1 package ExtUtils::XSSymSet;
5 use vars qw( $VERSION );
10 my($pkg,$maxlen,$silent) = @_;
12 # Allow absurdly long symbols here if we've told the compiler to
13 # do the shortening for us.
14 $maxlen = 2048 if $Config{'useshortenedsymbols'};
16 my($obj) = { '__M@xLen' => $maxlen, '__S!lent' => $silent };
22 my($self,$name,$maxlen,$silent) = @_;
24 unless (defined $maxlen) {
25 if (ref $self) { $maxlen ||= $self->{'__M@xLen'}; }
28 $maxlen = 2048 if $Config{'useshortenedsymbols'};
30 unless (defined $silent) {
31 if (ref $self) { $silent ||= $self->{'__S!lent'}; }
34 return $name if (length $name <= $maxlen);
37 # First, just try to remove duplicated delimiters
39 if (length $trimmed > $maxlen) {
40 # Next, all duplicated chars
41 $trimmed =~ s/(.)\1+/$1/g;
42 if (length $trimmed > $maxlen) {
43 my $squeezed = $trimmed;
44 my($xs,$prefix,$func) = $trimmed =~ /^(XS_)?(.*)_([^_]*)$/;
46 my $frac = 3; # replaces broken length-based calculations but w/same result
48 if (length $func <= 12) { # Try to preserve short function names
49 if ($frac > 1) { $pat .= '[^A-Z_]{' . ($frac - 1) . '}'; }
50 $prefix =~ s/$pat/$1/g;
51 $squeezed = "$xs$prefix" . "_$func";
52 if (length $squeezed > $maxlen) {
54 $prefix =~ s/$pat/$1/g;
55 $squeezed = "$xs$prefix" . "_$func";
59 if ($frac > 1) { $pat .= '[^A-Z_]{' . ($frac - 1) . '}'; }
60 $squeezed = "$prefix$func";
61 $squeezed =~ s/$pat/$1/g;
62 if (length "$xs$squeezed" > $maxlen) {
64 $squeezed =~ s/$pat/$1/g;
66 $squeezed = "$xs$squeezed";
68 if (length $squeezed <= $maxlen) { $trimmed = $squeezed; }
70 my $frac = int((length $trimmed - $maxlen) / length $trimmed + 0.5);
71 my $pat = '(.).{$frac}';
72 $trimmed =~ s/$pat/$1/g;
76 warn "Warning: long symbol $name\n\ttrimmed to $trimmed\n\t" unless $silent;
82 my($self,$sym,$maxlen,$silent) = @_;
83 my $trimmed = $self->get_trimmed($sym);
85 return $trimmed if defined $trimmed;
87 $maxlen ||= $self->{'__M@xLen'} || 31;
88 $silent ||= $self->{'__S!lent'} || 0;
89 $trimmed = $self->trimsym($sym,$maxlen,1);
90 if (exists $self->{$trimmed}) {
92 $trimmed = $self->trimsym($sym,$maxlen-3,$silent);
93 while (exists $self->{"${trimmed}_$i"}) { $i++; }
94 warn "Warning: duplicate symbol $trimmed\n\tchanged to ${trimmed}_$i\n\t(original was $sym)\n\t"
98 elsif (not $silent and $trimmed ne $sym) {
99 warn "Warning: long symbol $sym\n\ttrimmed to $trimmed\n\t";
101 $self->{$trimmed} = $sym;
102 $self->{'__N+Map'}->{$sym} = $trimmed;
109 my $trimmed = $self->{'__N+Map'}->{$sym};
110 if (defined $trimmed) {
111 delete $self->{'__N+Map'}->{$sym};
112 delete $self->{$trimmed};
120 $self->{'__N+Map'}->{$sym};
125 my($self,$trimmed) = @_;
130 sub all_orig { (keys %{$_[0]->{'__N+Map'}}); }
131 sub all_trimmed { (grep { /^\w+$/ } keys %{$_[0]}); }
137 ExtUtils::XSSymSet - keep sets of symbol names palatable to the VMS linker
141 use ExtUtils::XSSymSet;
143 $set = new ExtUtils::XSSymSet;
144 while ($sym = make_symbol()) { $set->addsym($sym); }
145 foreach $safesym ($set->all_trimmed) {
146 print "Processing $safesym (derived from ",
147 $self->get_orig($safesym), ")\n";
151 $safesym = ExtUtils::XSSymSet->trimsym($onesym);
155 Since the VMS linker distinguishes symbols based only on the first 31
156 characters of their names, it is occasionally necessary to shorten
157 symbol names in order to avoid collisions. (This is especially true of
158 names generated by xsubpp, since prefixes generated by nested package
159 names can become quite long.) C<ExtUtils::XSSymSet> provides functions to
160 shorten names in a consistent fashion, and to track a set of names to
161 insure that each is unique. While designed with F<xsubpp> in mind, it
162 may be used with any set of strings.
164 This package supplies the following functions, all of which should be
169 =item new([$maxlen[,$silent]])
171 Creates an empty C<ExtUtils::XSSymset> set of symbols. This function may be
172 called as a static method or via an existing object. If C<$maxlen> or
173 C<$silent> are specified, they are used as the defaults for maximum
174 name length and warning behavior in future calls to addsym() or
175 trimsym() via this object. If the compiler has been instructed to do its
176 own symbol shortening via C<$Config{'useshortenedsymbols'}>, a value of
177 2048 is assumed for C<$maxlen> as a way of bypassing the shortening done by
180 =item addsym($name[,$maxlen[,$silent]])
182 Creates a symbol name from C<$name>, using the methods described
183 under trimsym(), which is unique in this set of symbols, and returns
184 the new name. C<$name> and its resultant are added to the set, and
185 any future calls to addsym() specifying the same C<$name> will return
186 the same result, regardless of the value of C<$maxlen> specified.
187 Unless C<$silent> is true, warnings are output if C<$name> had to be
188 trimmed or changed in order to avoid collision with an existing symbol
189 name. C<$maxlen> and C<$silent> default to the values specified when
190 this set of symbols was created. This method must be called via an
193 =item trimsym($name[,$maxlen[,$silent]])
195 Creates a symbol name C<$maxlen> or fewer characters long from
196 C<$name> and returns it. If C<$name> is too long, it first tries to
197 shorten it by removing duplicate characters, then by periodically
198 removing non-underscore characters, and finally, if necessary, by
199 periodically removing characters of any type. C<$maxlen> defaults
200 to 31. Unless C<$silent> is true, a warning is output if C<$name>
201 is altered in any way. This function may be called either as a
202 static method or via an existing object, but in the latter case no
203 check is made to insure that the resulting name is unique in the
204 set of symbols. If the compiler has been instructed to do its
205 own symbol shortening via C<$Config{'useshortenedsymbols'}>, a value
206 of 2048 is assumed for C<$maxlen> as a way of bypassing the shortening
211 Removes C<$name> from the set of symbols, where C<$name> is the
212 original symbol name passed previously to addsym(). If C<$name>
213 existed in the set of symbols, returns its "trimmed" equivalent,
214 otherwise returns C<undef>. This method must be called via an
217 =item get_orig($trimmed)
219 Returns the original name which was trimmed to C<$trimmed> by a
220 previous call to addsym(), or C<undef> if C<$trimmed> does not
221 correspond to a member of this set of symbols. This method must be
222 called via an existing object.
224 =item get_trimmed($name)
226 Returns the trimmed name which was generated from C<$name> by a
227 previous call to addsym(), or C<undef> if C<$name> is not a member
228 of this set of symbols. This method must be called via an
233 Returns a list containing all of the original symbol names
238 Returns a list containing all of the trimmed symbol names
245 Charles Bailey E<lt>I<bailey@newman.upenn.edu>E<gt>
249 Last revised 8-Oct-2010, for Perl 5.13.6.