| Filename | /usr/lib64/perl5/5.16.0/utf8_heavy.pl |
| Statements | Executed 298 statements in 72.9ms |
| Calls | P | F | Exclusive Time |
Inclusive Time |
Subroutine |
|---|---|---|---|---|---|
| 5 | 5 | 4 | 64.6ms | 64.9ms | utf8::SWASHNEW (recurses: max depth 1, inclusive time 368µs) |
| 24 | 7 | 1 | 140µs | 140µs | utf8::CORE:subst (opcode) |
| 3 | 1 | 1 | 129µs | 147µs | utf8::_loose_name |
| 1 | 1 | 1 | 95µs | 174µs | utf8::BEGIN@147 |
| 1 | 1 | 1 | 85µs | 171µs | utf8::BEGIN@542 |
| 1 | 1 | 1 | 69µs | 103µs | utf8::BEGIN@3 |
| 1 | 1 | 1 | 66µs | 166µs | utf8::BEGIN@2 |
| 12 | 4 | 1 | 65µs | 65µs | utf8::CORE:match (opcode) |
| 1 | 1 | 1 | 63µs | 156µs | utf8::BEGIN@654 |
| 1 | 1 | 1 | 53µs | 131µs | utf8::BEGIN@153 |
| 1 | 1 | 1 | 45µs | 45µs | utf8::BEGIN@178 |
| 1 | 1 | 1 | 41µs | 41µs | utf8::BEGIN@166 |
| 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 | 144µs | 2 | 266µs | # spent 166µs (66+100) within utf8::BEGIN@2 which was called:
# once (66µs+100µs) by utf8::AUTOLOAD at line 2 # spent 166µs making 1 call to utf8::BEGIN@2
# spent 100µs making 1 call to strict::import |
| 3 | 2 | 2.67ms | 2 | 137µs | # spent 103µs (69+34) within utf8::BEGIN@3 which was called:
# once (69µs+34µs) by utf8::AUTOLOAD at line 3 # spent 103µs making 1 call to utf8::BEGIN@3
# spent 34µs making 1 call to warnings::import |
| 4 | |||||
| 5 | sub DEBUG () { 0 } | ||||
| 6 | $|=1 if DEBUG; | ||||
| 7 | |||||
| 8 | sub DESTROY {} | ||||
| 9 | |||||
| 10 | 1 | 1µs | my %Cache; | ||
| 11 | |||||
| 12 | sub croak { require Carp; Carp::croak(@_) } | ||||
| 13 | |||||
| 14 | # spent 147µs (129+18) within utf8::_loose_name which was called 3 times, avg 49µs/call:
# 3 times (129µs+18µs) by utf8::SWASHNEW at line 399, avg 49µs/call | ||||
| 15 | # Given a lowercase property or property-value name, return its | ||||
| 16 | # standardized version that is expected for look-up in the 'loose' hashes | ||||
| 17 | # in Heavy.pl (hence, this depends on what mktables does). This squeezes | ||||
| 18 | # out blanks, underscores and dashes. The complication stems from the | ||||
| 19 | # grandfathered-in 'L_', which retains a single trailing underscore. | ||||
| 20 | |||||
| 21 | 3 | 64µs | 3 | 10µs | my $loose = $_[0] =~ s/[-\s_]//rg; # spent 10µs making 3 calls to utf8::CORE:subst, avg 3µs/call |
| 22 | |||||
| 23 | 3 | 101µs | 3 | 8µs | return $loose if $loose !~ / ^ (?: is | to )? l $/x; # spent 8µs making 3 calls to utf8::CORE:match, avg 3µs/call |
| 24 | return 'l_' if $_[0] =~ / l .* _ /x; # If original had a trailing '_' | ||||
| 25 | return $loose; | ||||
| 26 | } | ||||
| 27 | |||||
| 28 | ## | ||||
| 29 | ## "SWASH" == "SWATCH HASH". A "swatch" is a swatch of the Unicode landscape. | ||||
| 30 | ## It's a data structure that encodes a set of Unicode characters. | ||||
| 31 | ## | ||||
| 32 | |||||
| 33 | { | ||||
| 34 | # If a floating point number is within this distance from the value of a | ||||
| 35 | # fraction, it is considered to be that fraction, even if many more digits | ||||
| 36 | # are specified that don't exactly match. | ||||
| 37 | 2 | 4µs | my $min_floating_slop; | ||
| 38 | |||||
| 39 | # To guard against this program calling something that in turn ends up | ||||
| 40 | # calling this program with the same inputs, and hence infinitely | ||||
| 41 | # recursing, we keep a stack of the properties that are currently in | ||||
| 42 | # progress, pushed upon entry, popped upon return. | ||||
| 43 | 1 | 5µs | my @recursed; | ||
| 44 | |||||
| 45 | # spent 64.9ms (64.6+334µs) within utf8::SWASHNEW which was called 5 times, avg 13.0ms/call:
# once (61.4ms+46µs) by constant::BEGIN@25 at line 18 of utf8.pm
# once (1.77ms+73µs) by File::MimeInfo::CORE:match at line 164 of File/MimeInfo.pm
# once (746µs+90µs) by XML::Simple::CORE:match at line 1042 of XML/Simple.pm
# once (336µs+426µs) by File::MimeInfo::CORE:subst at line 178 of File/MimeInfo.pm
# once (300µs+-300µs) by utf8::SWASHNEW at line 607 | ||||
| 46 | 5 | 34µs | my ($class, $type, $list, $minbits, $none) = @_; | ||
| 47 | 5 | 7µs | my $user_defined = 0; | ||
| 48 | 5 | 19µs | local $^D = 0 if $^D; | ||
| 49 | |||||
| 50 | 5 | 5µs | $class = "" unless defined $class; | ||
| 51 | print STDERR __LINE__, ": class=$class, type=$type, list=", | ||||
| 52 | (defined $list) ? $list : ':undef:', | ||||
| 53 | ", minbits=$minbits, none=$none\n" if DEBUG; | ||||
| 54 | |||||
| 55 | ## | ||||
| 56 | ## Get the list of codepoints for the type. | ||||
| 57 | ## Called from swash_init (see utf8.c) or SWASHNEW itself. | ||||
| 58 | ## | ||||
| 59 | ## Callers of swash_init: | ||||
| 60 | ## op.c:pmtrans -- for tr/// and y/// | ||||
| 61 | ## regexec.c:regclass_swash -- for /[]/, \p, and \P | ||||
| 62 | ## utf8.c:is_utf8_common -- for common Unicode properties | ||||
| 63 | ## utf8.c:to_utf8_case -- for lc, uc, ucfirst, etc. and //i | ||||
| 64 | ## Unicode::UCD::prop_invlist | ||||
| 65 | ## Unicode::UCD::prop_invmap | ||||
| 66 | ## | ||||
| 67 | ## Given a $type, our goal is to fill $list with the set of codepoint | ||||
| 68 | ## ranges. If $type is false, $list passed is used. | ||||
| 69 | ## | ||||
| 70 | ## $minbits: | ||||
| 71 | ## For binary properties, $minbits must be 1. | ||||
| 72 | ## For character mappings (case and transliteration), $minbits must | ||||
| 73 | ## be a number except 1. | ||||
| 74 | ## | ||||
| 75 | ## $list (or that filled according to $type): | ||||
| 76 | ## Refer to perlunicode.pod, "User-Defined Character Properties." | ||||
| 77 | ## | ||||
| 78 | ## For binary properties, only characters with the property value | ||||
| 79 | ## of True should be listed. The 3rd column, if any, will be ignored | ||||
| 80 | ## | ||||
| 81 | ## $none is undocumented, so I'm (khw) trying to do some documentation | ||||
| 82 | ## of it now. It appears to be if there is a mapping in an input file | ||||
| 83 | ## that maps to 'XXXX', then that is replaced by $none+1, expressed in | ||||
| 84 | ## hexadecimal. It is used somehow in tr///. | ||||
| 85 | ## | ||||
| 86 | ## To make the parsing of $type clear, this code takes the a rather | ||||
| 87 | ## unorthodox approach of last'ing out of the block once we have the | ||||
| 88 | ## info we need. Were this to be a subroutine, the 'last' would just | ||||
| 89 | ## be a 'return'. | ||||
| 90 | ## | ||||
| 91 | # If a problem is found $type is returned; | ||||
| 92 | # Upon success, a new (or cached) blessed object is returned with | ||||
| 93 | # keys TYPE, BITS, EXTRAS, LIST, and NONE with values having the | ||||
| 94 | # same meanings as the input parameters. | ||||
| 95 | # SPECIALS contains a reference to any special-treatment hash in the | ||||
| 96 | # INVERT_IT is non-zero if the result should be inverted before use | ||||
| 97 | # USER_DEFINED is non-zero if the result came from a user-defined | ||||
| 98 | # property. | ||||
| 99 | 5 | 5µs | my $file; ## file to load data from, and also part of the %Cache key. | ||
| 100 | |||||
| 101 | # Change this to get a different set of Unicode tables | ||||
| 102 | 5 | 10µs | my $unicore_dir = 'unicore'; | ||
| 103 | 5 | 4µs | my $invert_it = 0; | ||
| 104 | 5 | 6µs | my $list_is_from_mktables = 0; # Is $list returned from a mktables | ||
| 105 | # generated file? If so, we know it's | ||||
| 106 | # well behaved. | ||||
| 107 | |||||
| 108 | 5 | 8µs | if ($type) | ||
| 109 | { | ||||
| 110 | # Verify that this isn't a recursive call for this property. | ||||
| 111 | # Can't use croak, as it may try to recurse to here itself. | ||||
| 112 | 4 | 21µs | my $class_type = $class . "::$type"; | ||
| 113 | 4 | 17µs | if (grep { $_ eq $class_type } @recursed) { | ||
| 114 | CORE::die "panic: Infinite recursion in SWASHNEW for '$type'\n"; | ||||
| 115 | } | ||||
| 116 | 4 | 14µs | push @recursed, $class_type; | ||
| 117 | |||||
| 118 | 4 | 111µs | 4 | 36µs | $type =~ s/^\s+//; # spent 36µs making 4 calls to utf8::CORE:subst, avg 9µs/call |
| 119 | 4 | 72µs | 4 | 25µs | $type =~ s/\s+$//; # spent 25µs making 4 calls to utf8::CORE:subst, avg 6µs/call |
| 120 | |||||
| 121 | # regcomp.c surrounds the property name with '__" and '_i' if this | ||||
| 122 | # is to be caseless matching. | ||||
| 123 | 4 | 59µs | 4 | 9µs | my $caseless = $type =~ s/^(.*)__(.*)_i$/$1$2/; # spent 9µs making 4 calls to utf8::CORE:subst, avg 2µs/call |
| 124 | |||||
| 125 | print STDERR __LINE__, ": type=$type, caseless=$caseless\n" if DEBUG; | ||||
| 126 | |||||
| 127 | 4 | 21µs | GETFILE: | ||
| 128 | { | ||||
| 129 | ## | ||||
| 130 | ## It could be a user-defined property. Look in current | ||||
| 131 | ## package if no package given | ||||
| 132 | ## | ||||
| 133 | |||||
| 134 | |||||
| 135 | 4 | 5µs | my $caller0 = caller(0); | ||
| 136 | 4 | 76µs | 4 | 10µs | my $caller1 = $type =~ s/(.+)::// ? $1 : $caller0 eq 'main' ? # spent 10µs making 4 calls to utf8::CORE:subst, avg 3µs/call |
| 137 | 'main' : caller(1); | ||||
| 138 | |||||
| 139 | 4 | 90µs | 4 | 25µs | if (defined $caller1 && $type =~ /^I[ns]\w+$/) { # spent 25µs making 4 calls to utf8::CORE:match, avg 6µs/call |
| 140 | 2 | 16µs | my $prop = "${caller1}::$type"; | ||
| 141 | 2 | 20µs | if (exists &{$prop}) { | ||
| 142 | # stolen from Scalar::Util::PP::tainted() | ||||
| 143 | my $tainted; | ||||
| 144 | { | ||||
| 145 | local($@, $SIG{__DIE__}, $SIG{__WARN__}); | ||||
| 146 | local $^W = 0; | ||||
| 147 | 2 | 377µs | 2 | 253µs | # spent 174µs (95+79) within utf8::BEGIN@147 which was called:
# once (95µs+79µs) by utf8::AUTOLOAD at line 147 # spent 174µs making 1 call to utf8::BEGIN@147
# spent 79µs making 1 call to warnings::unimport |
| 148 | eval { kill 0 * $prop }; | ||||
| 149 | $tainted = 1 if $@ =~ /^Insecure/; | ||||
| 150 | } | ||||
| 151 | die "Insecure user-defined property \\p{$prop}\n" | ||||
| 152 | if $tainted; | ||||
| 153 | 2 | 352µs | 2 | 209µs | # spent 131µs (53+78) within utf8::BEGIN@153 which was called:
# once (53µs+78µs) by utf8::AUTOLOAD at line 153 # spent 131µs making 1 call to utf8::BEGIN@153
# spent 78µs making 1 call to strict::unimport |
| 154 | $list = &{$prop}($caseless); | ||||
| 155 | $user_defined = 1; | ||||
| 156 | last GETFILE; | ||||
| 157 | } | ||||
| 158 | } | ||||
| 159 | |||||
| 160 | # During Perl's compilation, this routine may be called before | ||||
| 161 | # the tables are constructed. If so, we have a chicken/egg | ||||
| 162 | # problem. If we die, the tables never get constructed, so | ||||
| 163 | # keep going, but return an empty table so only what the code | ||||
| 164 | # has compiled in internally (currently ASCII/Latin1 range | ||||
| 165 | # matching) will work. | ||||
| 166 | # spent 41µs within utf8::BEGIN@166 which was called:
# once (41µs+0s) by utf8::AUTOLOAD at line 170 | ||||
| 167 | # Poor man's constant, to avoid a run-time check. | ||||
| 168 | 1 | 70µs | $utf8::{miniperl} | ||
| 169 | = \! defined &DynaLoader::boot_DynaLoader; | ||||
| 170 | 1 | 498µs | 1 | 41µs | } # spent 41µs making 1 call to utf8::BEGIN@166 |
| 171 | 4 | 20µs | if (miniperl) { | ||
| 172 | eval "require '$unicore_dir/Heavy.pl'"; | ||||
| 173 | last GETFILE if $@; | ||||
| 174 | } | ||||
| 175 | else { | ||||
| 176 | 4 | 45.2ms | require "$unicore_dir/Heavy.pl"; | ||
| 177 | } | ||||
| 178 | 1 | 10.6ms | 1 | 45µs | # spent 45µs within utf8::BEGIN@178 which was called:
# once (45µs+0s) by utf8::AUTOLOAD at line 178 # spent 45µs making 1 call to utf8::BEGIN@178 |
| 179 | |||||
| 180 | # All property names are matched caselessly | ||||
| 181 | 4 | 27µs | my $property_and_table = CORE::lc $type; | ||
| 182 | print STDERR __LINE__, ": $property_and_table\n" if DEBUG; | ||||
| 183 | |||||
| 184 | # See if is of the compound form 'property=value', where the | ||||
| 185 | # value indicates the table we should use. | ||||
| 186 | 4 | 95µs | my ($property, $table, @remainder) = | ||
| 187 | split /\s*[:=]\s*/, $property_and_table, -1; | ||||
| 188 | 4 | 9µs | if (@remainder) { | ||
| 189 | pop @recursed if @recursed; | ||||
| 190 | return $type; | ||||
| 191 | } | ||||
| 192 | |||||
| 193 | 4 | 4µs | my $prefix; | ||
| 194 | 4 | 16µs | if (! defined $table) { | ||
| 195 | |||||
| 196 | # Here, is the single form. The property becomes empty, and | ||||
| 197 | # the whole value is the table. | ||||
| 198 | 4 | 12µs | $table = $property; | ||
| 199 | 4 | 15µs | $prefix = $property = ""; | ||
| 200 | } else { | ||||
| 201 | print STDERR __LINE__, ": $property\n" if DEBUG; | ||||
| 202 | |||||
| 203 | # Here it is the compound property=table form. The property | ||||
| 204 | # name is always loosely matched, and always can have an | ||||
| 205 | # optional 'is' prefix (which isn't true in the single | ||||
| 206 | # form). | ||||
| 207 | $property = _loose_name($property) =~ s/^is//r; | ||||
| 208 | |||||
| 209 | # And convert to canonical form. Quit if not valid. | ||||
| 210 | $property = $utf8::loose_property_name_of{$property}; | ||||
| 211 | if (! defined $property) { | ||||
| 212 | pop @recursed if @recursed; | ||||
| 213 | return $type; | ||||
| 214 | } | ||||
| 215 | |||||
| 216 | $prefix = "$property="; | ||||
| 217 | |||||
| 218 | # If the rhs looks like it is a number... | ||||
| 219 | print STDERR __LINE__, ": table=$table\n" if DEBUG; | ||||
| 220 | if ($table =~ qr{ ^ [ \s 0-9 _ + / . -]+ $ }x) { | ||||
| 221 | print STDERR __LINE__, ": table=$table\n" if DEBUG; | ||||
| 222 | |||||
| 223 | # Don't allow leading nor trailing slashes | ||||
| 224 | if ($table =~ / ^ \/ | \/ $ /x) { | ||||
| 225 | pop @recursed if @recursed; | ||||
| 226 | return $type; | ||||
| 227 | } | ||||
| 228 | |||||
| 229 | # Split on slash, in case it is a rational, like \p{1/5} | ||||
| 230 | my @parts = split qr{ \s* / \s* }x, $table, -1; | ||||
| 231 | print __LINE__, ": $type\n" if @parts > 2 && DEBUG; | ||||
| 232 | |||||
| 233 | # Can have maximum of one slash | ||||
| 234 | if (@parts > 2) { | ||||
| 235 | pop @recursed if @recursed; | ||||
| 236 | return $type; | ||||
| 237 | } | ||||
| 238 | |||||
| 239 | foreach my $part (@parts) { | ||||
| 240 | print __LINE__, ": part=$part\n" if DEBUG; | ||||
| 241 | |||||
| 242 | $part =~ s/^\+\s*//; # Remove leading plus | ||||
| 243 | $part =~ s/^-\s*/-/; # Remove blanks after unary | ||||
| 244 | # minus | ||||
| 245 | |||||
| 246 | # Remove underscores between digits. | ||||
| 247 | $part =~ s/( ?<= [0-9] ) _ (?= [0-9] ) //xg; | ||||
| 248 | |||||
| 249 | # No leading zeros (but don't make a single '0' | ||||
| 250 | # into a null string) | ||||
| 251 | $part =~ s/ ^ ( -? ) 0+ /$1/x; | ||||
| 252 | $part .= '0' if $part eq '-' || $part eq ""; | ||||
| 253 | |||||
| 254 | # No trailing zeros after a decimal point | ||||
| 255 | $part =~ s/ ( \. .*? ) 0+ $ /$1/x; | ||||
| 256 | |||||
| 257 | # Begin with a 0 if a leading decimal point | ||||
| 258 | $part =~ s/ ^ ( -? ) \. /${1}0./x; | ||||
| 259 | |||||
| 260 | # Ensure not a trailing decimal point: turn into an | ||||
| 261 | # integer | ||||
| 262 | $part =~ s/ \. $ //x; | ||||
| 263 | |||||
| 264 | print STDERR __LINE__, ": part=$part\n" if DEBUG; | ||||
| 265 | #return $type if $part eq ""; | ||||
| 266 | |||||
| 267 | # Result better look like a number. (This test is | ||||
| 268 | # needed because, for example could have a plus in | ||||
| 269 | # the middle.) | ||||
| 270 | if ($part !~ / ^ -? [0-9]+ ( \. [0-9]+)? $ /x) { | ||||
| 271 | pop @recursed if @recursed; | ||||
| 272 | return $type; | ||||
| 273 | } | ||||
| 274 | } | ||||
| 275 | |||||
| 276 | # If a rational... | ||||
| 277 | if (@parts == 2) { | ||||
| 278 | |||||
| 279 | # If denominator is negative, get rid of it, and ... | ||||
| 280 | if ($parts[1] =~ s/^-//) { | ||||
| 281 | |||||
| 282 | # If numerator is also negative, convert the | ||||
| 283 | # whole thing to positive, or move the minus to | ||||
| 284 | # the numerator | ||||
| 285 | if ($parts[0] !~ s/^-//) { | ||||
| 286 | $parts[0] = '-' . $parts[0]; | ||||
| 287 | } | ||||
| 288 | } | ||||
| 289 | $table = join '/', @parts; | ||||
| 290 | } | ||||
| 291 | elsif ($property ne 'nv' || $parts[0] !~ /\./) { | ||||
| 292 | |||||
| 293 | # Here is not numeric value, or doesn't have a | ||||
| 294 | # decimal point. No further manipulation is | ||||
| 295 | # necessary. (Note the hard-coded property name. | ||||
| 296 | # This could fail if other properties eventually | ||||
| 297 | # had fractions as well; perhaps the cjk ones | ||||
| 298 | # could evolve to do that. This hard-coding could | ||||
| 299 | # be fixed by mktables generating a list of | ||||
| 300 | # properties that could have fractions.) | ||||
| 301 | $table = $parts[0]; | ||||
| 302 | } else { | ||||
| 303 | |||||
| 304 | # Here is a floating point numeric_value. Try to | ||||
| 305 | # convert to rational. First see if is in the list | ||||
| 306 | # of known ones. | ||||
| 307 | if (exists $utf8::nv_floating_to_rational{$parts[0]}) { | ||||
| 308 | $table = $utf8::nv_floating_to_rational{$parts[0]}; | ||||
| 309 | } else { | ||||
| 310 | |||||
| 311 | # Here not in the list. See if is close | ||||
| 312 | # enough to something in the list. First | ||||
| 313 | # determine what 'close enough' means. It has | ||||
| 314 | # to be as tight as what mktables says is the | ||||
| 315 | # maximum slop, and as tight as how many | ||||
| 316 | # digits we were passed. That is, if the user | ||||
| 317 | # said .667, .6667, .66667, etc. we match as | ||||
| 318 | # many digits as they passed until get to | ||||
| 319 | # where it doesn't matter any more due to the | ||||
| 320 | # machine's precision. If they said .6666668, | ||||
| 321 | # we fail. | ||||
| 322 | (my $fraction = $parts[0]) =~ s/^.*\.//; | ||||
| 323 | my $epsilon = 10 ** - (length($fraction)); | ||||
| 324 | if ($epsilon > $utf8::max_floating_slop) { | ||||
| 325 | $epsilon = $utf8::max_floating_slop; | ||||
| 326 | } | ||||
| 327 | |||||
| 328 | # But it can't be tighter than the minimum | ||||
| 329 | # precision for this machine. If haven't | ||||
| 330 | # already calculated that minimum, do so now. | ||||
| 331 | if (! defined $min_floating_slop) { | ||||
| 332 | |||||
| 333 | # Keep going down an order of magnitude | ||||
| 334 | # until find that adding this quantity to | ||||
| 335 | # 1 remains 1; but put an upper limit on | ||||
| 336 | # this so in case this algorithm doesn't | ||||
| 337 | # work properly on some platform, that we | ||||
| 338 | # won't loop forever. | ||||
| 339 | my $count = 0; | ||||
| 340 | $min_floating_slop = 1; | ||||
| 341 | while (1+ $min_floating_slop != 1 | ||||
| 342 | && $count++ < 50) | ||||
| 343 | { | ||||
| 344 | my $next = $min_floating_slop / 10; | ||||
| 345 | last if $next == 0; # If underflows, | ||||
| 346 | # use previous one | ||||
| 347 | $min_floating_slop = $next; | ||||
| 348 | print STDERR __LINE__, ": min_float_slop=$min_floating_slop\n" if DEBUG; | ||||
| 349 | } | ||||
| 350 | |||||
| 351 | # Back off a couple orders of magnitude, | ||||
| 352 | # just to be safe. | ||||
| 353 | $min_floating_slop *= 100; | ||||
| 354 | } | ||||
| 355 | |||||
| 356 | if ($epsilon < $min_floating_slop) { | ||||
| 357 | $epsilon = $min_floating_slop; | ||||
| 358 | } | ||||
| 359 | print STDERR __LINE__, ": fraction=.$fraction; epsilon=$epsilon\n" if DEBUG; | ||||
| 360 | |||||
| 361 | undef $table; | ||||
| 362 | |||||
| 363 | # And for each possible rational in the table, | ||||
| 364 | # see if it is within epsilon of the input. | ||||
| 365 | foreach my $official | ||||
| 366 | (keys %utf8::nv_floating_to_rational) | ||||
| 367 | { | ||||
| 368 | print STDERR __LINE__, ": epsilon=$epsilon, official=$official, diff=", abs($parts[0] - $official), "\n" if DEBUG; | ||||
| 369 | if (abs($parts[0] - $official) < $epsilon) { | ||||
| 370 | $table = | ||||
| 371 | $utf8::nv_floating_to_rational{$official}; | ||||
| 372 | last; | ||||
| 373 | } | ||||
| 374 | } | ||||
| 375 | |||||
| 376 | # Quit if didn't find one. | ||||
| 377 | if (! defined $table) { | ||||
| 378 | pop @recursed if @recursed; | ||||
| 379 | return $type; | ||||
| 380 | } | ||||
| 381 | } | ||||
| 382 | } | ||||
| 383 | print STDERR __LINE__, ": $property=$table\n" if DEBUG; | ||||
| 384 | } | ||||
| 385 | } | ||||
| 386 | |||||
| 387 | # Combine lhs (if any) and rhs to get something that matches | ||||
| 388 | # the syntax of the lookups. | ||||
| 389 | 4 | 10µs | $property_and_table = "$prefix$table"; | ||
| 390 | print STDERR __LINE__, ": $property_and_table\n" if DEBUG; | ||||
| 391 | |||||
| 392 | # First try stricter matching. | ||||
| 393 | 4 | 29µs | $file = $utf8::stricter_to_file_of{$property_and_table}; | ||
| 394 | |||||
| 395 | # If didn't find it, try again with looser matching by editing | ||||
| 396 | # out the applicable characters on the rhs and looking up | ||||
| 397 | # again. | ||||
| 398 | 4 | 10µs | if (! defined $file) { | ||
| 399 | 3 | 36µs | 3 | 147µs | $table = _loose_name($table); # spent 147µs making 3 calls to utf8::_loose_name, avg 49µs/call |
| 400 | 3 | 8µs | $property_and_table = "$prefix$table"; | ||
| 401 | print STDERR __LINE__, ": $property_and_table\n" if DEBUG; | ||||
| 402 | 3 | 18µs | $file = $utf8::loose_to_file_of{$property_and_table}; | ||
| 403 | } | ||||
| 404 | |||||
| 405 | # Add the constant and go fetch it in. | ||||
| 406 | 4 | 9µs | if (defined $file) { | ||
| 407 | |||||
| 408 | # A beginning ! means to invert. The 0+ makes sure is | ||||
| 409 | # numeric | ||||
| 410 | 4 | 161µs | 4 | 23µs | $invert_it = 0 + $file =~ s/^!//; # spent 23µs making 4 calls to utf8::CORE:subst, avg 6µs/call |
| 411 | |||||
| 412 | 4 | 11µs | if ($utf8::why_deprecated{$file}) { | ||
| 413 | warnings::warnif('deprecated', "Use of '$type' in \\p{} or \\P{} is deprecated because: $utf8::why_deprecated{$file};"); | ||||
| 414 | } | ||||
| 415 | |||||
| 416 | 4 | 5µs | if ($caseless | ||
| 417 | && exists $utf8::caseless_equivalent{$property_and_table}) | ||||
| 418 | { | ||||
| 419 | $file = $utf8::caseless_equivalent{$property_and_table}; | ||||
| 420 | } | ||||
| 421 | 4 | 29µs | $file= "$unicore_dir/lib/$file.pl"; | ||
| 422 | 4 | 45µs | last GETFILE; | ||
| 423 | } | ||||
| 424 | print STDERR __LINE__, ": didn't find $property_and_table\n" if DEBUG; | ||||
| 425 | |||||
| 426 | ## | ||||
| 427 | ## Last attempt -- see if it's a standard "To" name | ||||
| 428 | ## (e.g. "ToLower") ToTitle is used by ucfirst(). | ||||
| 429 | ## The user-level way to access ToDigit() and ToFold() | ||||
| 430 | ## is to use Unicode::UCD. | ||||
| 431 | ## | ||||
| 432 | # Only check if caller wants non-binary | ||||
| 433 | my $retried = 0; | ||||
| 434 | if ($minbits != 1 && $property_and_table =~ s/^to//) {{ | ||||
| 435 | # Look input up in list of properties for which we have | ||||
| 436 | # mapping files. | ||||
| 437 | if (defined ($file = | ||||
| 438 | $utf8::loose_property_to_file_of{$property_and_table})) | ||||
| 439 | { | ||||
| 440 | $type = $utf8::file_to_swash_name{$file}; | ||||
| 441 | print STDERR __LINE__, ": type set to $type\n" if DEBUG; | ||||
| 442 | $file = "$unicore_dir/$file.pl"; | ||||
| 443 | last GETFILE; | ||||
| 444 | } # If that fails see if there is a corresponding binary | ||||
| 445 | # property file | ||||
| 446 | elsif (defined ($file = | ||||
| 447 | $utf8::loose_to_file_of{$property_and_table})) | ||||
| 448 | { | ||||
| 449 | |||||
| 450 | # Here, there is no map file for the property we are | ||||
| 451 | # trying to get the map of, but this is a binary | ||||
| 452 | # property, and there is a file for it that can easily | ||||
| 453 | # be translated to a mapping. | ||||
| 454 | |||||
| 455 | # In the case of properties that are forced to binary, | ||||
| 456 | # they are a combination. We return the actual | ||||
| 457 | # mapping instead of the binary. If the input is | ||||
| 458 | # something like 'Tocjkkiicore', it will be found in | ||||
| 459 | # %loose_property_to_file_of above as => 'To/kIICore'. | ||||
| 460 | # But the form like ToIskiicore won't be. To fix | ||||
| 461 | # this, it was easiest to do it here. These | ||||
| 462 | # properties are the complements of the default | ||||
| 463 | # property, so there is an entry in %loose_to_file_of | ||||
| 464 | # that is 'iskiicore' => '!kIICore/N', If we find such | ||||
| 465 | # an entry, strip off things and try again, which | ||||
| 466 | # should find the entry in %loose_property_to_file_of. | ||||
| 467 | # Actual binary properties that are of this form, such | ||||
| 468 | # as this entry: 'ishrkt' => '!Perl/Any' will also be | ||||
| 469 | # retried, but won't be in %loose_property_to_file_of, | ||||
| 470 | # and instead the next time through, it will find | ||||
| 471 | # 'hrkt' => '!Perl/Any' and proceed. | ||||
| 472 | redo if ! $retried | ||||
| 473 | && $file =~ /^!/ | ||||
| 474 | && $property_and_table =~ s/^is//; | ||||
| 475 | |||||
| 476 | # This is a binary property. Setting this here causes | ||||
| 477 | # it to be stored as such in the cache, so if someone | ||||
| 478 | # comes along later looking for just a binary, they | ||||
| 479 | # get it. | ||||
| 480 | $minbits = 1; | ||||
| 481 | |||||
| 482 | # The 0+ makes sure is numeric | ||||
| 483 | $invert_it = 0 + $file =~ s/^!//; | ||||
| 484 | $file = "$unicore_dir/lib/$file.pl"; | ||||
| 485 | last GETFILE; | ||||
| 486 | } | ||||
| 487 | } } | ||||
| 488 | |||||
| 489 | ## | ||||
| 490 | ## If we reach this line, it's because we couldn't figure | ||||
| 491 | ## out what to do with $type. Ouch. | ||||
| 492 | ## | ||||
| 493 | |||||
| 494 | pop @recursed if @recursed; | ||||
| 495 | return $type; | ||||
| 496 | } # end of GETFILE block | ||||
| 497 | |||||
| 498 | 4 | 20µs | if (defined $file) { | ||
| 499 | print STDERR __LINE__, ": found it (file='$file')\n" if DEBUG; | ||||
| 500 | |||||
| 501 | ## | ||||
| 502 | ## If we reach here, it was due to a 'last GETFILE' above | ||||
| 503 | ## (exception: user-defined properties and mappings), so we | ||||
| 504 | ## have a filename, so now we load it if we haven't already. | ||||
| 505 | ## If we have, return the cached results. The cache key is the | ||||
| 506 | ## class and file to load, and whether the results need to be | ||||
| 507 | ## inverted. | ||||
| 508 | ## | ||||
| 509 | 4 | 48µs | my $found = $Cache{$class, $file, $invert_it}; | ||
| 510 | 4 | 7µs | if ($found and ref($found) eq $class) { | ||
| 511 | print STDERR __LINE__, ": Returning cached swash for '$class,$file,$invert_it' for \\p{$type}\n" if DEBUG; | ||||
| 512 | 1 | 4µs | pop @recursed if @recursed; | ||
| 513 | 1 | 27µs | return $found; | ||
| 514 | } | ||||
| 515 | |||||
| 516 | 3 | 6µs | local $@; | ||
| 517 | 3 | 360µs | local $!; | ||
| 518 | 6 | 5.48ms | $list = do $file; die $@ if $@; | ||
| 519 | 3 | 39µs | $list_is_from_mktables = 1; | ||
| 520 | } | ||||
| 521 | } # End of $type is non-null | ||||
| 522 | |||||
| 523 | # Here, either $type was null, or we found the requested property and | ||||
| 524 | # read it into $list | ||||
| 525 | |||||
| 526 | 4 | 11µs | my $extras = ""; | ||
| 527 | |||||
| 528 | 4 | 5µs | my $bits = $minbits; | ||
| 529 | |||||
| 530 | # mktables lists don't have extras, like '&utf8::prop', so don't need | ||||
| 531 | # to separate them; also lists are already sorted, so don't need to do | ||||
| 532 | # that. | ||||
| 533 | 4 | 9µs | if ($list && ! $list_is_from_mktables) { | ||
| 534 | 1 | 6µs | my $taint = substr($list,0,0); # maintain taint | ||
| 535 | |||||
| 536 | # Separate the extras from the code point list, and for | ||||
| 537 | # user-defined properties, make sure the latter are well-behaved | ||||
| 538 | # for downstream code. | ||||
| 539 | 1 | 3µs | if ($user_defined) { | ||
| 540 | my @tmp = split(/^/m, $list); | ||||
| 541 | my %seen; | ||||
| 542 | 2 | 3.96ms | 2 | 257µs | # spent 171µs (85+86) within utf8::BEGIN@542 which was called:
# once (85µs+86µs) by utf8::AUTOLOAD at line 542 # spent 171µs making 1 call to utf8::BEGIN@542
# spent 86µs making 1 call to warnings::unimport |
| 543 | |||||
| 544 | # The extras are anything that doesn't begin with a hex digit. | ||||
| 545 | $extras = join '', $taint, grep /^[^0-9a-fA-F]/, @tmp; | ||||
| 546 | |||||
| 547 | # Remove the extras, and sort the remaining entries by the | ||||
| 548 | # numeric value of their beginning hex digits, removing any | ||||
| 549 | # duplicates. | ||||
| 550 | $list = join '', $taint, | ||||
| 551 | map { $_->[1] } | ||||
| 552 | sort { $a->[0] <=> $b->[0] } | ||||
| 553 | map { /^([0-9a-fA-F]+)/; [ CORE::hex($1), $_ ] } | ||||
| 554 | grep { /^([0-9a-fA-F]+)/ and not $seen{$1}++ } @tmp; # XXX doesn't do ranges right | ||||
| 555 | } | ||||
| 556 | else { | ||||
| 557 | # mktables has gone to some trouble to make non-user defined | ||||
| 558 | # properties well-behaved, so we can skip the effort we do for | ||||
| 559 | # user-defined ones. Any extras are at the very beginning of | ||||
| 560 | # the string. | ||||
| 561 | |||||
| 562 | # This regex splits out the first lines of $list into $1 and | ||||
| 563 | # strips them off from $list, until we get one that begins | ||||
| 564 | # with a hex number, alone on the line, or followed by a tab. | ||||
| 565 | # Either portion may be empty. | ||||
| 566 | 1 | 48µs | 1 | 26µs | $list =~ s/ \A ( .*? ) # spent 26µs making 1 call to utf8::CORE:subst |
| 567 | (?: \z | (?= ^ [0-9a-fA-F]+ (?: \t | $) ) ) | ||||
| 568 | //msx; | ||||
| 569 | |||||
| 570 | 1 | 11µs | $extras = "$taint$1"; | ||
| 571 | } | ||||
| 572 | } | ||||
| 573 | |||||
| 574 | 4 | 4µs | if ($none) { | ||
| 575 | my $hextra = sprintf "%04x", $none + 1; | ||||
| 576 | $list =~ s/\tXXXX$/\t$hextra/mg; | ||||
| 577 | } | ||||
| 578 | |||||
| 579 | 4 | 5µs | if ($minbits != 1 && $minbits < 32) { # not binary property | ||
| 580 | my $top = 0; | ||||
| 581 | while ($list =~ /^([0-9a-fA-F]+)(?:[\t]([0-9a-fA-F]+)?)(?:[ \t]([0-9a-fA-F]+))?/mg) { | ||||
| 582 | my $min = CORE::hex $1; | ||||
| 583 | my $max = defined $2 ? CORE::hex $2 : $min; | ||||
| 584 | my $val = defined $3 ? CORE::hex $3 : 0; | ||||
| 585 | $val += $max - $min if defined $3; | ||||
| 586 | $top = $val if $val > $top; | ||||
| 587 | } | ||||
| 588 | my $topbits = | ||||
| 589 | $top > 0xffff ? 32 : | ||||
| 590 | $top > 0xff ? 16 : 8; | ||||
| 591 | $bits = $topbits if $bits < $topbits; | ||||
| 592 | } | ||||
| 593 | |||||
| 594 | 4 | 6µs | my @extras; | ||
| 595 | 4 | 7µs | if ($extras) { | ||
| 596 | 1 | 4µs | for my $x ($extras) { | ||
| 597 | 1 | 4µs | my $taint = substr($x,0,0); # maintain taint | ||
| 598 | 1 | 10µs | pos $x = 0; | ||
| 599 | 1 | 34µs | 1 | 8µs | while ($x =~ /^([^0-9a-fA-F\n])(.*)/mg) { # spent 8µs making 1 call to utf8::CORE:match |
| 600 | 2 | 13µs | my $char = "$1$taint"; | ||
| 601 | 2 | 10µs | my $name = "$2$taint"; | ||
| 602 | print STDERR __LINE__, ": char [$char] => name [$name]\n" | ||||
| 603 | if DEBUG; | ||||
| 604 | 2 | 78µs | 4 | 24µs | if ($char =~ /[-+!&]/) { # spent 24µs making 4 calls to utf8::CORE:match, avg 6µs/call |
| 605 | 1 | 11µs | my ($c,$t) = split(/::/, $name, 2); # bogus use of ::, really | ||
| 606 | 1 | 900ns | my $subobj; | ||
| 607 | 1 | 82µs | 1 | 0s | if ($c eq 'utf8') { # spent 368µs making 1 call to utf8::SWASHNEW, recursion: max depth 1, sum of overlapping time 368µs |
| 608 | $subobj = utf8->SWASHNEW($t, "", $minbits, 0); | ||||
| 609 | } | ||||
| 610 | elsif (exists &$name) { | ||||
| 611 | $subobj = utf8->SWASHNEW($name, "", $minbits, 0); | ||||
| 612 | } | ||||
| 613 | elsif ($c =~ /^([0-9a-fA-F]+)/) { | ||||
| 614 | $subobj = utf8->SWASHNEW("", $c, $minbits, 0); | ||||
| 615 | } | ||||
| 616 | print STDERR __LINE__, ": returned from getting sub object for $name\n" if DEBUG; | ||||
| 617 | 1 | 2µs | if (! ref $subobj) { | ||
| 618 | pop @recursed if @recursed && $type; | ||||
| 619 | return $subobj; | ||||
| 620 | } | ||||
| 621 | 1 | 4µs | push @extras, $name => $subobj; | ||
| 622 | 1 | 3µs | $bits = $subobj->{BITS} if $bits < $subobj->{BITS}; | ||
| 623 | 1 | 4µs | $user_defined = $subobj->{USER_DEFINED} | ||
| 624 | if $subobj->{USER_DEFINED}; | ||||
| 625 | } | ||||
| 626 | } | ||||
| 627 | } | ||||
| 628 | } | ||||
| 629 | |||||
| 630 | if (DEBUG) { | ||||
| 631 | print STDERR __LINE__, ": CLASS = $class, TYPE => $type, BITS => $bits, NONE => $none, INVERT_IT => $invert_it, USER_DEFINED => $user_defined"; | ||||
| 632 | print STDERR "\nLIST =>\n$list" if defined $list; | ||||
| 633 | print STDERR "\nEXTRAS =>\n$extras" if defined $extras; | ||||
| 634 | print STDERR "\n"; | ||||
| 635 | } | ||||
| 636 | |||||
| 637 | 4 | 192µs | my $SWASH = bless { | ||
| 638 | TYPE => $type, | ||||
| 639 | BITS => $bits, | ||||
| 640 | EXTRAS => $extras, | ||||
| 641 | LIST => $list, | ||||
| 642 | NONE => $none, | ||||
| 643 | USER_DEFINED => $user_defined, | ||||
| 644 | @extras, | ||||
| 645 | } => $class; | ||||
| 646 | |||||
| 647 | 4 | 10µs | if ($file) { | ||
| 648 | 3 | 35µs | $Cache{$class, $file, $invert_it} = $SWASH; | ||
| 649 | 3 | 8µs | if ($type | ||
| 650 | && exists $utf8::SwashInfo{$type} | ||||
| 651 | && exists $utf8::SwashInfo{$type}{'specials_name'}) | ||||
| 652 | { | ||||
| 653 | my $specials_name = $utf8::SwashInfo{$type}{'specials_name'}; | ||||
| 654 | 2 | 839µs | 2 | 249µs | # spent 156µs (63+93) within utf8::BEGIN@654 which was called:
# once (63µs+93µs) by utf8::AUTOLOAD at line 654 # spent 156µs making 1 call to utf8::BEGIN@654
# spent 93µs making 1 call to strict::unimport |
| 655 | print STDERR "\nspecials_name => $specials_name\n" if DEBUG; | ||||
| 656 | $SWASH->{'SPECIALS'} = \%$specials_name; | ||||
| 657 | } | ||||
| 658 | 3 | 17µs | $SWASH->{'INVERT_IT'} = $invert_it; | ||
| 659 | } | ||||
| 660 | |||||
| 661 | 4 | 19µs | pop @recursed if @recursed && $type; | ||
| 662 | |||||
| 663 | 4 | 137µs | return $SWASH; | ||
| 664 | } | ||||
| 665 | } | ||||
| 666 | |||||
| 667 | # Now SWASHGET is recasted into a C function S_swatch_get (see utf8.c). | ||||
| 668 | |||||
| 669 | 1 | 17µs | 1; | ||
# spent 65µs within utf8::CORE:match which was called 12 times, avg 5µs/call:
# 4 times (25µs+0s) by utf8::SWASHNEW at line 139, avg 6µs/call
# 4 times (24µs+0s) by utf8::SWASHNEW at line 604, avg 6µs/call
# 3 times (8µs+0s) by utf8::_loose_name at line 23, avg 3µs/call
# once (8µs+0s) by utf8::SWASHNEW at line 599 | |||||
# spent 140µs within utf8::CORE:subst which was called 24 times, avg 6µs/call:
# 4 times (36µs+0s) by utf8::SWASHNEW at line 118, avg 9µs/call
# 4 times (25µs+0s) by utf8::SWASHNEW at line 119, avg 6µs/call
# 4 times (23µs+0s) by utf8::SWASHNEW at line 410, avg 6µs/call
# 4 times (10µs+0s) by utf8::SWASHNEW at line 136, avg 3µs/call
# 4 times (9µs+0s) by utf8::SWASHNEW at line 123, avg 2µs/call
# 3 times (10µs+0s) by utf8::_loose_name at line 21, avg 3µs/call
# once (26µs+0s) by utf8::SWASHNEW at line 566 |