$1}++ } @tmp; # XXX doesn't do ranges right # spent 109µs making 2 calls to utf8::CORE:sort, avg 54µs/call |
| 467 | } | ||||
| 468 | |||||
| 469 | 2 | 200ns | if ($none) { | ||
| 470 | my $hextra = sprintf "%04x", $none + 1; | ||||
| 471 | $list =~ s/\tXXXX$/\t$hextra/mg; | ||||
| 472 | } | ||||
| 473 | |||||
| 474 | 2 | 500ns | if ($minbits != 1 && $minbits < 32) { # not binary property | ||
| 475 | my $top = 0; | ||||
| 476 | while ($list =~ /^([0-9a-fA-F]+)(?:[\t]([0-9a-fA-F]+)?)(?:[ \t]([0-9a-fA-F]+))?/mg) { | ||||
| 477 | my $min = CORE::hex $1; | ||||
| 478 | my $max = defined $2 ? CORE::hex $2 : $min; | ||||
| 479 | my $val = defined $3 ? CORE::hex $3 : 0; | ||||
| 480 | $val += $max - $min if defined $3; | ||||
| 481 | $top = $val if $val > $top; | ||||
| 482 | } | ||||
| 483 | my $topbits = | ||||
| 484 | $top > 0xffff ? 32 : | ||||
| 485 | $top > 0xff ? 16 : 8; | ||||
| 486 | $bits = $topbits if $bits < $topbits; | ||||
| 487 | } | ||||
| 488 | |||||
| 489 | 2 | 300ns | my @extras; | ||
| 490 | 2 | 200ns | if ($extras) { | ||
| 491 | for my $x ($extras) { | ||||
| 492 | my $taint = substr($x,0,0); # maintain taint | ||||
| 493 | pos $x = 0; | ||||
| 494 | while ($x =~ /^([^0-9a-fA-F\n])(.*)/mg) { | ||||
| 495 | my $char = "$1$taint"; | ||||
| 496 | my $name = "$2$taint"; | ||||
| 497 | print STDERR __LINE__, ": char [$char] => name [$name]\n" | ||||
| 498 | if DEBUG; | ||||
| 499 | if ($char =~ /[-+!&]/) { | ||||
| 500 | my ($c,$t) = split(/::/, $name, 2); # bogus use of ::, really | ||||
| 501 | my $subobj; | ||||
| 502 | if ($c eq 'utf8') { | ||||
| 503 | $subobj = utf8->SWASHNEW($t, "", $minbits, 0); | ||||
| 504 | } | ||||
| 505 | elsif (exists &$name) { | ||||
| 506 | $subobj = utf8->SWASHNEW($name, "", $minbits, 0); | ||||
| 507 | } | ||||
| 508 | elsif ($c =~ /^([0-9a-fA-F]+)/) { | ||||
| 509 | $subobj = utf8->SWASHNEW("", $c, $minbits, 0); | ||||
| 510 | } | ||||
| 511 | if (! ref $subobj) { | ||||
| 512 | pop @recursed if @recursed && $type; | ||||
| 513 | return $subobj; | ||||
| 514 | } | ||||
| 515 | push @extras, $name => $subobj; | ||||
| 516 | $bits = $subobj->{BITS} if $bits < $subobj->{BITS}; | ||||
| 517 | } | ||||
| 518 | } | ||||
| 519 | } | ||||
| 520 | } | ||||
| 521 | |||||
| 522 | if (DEBUG) { | ||||
| 523 | print STDERR __LINE__, ": CLASS = $class, TYPE => $type, BITS => $bits, NONE => $none"; | ||||
| 524 | print STDERR "\nLIST =>\n$list" if defined $list; | ||||
| 525 | print STDERR "\nEXTRAS =>\n$extras" if defined $extras; | ||||
| 526 | print STDERR "\n"; | ||||
| 527 | } | ||||
| 528 | |||||
| 529 | 2 | 50µs | my $SWASH = bless { | ||
| 530 | TYPE => $type, | ||||
| 531 | BITS => $bits, | ||||
| 532 | EXTRAS => $extras, | ||||
| 533 | LIST => $list, | ||||
| 534 | NONE => $none, | ||||
| 535 | @extras, | ||||
| 536 | } => $class; | ||||
| 537 | |||||
| 538 | 2 | 2µs | if ($file) { | ||
| 539 | $Cache{$class, $file} = $SWASH; | ||||
| 540 | } | ||||
| 541 | |||||
| 542 | 2 | 2µs | pop @recursed if @recursed && $type; | ||
| 543 | |||||
| 544 | 2 | 6µs | return $SWASH; | ||
| 545 | } | ||||
| 546 | } | ||||
| 547 | |||||
| 548 | # Now SWASHGET is recasted into a C function S_swash_get (see utf8.c). | ||||
| 549 | |||||
| 550 | 1 | 4µs | 1; | ||
# spent 1.27ms within utf8::CORE:match which was called 2897 times, avg 439ns/call:
# 965 times (556µs+0s) by utf8::SWASHNEW at line 465, avg 576ns/call
# 965 times (531µs+0s) by utf8::SWASHNEW at line 464, avg 550ns/call
# 965 times (184µs+0s) by utf8::SWASHNEW at line 461, avg 190ns/call
# 2 times (1µs+0s) by utf8::SWASHNEW at line 109, avg 550ns/call | |||||
# spent 109µs within utf8::CORE:sort which was called 2 times, avg 54µs/call:
# 2 times (109µs+0s) by utf8::SWASHNEW at line 466, avg 54µs/call | |||||
# spent 10µs within utf8::CORE:subst which was called 10 times, avg 1µs/call:
# 2 times (4µs+0s) by utf8::SWASHNEW at line 92, avg 2µs/call
# 2 times (3µs+0s) by utf8::SWASHNEW at line 91, avg 1µs/call
# 2 times (1µs+0s) by utf8::SWASHNEW at line 367, avg 650ns/call
# 2 times (900ns+0s) by utf8::SWASHNEW at line 96, avg 450ns/call
# 2 times (800ns+0s) by utf8::SWASHNEW at line 100, avg 400ns/call | |||||
# spent 12µs within utf8::is_utf8 which was called 17 times, avg 700ns/call:
# 17 times (12µs+0s) by constant::import at line 123 of constant.pm, avg 700ns/call |