+ GETFILE:
+ {
+ ##
+ ## It could be a user-defined property. Look in current
+ ## package if no package given
+ ##
+
+
+ my $caller0 = caller(0);
+ my $caller1 = $type =~ s/(.+):://
+ ? $1
+ : $caller0 eq 'main'
+ ? 'main'
+ : caller(1);
+
+ if (defined $caller1 && $type =~ /^I[ns]\w+$/) {
+ my $prop = "${caller1}::$type";
+ if (exists &{$prop}) {
+ # stolen from Scalar::Util::PP::tainted()
+ my $tainted;
+ {
+ local($@, $SIG{__DIE__}, $SIG{__WARN__});
+ local $^W = 0;
+ no warnings;
+ eval { kill 0 * $prop };
+ $tainted = 1 if $@ =~ /^Insecure/;
+ }
+ die "Insecure user-defined property \\p{$prop}\n"
+ if $tainted;
+ no strict 'refs';
+ $list = &{$prop}($caseless);
+ $user_defined = 1;
+ last GETFILE;
+ }
+ }
+
+ # During Perl's compilation, this routine may be called before
+ # the tables are constructed. If so, we have a chicken/egg
+ # problem. If we die, the tables never get constructed, so
+ # keep going, but return an empty table so only what the code
+ # has compiled in internally (currently ASCII/Latin1 range
+ # matching) will work.
+ BEGIN {
+ # Poor man's constant, to avoid a run-time check.
+ $utf8::{miniperl}
+ = \! defined &DynaLoader::boot_DynaLoader;
+ }
+ if (miniperl) {
+ eval "require '$unicore_dir/Heavy.pl'";
+ if ($@) {
+ print STDERR __LINE__, ": '$@'\n" if DEBUG;
+ pop @recursed if @recursed;
+ return $type;
+ }
+ }
+ else {
+ require "$unicore_dir/Heavy.pl";
+ }
+ BEGIN { delete $utf8::{miniperl} }
+
+ # All property names are matched caselessly
+ my $property_and_table = CORE::lc $type;
+ print STDERR __LINE__, ": $property_and_table\n" if DEBUG;
+
+ # See if is of the compound form 'property=value', where the
+ # value indicates the table we should use.
+ my ($property, $table, @remainder) =
+ split /\s*[:=]\s*/, $property_and_table, -1;
+ if (@remainder) {
+ pop @recursed if @recursed;
+ return $type;
+ }
+
+ my $prefix;
+ if (! defined $table) {
+
+ # Here, is the single form. The property becomes empty, and
+ # the whole value is the table.
+ $table = $property;
+ $prefix = $property = "";
+ } else {
+ print STDERR __LINE__, ": $property\n" if DEBUG;
+
+ # Here it is the compound property=table form. The property
+ # name is always loosely matched, and always can have an
+ # optional 'is' prefix (which isn't true in the single
+ # form).
+ $property = _loose_name($property) =~ s/^is//r;
+
+ # And convert to canonical form. Quit if not valid.
+ $property = $utf8::loose_property_name_of{$property};
+ if (! defined $property) {
+ pop @recursed if @recursed;
+ return $type;
+ }
+
+ $prefix = "$property=";
+
+ # If the rhs looks like it is a number...
+ print STDERR __LINE__, ": table=$table\n" if DEBUG;
+
+ if ($table =~ $number) {
+ print STDERR __LINE__, ": table=$table\n" if DEBUG;
+
+ # Split on slash, in case it is a rational, like \p{1/5}
+ my @parts = split m{ \s* / \s* }x, $table, -1;
+ print __LINE__, ": $type\n" if @parts > 2 && DEBUG;
+
+ foreach my $part (@parts) {
+ print __LINE__, ": part=$part\n" if DEBUG;
+
+ $part =~ s/^\+\s*//; # Remove leading plus
+ $part =~ s/^-\s*/-/; # Remove blanks after unary
+ # minus
+
+ # Remove underscores between digits.
+ $part =~ s/(?<= [0-9] ) _ (?= [0-9] ) //xg;
+
+ # No leading zeros (but don't make a single '0'
+ # into a null string)
+ $part =~ s/ ^ ( -? ) 0+ /$1/x;
+ $part .= '0' if $part eq '-' || $part eq "";
+
+ # No trailing zeros after a decimal point
+ $part =~ s/ ( \. [0-9]*? ) 0+ $ /$1/x;
+
+ # Begin with a 0 if a leading decimal point
+ $part =~ s/ ^ ( -? ) \. /${1}0./x;
+
+ # Ensure not a trailing decimal point: turn into an
+ # integer
+ $part =~ s/ \. $ //x;
+
+ print STDERR __LINE__, ": part=$part\n" if DEBUG;
+ #return $type if $part eq "";
+ }
+
+ # If a rational...
+ if (@parts == 2) {
+
+ # If denominator is negative, get rid of it, and ...
+ if ($parts[1] =~ s/^-//) {
+
+ # If numerator is also negative, convert the
+ # whole thing to positive, else move the minus
+ # to the numerator
+ if ($parts[0] !~ s/^-//) {
+ $parts[0] = '-' . $parts[0];
+ }
+ }
+ $table = join '/', @parts;
+ }
+ elsif ($property ne 'nv' || $parts[0] !~ /\./) {
+
+ # Here is not numeric value, or doesn't have a
+ # decimal point. No further manipulation is
+ # necessary. (Note the hard-coded property name.
+ # This could fail if other properties eventually
+ # had fractions as well; perhaps the cjk ones
+ # could evolve to do that. This hard-coding could
+ # be fixed by mktables generating a list of
+ # properties that could have fractions.)
+ $table = $parts[0];
+ } else {
+
+ # Here is a floating point numeric_value. Convert
+ # to rational. Get a normalized form, like
+ # 5.00E-01, and look that up in the hash
+
+ my $float = sprintf "%.*e",
+ $utf8::e_precision,
+ 0 + $parts[0];
+
+ if (exists $utf8::nv_floating_to_rational{$float}) {
+ $table = $utf8::nv_floating_to_rational{$float};
+ } else {
+ pop @recursed if @recursed;
+ return $type;
+ }
+ }
+ print STDERR __LINE__, ": $property=$table\n" if DEBUG;
+ }
+ }
+
+ # Combine lhs (if any) and rhs to get something that matches
+ # the syntax of the lookups.
+ $property_and_table = "$prefix$table";
+ print STDERR __LINE__, ": $property_and_table\n" if DEBUG;
+
+ # First try stricter matching.
+ $file = $utf8::stricter_to_file_of{$property_and_table};
+
+ # If didn't find it, try again with looser matching by editing
+ # out the applicable characters on the rhs and looking up
+ # again.
+ my $strict_property_and_table;
+ if (! defined $file) {
+
+ # This isn't used unless the name begins with 'to'
+ $strict_property_and_table = $property_and_table =~ s/^to//r;
+ $table = _loose_name($table);
+ $property_and_table = "$prefix$table";
+ print STDERR __LINE__, ": $property_and_table\n" if DEBUG;
+ $file = $utf8::loose_to_file_of{$property_and_table};
+ }
+
+ # Add the constant and go fetch it in.
+ if (defined $file) {
+
+ # If the file name contains a !, it means to invert. The
+ # 0+ makes sure result is numeric
+ $invert_it = 0 + $file =~ s/!//;
+
+ if ($utf8::why_deprecated{$file}) {
+ warnings::warnif('deprecated', "Use of '$type' in \\p{} or \\P{} is deprecated because: $utf8::why_deprecated{$file};");
+ }
+
+ if ($caseless
+ && exists $utf8::caseless_equivalent{$property_and_table})
+ {
+ $file = $utf8::caseless_equivalent{$property_and_table};
+ }
+
+ # The pseudo-directory '#' means that there really isn't a
+ # file to read, the data is in-line as part of the string;
+ # we extract it below.
+ $file = "$unicore_dir/lib/$file.pl" unless $file =~ m!^#/!;
+ last GETFILE;
+ }
+ print STDERR __LINE__, ": didn't find $property_and_table\n" if DEBUG;
+
+ ##
+ ## Last attempt -- see if it's a standard "To" name
+ ## (e.g. "ToLower") ToTitle is used by ucfirst().
+ ## The user-level way to access ToDigit() and ToFold()
+ ## is to use Unicode::UCD.
+ ##
+ # Only check if caller wants non-binary
+ if ($minbits != 1) {
+ if ($property_and_table =~ s/^to//) {
+ # Look input up in list of properties for which we have
+ # mapping files. First do it with the strict approach
+ if (defined ($file = $utf8::strict_property_to_file_of{
+ $strict_property_and_table}))
+ {
+ $type = $utf8::file_to_swash_name{$file};
+ print STDERR __LINE__, ": type set to $type\n"
+ if DEBUG;
+ $file = "$unicore_dir/$file.pl";
+ last GETFILE;
+ }
+ elsif (defined ($file =
+ $utf8::loose_property_to_file_of{$property_and_table}))
+ {
+ $type = $utf8::file_to_swash_name{$file};
+ print STDERR __LINE__, ": type set to $type\n"
+ if DEBUG;
+ $file = "$unicore_dir/$file.pl";
+ last GETFILE;
+ } # If that fails see if there is a corresponding binary
+ # property file
+ elsif (defined ($file =
+ $utf8::loose_to_file_of{$property_and_table}))
+ {
+
+ # Here, there is no map file for the property we
+ # are trying to get the map of, but this is a
+ # binary property, and there is a file for it that
+ # can easily be translated to a mapping, so use
+ # that, treating this as a binary property.
+ # Setting 'minbits' here causes it to be stored as
+ # such in the cache, so if someone comes along
+ # later looking for just a binary, they get it.
+ $minbits = 1;
+
+ # The 0+ makes sure is numeric
+ $invert_it = 0 + $file =~ s/!//;
+ $file = "$unicore_dir/lib/$file.pl"
+ unless $file =~ m!^#/!;
+ last GETFILE;
+ }
+ }
+ }
+
+ ##
+ ## If we reach this line, it's because we couldn't figure
+ ## out what to do with $type. Ouch.
+ ##
+
+ pop @recursed if @recursed;
+ return $type;
+ } # end of GETFILE block
+
+ if (defined $file) {
+ print STDERR __LINE__, ": found it (file='$file')\n" if DEBUG;
+
+ ##
+ ## If we reach here, it was due to a 'last GETFILE' above
+ ## (exception: user-defined properties and mappings), so we
+ ## have a filename, so now we load it if we haven't already.
+
+ # The pseudo-directory '#' means the result isn't really a
+ # file, but is in-line, with semi-colons to be turned into
+ # new-lines. Since it is in-line there is no advantage to
+ # caching the result
+ if ($file =~ s!^#/!!) {
+ $list = $utf8::inline_definitions[$file];
+ }
+ else {
+ # Here, we have an actual file to read in and load, but it
+ # may already have been read-in and cached. The cache key
+ # is the class and file to load, and whether the results
+ # need to be inverted.
+ my $found = $Cache{$class, $file, $invert_it};
+ if ($found and ref($found) eq $class) {
+ print STDERR __LINE__, ": Returning cached swash for '$class,$file,$invert_it' for \\p{$type}\n" if DEBUG;
+ pop @recursed if @recursed;
+ return $found;
+ }
+
+ local $@;
+ local $!;
+ $list = do $file; die $@ if $@;
+ }
+
+ $list_is_from_mktables = 1;
+ }
+ } # End of $type is non-null
+
+ # Here, either $type was null, or we found the requested property and
+ # read it into $list
+
+ my $extras = "";
+
+ my $bits = $minbits;
+
+ # mktables lists don't have extras, like '&utf8::prop', so don't need
+ # to separate them; also lists are already sorted, so don't need to do
+ # that.
+ if ($list && ! $list_is_from_mktables) {
+ my $taint = substr($list,0,0); # maintain taint
+
+ # Separate the extras from the code point list, and make sure
+ # user-defined properties and tr/// are well-behaved for
+ # downstream code.
+ if ($user_defined || $none) {
+ my @tmp = split(/^/m, $list);
+ my %seen;
+ no warnings;
+
+ # The extras are anything that doesn't begin with a hex digit.
+ $extras = join '', $taint, grep /^[^0-9a-fA-F]/, @tmp;
+
+ # Remove the extras, and sort the remaining entries by the
+ # numeric value of their beginning hex digits, removing any
+ # duplicates.
+ $list = join '', $taint,
+ map { $_->[1] }
+ sort { $a->[0] <=> $b->[0] }
+ map { /^([0-9a-fA-F]+)/ && !$seen{$1}++ ? [ CORE::hex($1), $_ ] : () }
+ @tmp; # XXX doesn't do ranges right
+ }
+ else {
+ # mktables has gone to some trouble to make non-user defined
+ # properties well-behaved, so we can skip the effort we do for
+ # user-defined ones. Any extras are at the very beginning of
+ # the string.
+
+ # This regex splits out the first lines of $list into $1 and
+ # strips them off from $list, until we get one that begins
+ # with a hex number, alone on the line, or followed by a tab.
+ # Either portion may be empty.
+ $list =~ s/ \A ( .*? )
+ (?: \z | (?= ^ [0-9a-fA-F]+ (?: \t | $) ) )
+ //msx;
+
+ $extras = "$taint$1";
+ }