Commit | Line | Data |
---|---|---|
4633a7c4 LW |
1 | package overload; |
2 | ||
af048c18 | 3 | our $VERSION = '1.08'; |
b75c8c73 | 4 | |
a6006777 | 5 | sub nil {} |
6 | ||
4633a7c4 LW |
7 | sub OVERLOAD { |
8 | $package = shift; | |
9 | my %arg = @_; | |
a6006777 | 10 | my ($sub, $fb); |
11 | $ {$package . "::OVERLOAD"}{dummy}++; # Register with magic by touching. | |
12 | *{$package . "::()"} = \&nil; # Make it findable via fetchmethod. | |
4633a7c4 | 13 | for (keys %arg) { |
a6006777 | 14 | if ($_ eq 'fallback') { |
15 | $fb = $arg{$_}; | |
16 | } else { | |
17 | $sub = $arg{$_}; | |
18 | if (not ref $sub and $sub !~ /::/) { | |
44a8e56a | 19 | $ {$package . "::(" . $_} = $sub; |
20 | $sub = \&nil; | |
a6006777 | 21 | } |
22 | #print STDERR "Setting `$ {'package'}::\cO$_' to \\&`$sub'.\n"; | |
23 | *{$package . "::(" . $_} = \&{ $sub }; | |
24 | } | |
4633a7c4 | 25 | } |
a6006777 | 26 | ${$package . "::()"} = $fb; # Make it findable too (fallback only). |
4633a7c4 LW |
27 | } |
28 | ||
29 | sub import { | |
30 | $package = (caller())[0]; | |
31 | # *{$package . "::OVERLOAD"} = \&OVERLOAD; | |
32 | shift; | |
33 | $package->overload::OVERLOAD(@_); | |
34 | } | |
35 | ||
36 | sub unimport { | |
37 | $package = (caller())[0]; | |
a6006777 | 38 | ${$package . "::OVERLOAD"}{dummy}++; # Upgrade the table |
4633a7c4 LW |
39 | shift; |
40 | for (@_) { | |
a6006777 | 41 | if ($_ eq 'fallback') { |
42 | undef $ {$package . "::()"}; | |
43 | } else { | |
44 | delete $ {$package . "::"}{"(" . $_}; | |
45 | } | |
4633a7c4 LW |
46 | } |
47 | } | |
48 | ||
49 | sub Overloaded { | |
a6006777 | 50 | my $package = shift; |
51 | $package = ref $package if ref $package; | |
52 | $package->can('()'); | |
4633a7c4 LW |
53 | } |
54 | ||
44a8e56a | 55 | sub ov_method { |
56 | my $globref = shift; | |
57 | return undef unless $globref; | |
58 | my $sub = \&{*$globref}; | |
59 | return $sub if $sub ne \&nil; | |
60 | return shift->can($ {*$globref}); | |
61 | } | |
62 | ||
4633a7c4 | 63 | sub OverloadedStringify { |
a6006777 | 64 | my $package = shift; |
65 | $package = ref $package if ref $package; | |
44a8e56a | 66 | #$package->can('(""') |
ee239bfe IZ |
67 | ov_method mycan($package, '(""'), $package |
68 | or ov_method mycan($package, '(0+'), $package | |
69 | or ov_method mycan($package, '(bool'), $package | |
70 | or ov_method mycan($package, '(nomethod'), $package; | |
4633a7c4 LW |
71 | } |
72 | ||
73 | sub Method { | |
a6006777 | 74 | my $package = shift; |
05a4b9b1 | 75 | if(ref $package) { |
4de05ceb RGS |
76 | local $@; |
77 | local $!; | |
78 | require Scalar::Util; | |
05a4b9b1 BB |
79 | $package = Scalar::Util::blessed($package); |
80 | return undef if !defined $package; | |
81 | } | |
44a8e56a | 82 | #my $meth = $package->can('(' . shift); |
83 | ov_method mycan($package, '(' . shift), $package; | |
84 | #return $meth if $meth ne \&nil; | |
85 | #return $ {*{$meth}}; | |
4633a7c4 LW |
86 | } |
87 | ||
88 | sub AddrRef { | |
a6006777 | 89 | my $package = ref $_[0]; |
90 | return "$_[0]" unless $package; | |
b3c0ec7c | 91 | |
4de05ceb RGS |
92 | local $@; |
93 | local $!; | |
94 | require Scalar::Util; | |
295812e0 RGS |
95 | my $class = Scalar::Util::blessed($_[0]); |
96 | my $class_prefix = defined($class) ? "$class=" : ""; | |
97 | my $type = Scalar::Util::reftype($_[0]); | |
98 | my $addr = Scalar::Util::refaddr($_[0]); | |
99 | return sprintf("$class_prefix$type(0x%x)", $addr); | |
4633a7c4 LW |
100 | } |
101 | ||
1b1d102f | 102 | *StrVal = *AddrRef; |
4633a7c4 | 103 | |
44a8e56a | 104 | sub mycan { # Real can would leave stubs. |
105 | my ($package, $meth) = @_; | |
e1a479c5 | 106 | |
1e9bd118 NC |
107 | local $@; |
108 | local $!; | |
109 | require mro; | |
110 | ||
e1a479c5 BB |
111 | my $mro = mro::get_linear_isa($package); |
112 | foreach my $p (@$mro) { | |
113 | my $fqmeth = $p . q{::} . $meth; | |
114 | return \*{$fqmeth} if defined &{$fqmeth}; | |
44a8e56a | 115 | } |
e1a479c5 | 116 | |
44a8e56a | 117 | return undef; |
118 | } | |
119 | ||
b3ac6de7 | 120 | %constants = ( |
9cfe5470 RGS |
121 | 'integer' => 0x1000, # HINT_NEW_INTEGER |
122 | 'float' => 0x2000, # HINT_NEW_FLOAT | |
123 | 'binary' => 0x4000, # HINT_NEW_BINARY | |
124 | 'q' => 0x8000, # HINT_NEW_STRING | |
125 | 'qr' => 0x10000, # HINT_NEW_RE | |
b3ac6de7 IZ |
126 | ); |
127 | ||
ee239bfe IZ |
128 | %ops = ( with_assign => "+ - * / % ** << >> x .", |
129 | assign => "+= -= *= /= %= **= <<= >>= x= .=", | |
2877bd81 | 130 | num_comparison => "< <= > >= == !=", |
ee239bfe | 131 | '3way_comparison'=> "<=> cmp", |
2877bd81 | 132 | str_comparison => "lt le gt ge eq ne", |
6dd85743 | 133 | binary => '& &= | |= ^ ^=', |
ee239bfe IZ |
134 | unary => "neg ! ~", |
135 | mutators => '++ --', | |
f216259d | 136 | func => "atan2 cos sin exp abs log sqrt int", |
ee239bfe | 137 | conversion => 'bool "" 0+', |
f5284f61 | 138 | iterators => '<>', |
e2210791 | 139 | filetest => "-X", |
f5284f61 | 140 | dereferencing => '${} @{} %{} &{} *{}', |
ae20c3aa | 141 | matching => '~~', |
ee239bfe IZ |
142 | special => 'nomethod fallback ='); |
143 | ||
6b82e2f5 | 144 | use warnings::register; |
b3ac6de7 IZ |
145 | sub constant { |
146 | # Arguments: what, sub | |
147 | while (@_) { | |
6b82e2f5 | 148 | if (@_ == 1) { |
4498a751 | 149 | warnings::warnif ("Odd number of arguments for overload::constant"); |
6b82e2f5 A |
150 | last; |
151 | } | |
152 | elsif (!exists $constants {$_ [0]}) { | |
4498a751 | 153 | warnings::warnif ("`$_[0]' is not an overloadable type"); |
6b82e2f5 | 154 | } |
1e70e886 | 155 | elsif (!ref $_ [1] || "$_[1]" !~ /(^|=)CODE\(0x[0-9a-f]+\)$/) { |
6b82e2f5 A |
156 | # Can't use C<ref $_[1] eq "CODE"> above as code references can be |
157 | # blessed, and C<ref> would return the package the ref is blessed into. | |
158 | if (warnings::enabled) { | |
6b82e2f5 | 159 | $_ [1] = "undef" unless defined $_ [1]; |
4498a751 | 160 | warnings::warn ("`$_[1]' is not a code reference"); |
6b82e2f5 A |
161 | } |
162 | } | |
163 | else { | |
164 | $^H{$_[0]} = $_[1]; | |
f22a2069 | 165 | $^H |= $constants{$_[0]}; |
6b82e2f5 | 166 | } |
b3ac6de7 IZ |
167 | shift, shift; |
168 | } | |
169 | } | |
170 | ||
171 | sub remove_constant { | |
172 | # Arguments: what, sub | |
173 | while (@_) { | |
174 | delete $^H{$_[0]}; | |
175 | $^H &= ~ $constants{$_[0]}; | |
176 | shift, shift; | |
177 | } | |
178 | } | |
179 | ||
4633a7c4 LW |
180 | 1; |
181 | ||
182 | __END__ | |
183 | ||
b267980d | 184 | =head1 NAME |
4633a7c4 | 185 | |
7adf7a02 | 186 | overload - Package for overloading Perl operations |
4633a7c4 LW |
187 | |
188 | =head1 SYNOPSIS | |
189 | ||
190 | package SomeThing; | |
191 | ||
b267980d | 192 | use overload |
4633a7c4 LW |
193 | '+' => \&myadd, |
194 | '-' => \&mysub; | |
195 | # etc | |
196 | ... | |
197 | ||
198 | package main; | |
2b393bf4 | 199 | $a = SomeThing->new( 57 ); |
4633a7c4 LW |
200 | $b=5+$a; |
201 | ... | |
202 | if (overload::Overloaded $b) {...} | |
203 | ... | |
204 | $strval = overload::StrVal $b; | |
205 | ||
4633a7c4 LW |
206 | =head1 DESCRIPTION |
207 | ||
caf82a01 YST |
208 | This pragma allows overloading of Perl's operators for a class. |
209 | To overload built-in functions, see L<perlsub/Overriding Built-in Functions> instead. | |
210 | ||
4633a7c4 LW |
211 | =head2 Declaration of overloaded functions |
212 | ||
213 | The compilation directive | |
214 | ||
215 | package Number; | |
216 | use overload | |
b267980d | 217 | "+" => \&add, |
4633a7c4 LW |
218 | "*=" => "muas"; |
219 | ||
220 | declares function Number::add() for addition, and method muas() in | |
221 | the "class" C<Number> (or one of its base classes) | |
b267980d | 222 | for the assignment form C<*=> of multiplication. |
4633a7c4 LW |
223 | |
224 | Arguments of this directive come in (key, value) pairs. Legal values | |
e7ea3e70 IZ |
225 | are values legal inside a C<&{ ... }> call, so the name of a |
226 | subroutine, a reference to a subroutine, or an anonymous subroutine | |
227 | will all work. Note that values specified as strings are | |
228 | interpreted as methods, not subroutines. Legal keys are listed below. | |
4633a7c4 LW |
229 | |
230 | The subroutine C<add> will be called to execute C<$a+$b> if $a | |
231 | is a reference to an object blessed into the package C<Number>, or if $a is | |
232 | not an object from a package with defined mathemagic addition, but $b is a | |
233 | reference to a C<Number>. It can also be called in other situations, like | |
234 | C<$a+=7>, or C<$a++>. See L<MAGIC AUTOGENERATION>. (Mathemagical | |
235 | methods refer to methods triggered by an overloaded mathematical | |
236 | operator.) | |
237 | ||
774d564b | 238 | Since overloading respects inheritance via the @ISA hierarchy, the |
239 | above declaration would also trigger overloading of C<+> and C<*=> in | |
240 | all the packages which inherit from C<Number>. | |
e7ea3e70 | 241 | |
4633a7c4 LW |
242 | =head2 Calling Conventions for Binary Operations |
243 | ||
244 | The functions specified in the C<use overload ...> directive are called | |
245 | with three (in one particular case with four, see L<Last Resort>) | |
246 | arguments. If the corresponding operation is binary, then the first | |
247 | two arguments are the two arguments of the operation. However, due to | |
248 | general object calling conventions, the first argument should always be | |
249 | an object in the package, so in the situation of C<7+$a>, the | |
250 | order of the arguments is interchanged. It probably does not matter | |
251 | when implementing the addition method, but whether the arguments | |
252 | are reversed is vital to the subtraction method. The method can | |
253 | query this information by examining the third argument, which can take | |
254 | three different values: | |
255 | ||
256 | =over 7 | |
257 | ||
258 | =item FALSE | |
259 | ||
260 | the order of arguments is as in the current operation. | |
261 | ||
262 | =item TRUE | |
263 | ||
264 | the arguments are reversed. | |
265 | ||
266 | =item C<undef> | |
267 | ||
268 | the current operation is an assignment variant (as in | |
269 | C<$a+=7>), but the usual function is called instead. This additional | |
ee239bfe IZ |
270 | information can be used to generate some optimizations. Compare |
271 | L<Calling Conventions for Mutators>. | |
4633a7c4 LW |
272 | |
273 | =back | |
274 | ||
275 | =head2 Calling Conventions for Unary Operations | |
276 | ||
277 | Unary operation are considered binary operations with the second | |
278 | argument being C<undef>. Thus the functions that overloads C<{"++"}> | |
279 | is called with arguments C<($a,undef,'')> when $a++ is executed. | |
280 | ||
ee239bfe IZ |
281 | =head2 Calling Conventions for Mutators |
282 | ||
283 | Two types of mutators have different calling conventions: | |
284 | ||
88c28ceb | 285 | =over |
ee239bfe IZ |
286 | |
287 | =item C<++> and C<--> | |
288 | ||
289 | The routines which implement these operators are expected to actually | |
290 | I<mutate> their arguments. So, assuming that $obj is a reference to a | |
291 | number, | |
292 | ||
293 | sub incr { my $n = $ {$_[0]}; ++$n; $_[0] = bless \$n} | |
294 | ||
295 | is an appropriate implementation of overloaded C<++>. Note that | |
296 | ||
297 | sub incr { ++$ {$_[0]} ; shift } | |
298 | ||
299 | is OK if used with preincrement and with postincrement. (In the case | |
300 | of postincrement a copying will be performed, see L<Copy Constructor>.) | |
301 | ||
302 | =item C<x=> and other assignment versions | |
303 | ||
304 | There is nothing special about these methods. They may change the | |
305 | value of their arguments, and may leave it as is. The result is going | |
306 | to be assigned to the value in the left-hand-side if different from | |
307 | this value. | |
308 | ||
f610777f | 309 | This allows for the same method to be used as overloaded C<+=> and |
ee239bfe IZ |
310 | C<+>. Note that this is I<allowed>, but not recommended, since by the |
311 | semantic of L<"Fallback"> Perl will call the method for C<+> anyway, | |
312 | if C<+=> is not overloaded. | |
313 | ||
314 | =back | |
315 | ||
d1be9408 | 316 | B<Warning.> Due to the presence of assignment versions of operations, |
b267980d NIS |
317 | routines which may be called in assignment context may create |
318 | self-referential structures. Currently Perl will not free self-referential | |
ee239bfe IZ |
319 | structures until cycles are C<explicitly> broken. You may get problems |
320 | when traversing your structures too. | |
321 | ||
b267980d | 322 | Say, |
ee239bfe IZ |
323 | |
324 | use overload '+' => sub { bless [ \$_[0], \$_[1] ] }; | |
325 | ||
326 | is asking for trouble, since for code C<$obj += $foo> the subroutine | |
b267980d | 327 | is called as C<$obj = add($obj, $foo, undef)>, or C<$obj = [\$obj, |
ee239bfe IZ |
328 | \$foo]>. If using such a subroutine is an important optimization, one |
329 | can overload C<+=> explicitly by a non-"optimized" version, or switch | |
b267980d | 330 | to non-optimized version if C<not defined $_[2]> (see |
ee239bfe IZ |
331 | L<Calling Conventions for Binary Operations>). |
332 | ||
333 | Even if no I<explicit> assignment-variants of operators are present in | |
334 | the script, they may be generated by the optimizer. Say, C<",$obj,"> or | |
335 | C<',' . $obj . ','> may be both optimized to | |
336 | ||
337 | my $tmp = ',' . $obj; $tmp .= ','; | |
338 | ||
4633a7c4 LW |
339 | =head2 Overloadable Operations |
340 | ||
ee239bfe | 341 | The following symbols can be specified in C<use overload> directive: |
4633a7c4 LW |
342 | |
343 | =over 5 | |
344 | ||
345 | =item * I<Arithmetic operations> | |
346 | ||
347 | "+", "+=", "-", "-=", "*", "*=", "/", "/=", "%", "%=", | |
348 | "**", "**=", "<<", "<<=", ">>", ">>=", "x", "x=", ".", ".=", | |
349 | ||
350 | For these operations a substituted non-assignment variant can be called if | |
fa8a6580 MS |
351 | the assignment variant is not available. Methods for operations C<+>, |
352 | C<->, C<+=>, and C<-=> can be called to automatically generate | |
353 | increment and decrement methods. The operation C<-> can be used to | |
4633a7c4 LW |
354 | autogenerate missing methods for unary minus or C<abs>. |
355 | ||
ee239bfe IZ |
356 | See L<"MAGIC AUTOGENERATION">, L<"Calling Conventions for Mutators"> and |
357 | L<"Calling Conventions for Binary Operations">) for details of these | |
358 | substitutions. | |
359 | ||
4633a7c4 LW |
360 | =item * I<Comparison operations> |
361 | ||
362 | "<", "<=", ">", ">=", "==", "!=", "<=>", | |
363 | "lt", "le", "gt", "ge", "eq", "ne", "cmp", | |
364 | ||
365 | If the corresponding "spaceship" variant is available, it can be | |
366 | used to substitute for the missing operation. During C<sort>ing | |
367 | arrays, C<cmp> is used to compare values subject to C<use overload>. | |
368 | ||
369 | =item * I<Bit operations> | |
370 | ||
6dd85743 | 371 | "&", "&=", "^", "^=", "|", "|=", "neg", "!", "~", |
4633a7c4 | 372 | |
fa8a6580 | 373 | C<neg> stands for unary minus. If the method for C<neg> is not |
3bc6ec80 | 374 | specified, it can be autogenerated using the method for |
fa8a6580 MS |
375 | subtraction. If the method for C<!> is not specified, it can be |
376 | autogenerated using the methods for C<bool>, or C<"">, or C<0+>. | |
4633a7c4 | 377 | |
6dd85743 AF |
378 | The same remarks in L<"Arithmetic operations"> about |
379 | assignment-variants and autogeneration apply for | |
380 | bit operations C<"&">, C<"^">, and C<"|"> as well. | |
381 | ||
4633a7c4 LW |
382 | =item * I<Increment and decrement> |
383 | ||
384 | "++", "--", | |
385 | ||
386 | If undefined, addition and subtraction methods can be | |
387 | used instead. These operations are called both in prefix and | |
388 | postfix form. | |
389 | ||
390 | =item * I<Transcendental functions> | |
391 | ||
f216259d | 392 | "atan2", "cos", "sin", "exp", "abs", "log", "sqrt", "int" |
4633a7c4 LW |
393 | |
394 | If C<abs> is unavailable, it can be autogenerated using methods | |
1fef88e7 | 395 | for "E<lt>" or "E<lt>=E<gt>" combined with either unary minus or subtraction. |
4633a7c4 | 396 | |
f216259d IZ |
397 | Note that traditionally the Perl function L<int> rounds to 0, thus for |
398 | floating-point-like types one should follow the same semantic. If | |
399 | C<int> is unavailable, it can be autogenerated using the overloading of | |
400 | C<0+>. | |
401 | ||
4633a7c4 LW |
402 | =item * I<Boolean, string and numeric conversion> |
403 | ||
fa8a6580 | 404 | 'bool', '""', '0+', |
4633a7c4 | 405 | |
f5284f61 | 406 | If one or two of these operations are not overloaded, the remaining ones can |
4633a7c4 | 407 | be used instead. C<bool> is used in the flow control operators |
fa8a6580 | 408 | (like C<while>) and for the ternary C<?:> operation. These functions can |
4633a7c4 LW |
409 | return any arbitrary Perl value. If the corresponding operation for this value |
410 | is overloaded too, that operation will be called again with this value. | |
411 | ||
1554e226 DC |
412 | As a special case if the overload returns the object itself then it will |
413 | be used directly. An overloaded conversion returning the object is | |
414 | probably a bug, because you're likely to get something that looks like | |
415 | C<YourPackage=HASH(0x8172b34)>. | |
416 | ||
f5284f61 IZ |
417 | =item * I<Iteration> |
418 | ||
419 | "<>" | |
420 | ||
421 | If not overloaded, the argument will be converted to a filehandle or | |
422 | glob (which may require a stringification). The same overloading | |
423 | happens both for the I<read-filehandle> syntax C<E<lt>$varE<gt>> and | |
424 | I<globbing> syntax C<E<lt>${var}E<gt>>. | |
425 | ||
54f8c773 YST |
426 | B<BUGS> Even in list context, the iterator is currently called only |
427 | once and with scalar context. | |
428 | ||
e2210791 BM |
429 | =item * I<File tests> |
430 | ||
431 | "-X" | |
432 | ||
433 | This overload is used for all the filetest operators (C<-f>, C<-x> and | |
434 | so on: see L<perlfunc/-X> for the full list). Even though these are | |
435 | unary operators, the method will be called with a second argument which | |
436 | is a single letter indicating which test was performed. Note that the | |
437 | overload key is the literal string C<"-X">: you can't provide separate | |
438 | overloads for the different tests. | |
439 | ||
440 | Calling an overloaded filetest operator does not affect the stat value | |
441 | associated with the special filehandle C<_>. It still refers to the | |
442 | result of the last C<stat>, C<lstat> or unoverloaded filetest. | |
443 | ||
444 | If not overloaded, these operators will fall back to the default | |
445 | behaviour even without C<< fallback => 1 >>. This means that if the | |
446 | object is a blessed glob or blessed IO ref it will be treated as a | |
447 | filehandle, otherwise string overloading will be invoked and the result | |
448 | treated as a filename. | |
449 | ||
450 | This overload was introduced in perl 5.12. | |
451 | ||
ae20c3aa RGS |
452 | =item * I<Matching> |
453 | ||
454 | The key C<"~~"> allows you to override the smart matching used by | |
455 | the switch construct. See L<feature>. | |
456 | ||
f5284f61 IZ |
457 | =item * I<Dereferencing> |
458 | ||
459 | '${}', '@{}', '%{}', '&{}', '*{}'. | |
460 | ||
461 | If not overloaded, the argument will be dereferenced I<as is>, thus | |
462 | should be of correct type. These functions should return a reference | |
463 | of correct type, or another object with overloaded dereferencing. | |
464 | ||
b267980d NIS |
465 | As a special case if the overload returns the object itself then it |
466 | will be used directly (provided it is the correct type). | |
467 | ||
468 | The dereference operators must be specified explicitly they will not be passed to | |
469 | "nomethod". | |
470 | ||
4633a7c4 LW |
471 | =item * I<Special> |
472 | ||
ae20c3aa | 473 | "nomethod", "fallback", "=". |
4633a7c4 LW |
474 | |
475 | see L<SPECIAL SYMBOLS FOR C<use overload>>. | |
476 | ||
477 | =back | |
478 | ||
ee239bfe IZ |
479 | See L<"Fallback"> for an explanation of when a missing method can be |
480 | autogenerated. | |
481 | ||
482 | A computer-readable form of the above table is available in the hash | |
483 | %overload::ops, with values being space-separated lists of names: | |
484 | ||
485 | with_assign => '+ - * / % ** << >> x .', | |
486 | assign => '+= -= *= /= %= **= <<= >>= x= .=', | |
2877bd81 | 487 | num_comparison => '< <= > >= == !=', |
ee239bfe | 488 | '3way_comparison'=> '<=> cmp', |
2877bd81 | 489 | str_comparison => 'lt le gt ge eq ne', |
6dd85743 | 490 | binary => '& &= | |= ^ ^=', |
ee239bfe IZ |
491 | unary => 'neg ! ~', |
492 | mutators => '++ --', | |
493 | func => 'atan2 cos sin exp abs log sqrt', | |
494 | conversion => 'bool "" 0+', | |
f5284f61 | 495 | iterators => '<>', |
e2210791 | 496 | filetest => '-X', |
f5284f61 | 497 | dereferencing => '${} @{} %{} &{} *{}', |
ae20c3aa | 498 | matching => '~~', |
ee239bfe | 499 | special => 'nomethod fallback =' |
4633a7c4 | 500 | |
e7ea3e70 IZ |
501 | =head2 Inheritance and overloading |
502 | ||
774d564b | 503 | Inheritance interacts with overloading in two ways. |
e7ea3e70 | 504 | |
88c28ceb | 505 | =over |
e7ea3e70 IZ |
506 | |
507 | =item Strings as values of C<use overload> directive | |
508 | ||
774d564b | 509 | If C<value> in |
e7ea3e70 IZ |
510 | |
511 | use overload key => value; | |
512 | ||
774d564b | 513 | is a string, it is interpreted as a method name. |
e7ea3e70 IZ |
514 | |
515 | =item Overloading of an operation is inherited by derived classes | |
516 | ||
774d564b | 517 | Any class derived from an overloaded class is also overloaded. The |
518 | set of overloaded methods is the union of overloaded methods of all | |
519 | the ancestors. If some method is overloaded in several ancestor, then | |
e7ea3e70 | 520 | which description will be used is decided by the usual inheritance |
774d564b | 521 | rules: |
e7ea3e70 | 522 | |
774d564b | 523 | If C<A> inherits from C<B> and C<C> (in this order), C<B> overloads |
524 | C<+> with C<\&D::plus_sub>, and C<C> overloads C<+> by C<"plus_meth">, | |
525 | then the subroutine C<D::plus_sub> will be called to implement | |
526 | operation C<+> for an object in package C<A>. | |
e7ea3e70 IZ |
527 | |
528 | =back | |
529 | ||
774d564b | 530 | Note that since the value of the C<fallback> key is not a subroutine, |
531 | its inheritance is not governed by the above rules. In the current | |
532 | implementation, the value of C<fallback> in the first overloaded | |
533 | ancestor is used, but this is accidental and subject to change. | |
e7ea3e70 | 534 | |
4633a7c4 LW |
535 | =head1 SPECIAL SYMBOLS FOR C<use overload> |
536 | ||
537 | Three keys are recognized by Perl that are not covered by the above | |
538 | description. | |
539 | ||
774d564b | 540 | =head2 Last Resort |
4633a7c4 LW |
541 | |
542 | C<"nomethod"> should be followed by a reference to a function of four | |
543 | parameters. If defined, it is called when the overloading mechanism | |
544 | cannot find a method for some operation. The first three arguments of | |
545 | this function coincide with the arguments for the corresponding method if | |
546 | it were found, the fourth argument is the symbol | |
547 | corresponding to the missing method. If several methods are tried, | |
548 | the last one is used. Say, C<1-$a> can be equivalent to | |
549 | ||
550 | &nomethodMethod($a,1,1,"-") | |
551 | ||
552 | if the pair C<"nomethod" =E<gt> "nomethodMethod"> was specified in the | |
553 | C<use overload> directive. | |
554 | ||
b267980d NIS |
555 | The C<"nomethod"> mechanism is I<not> used for the dereference operators |
556 | ( ${} @{} %{} &{} *{} ). | |
557 | ||
558 | ||
4633a7c4 LW |
559 | If some operation cannot be resolved, and there is no function |
560 | assigned to C<"nomethod">, then an exception will be raised via die()-- | |
561 | unless C<"fallback"> was specified as a key in C<use overload> directive. | |
562 | ||
b267980d NIS |
563 | |
564 | =head2 Fallback | |
4633a7c4 LW |
565 | |
566 | The key C<"fallback"> governs what to do if a method for a particular | |
567 | operation is not found. Three different cases are possible depending on | |
568 | the value of C<"fallback">: | |
569 | ||
570 | =over 16 | |
571 | ||
572 | =item * C<undef> | |
573 | ||
574 | Perl tries to use a | |
575 | substituted method (see L<MAGIC AUTOGENERATION>). If this fails, it | |
576 | then tries to calls C<"nomethod"> value; if missing, an exception | |
577 | will be raised. | |
578 | ||
579 | =item * TRUE | |
580 | ||
581 | The same as for the C<undef> value, but no exception is raised. Instead, | |
582 | it silently reverts to what it would have done were there no C<use overload> | |
583 | present. | |
584 | ||
585 | =item * defined, but FALSE | |
586 | ||
587 | No autogeneration is tried. Perl tries to call | |
b267980d | 588 | C<"nomethod"> value, and if this is missing, raises an exception. |
4633a7c4 LW |
589 | |
590 | =back | |
591 | ||
e7ea3e70 IZ |
592 | B<Note.> C<"fallback"> inheritance via @ISA is not carved in stone |
593 | yet, see L<"Inheritance and overloading">. | |
594 | ||
4633a7c4 LW |
595 | =head2 Copy Constructor |
596 | ||
597 | The value for C<"="> is a reference to a function with three | |
598 | arguments, i.e., it looks like the other values in C<use | |
599 | overload>. However, it does not overload the Perl assignment | |
600 | operator. This would go against Camel hair. | |
601 | ||
602 | This operation is called in the situations when a mutator is applied | |
603 | to a reference that shares its object with some other reference, such | |
604 | as | |
605 | ||
b267980d | 606 | $a=$b; |
ee239bfe | 607 | ++$a; |
4633a7c4 LW |
608 | |
609 | To make this change $a and not change $b, a copy of C<$$a> is made, | |
610 | and $a is assigned a reference to this new object. This operation is | |
ee239bfe | 611 | done during execution of the C<++$a>, and not during the assignment, |
4633a7c4 | 612 | (so before the increment C<$$a> coincides with C<$$b>). This is only |
ee239bfe IZ |
613 | done if C<++> is expressed via a method for C<'++'> or C<'+='> (or |
614 | C<nomethod>). Note that if this operation is expressed via C<'+'> | |
615 | a nonmutator, i.e., as in | |
4633a7c4 | 616 | |
b267980d | 617 | $a=$b; |
4633a7c4 LW |
618 | $a=$a+1; |
619 | ||
620 | then C<$a> does not reference a new copy of C<$$a>, since $$a does not | |
621 | appear as lvalue when the above code is executed. | |
622 | ||
623 | If the copy constructor is required during the execution of some mutator, | |
624 | but a method for C<'='> was not specified, it can be autogenerated as a | |
49c95d58 RD |
625 | string copy if the object is a plain scalar or a simple assignment if it |
626 | is not. | |
4633a7c4 LW |
627 | |
628 | =over 5 | |
629 | ||
630 | =item B<Example> | |
631 | ||
b267980d | 632 | The actually executed code for |
4633a7c4 | 633 | |
b267980d | 634 | $a=$b; |
4633a7c4 LW |
635 | Something else which does not modify $a or $b.... |
636 | ++$a; | |
637 | ||
638 | may be | |
639 | ||
b267980d | 640 | $a=$b; |
4633a7c4 LW |
641 | Something else which does not modify $a or $b.... |
642 | $a = $a->clone(undef,""); | |
643 | $a->incr(undef,""); | |
644 | ||
645 | if $b was mathemagical, and C<'++'> was overloaded with C<\&incr>, | |
646 | C<'='> was overloaded with C<\&clone>. | |
647 | ||
648 | =back | |
649 | ||
f610777f | 650 | Same behaviour is triggered by C<$b = $a++>, which is consider a synonym for |
ee239bfe IZ |
651 | C<$b = $a; ++$a>. |
652 | ||
4633a7c4 LW |
653 | =head1 MAGIC AUTOGENERATION |
654 | ||
655 | If a method for an operation is not found, and the value for C<"fallback"> is | |
656 | TRUE or undefined, Perl tries to autogenerate a substitute method for | |
657 | the missing operation based on the defined operations. Autogenerated method | |
658 | substitutions are possible for the following operations: | |
659 | ||
660 | =over 16 | |
661 | ||
662 | =item I<Assignment forms of arithmetic operations> | |
663 | ||
664 | C<$a+=$b> can use the method for C<"+"> if the method for C<"+="> | |
665 | is not defined. | |
666 | ||
b267980d | 667 | =item I<Conversion operations> |
4633a7c4 LW |
668 | |
669 | String, numeric, and boolean conversion are calculated in terms of one | |
670 | another if not all of them are defined. | |
671 | ||
672 | =item I<Increment and decrement> | |
673 | ||
674 | The C<++$a> operation can be expressed in terms of C<$a+=1> or C<$a+1>, | |
675 | and C<$a--> in terms of C<$a-=1> and C<$a-1>. | |
676 | ||
677 | =item C<abs($a)> | |
678 | ||
679 | can be expressed in terms of C<$aE<lt>0> and C<-$a> (or C<0-$a>). | |
680 | ||
681 | =item I<Unary minus> | |
682 | ||
683 | can be expressed in terms of subtraction. | |
684 | ||
3bc6ec80 | 685 | =item I<Negation> |
686 | ||
687 | C<!> and C<not> can be expressed in terms of boolean conversion, or | |
688 | string or numerical conversion. | |
689 | ||
4633a7c4 LW |
690 | =item I<Concatenation> |
691 | ||
692 | can be expressed in terms of string conversion. | |
693 | ||
b267980d | 694 | =item I<Comparison operations> |
4633a7c4 LW |
695 | |
696 | can be expressed in terms of its "spaceship" counterpart: either | |
697 | C<E<lt>=E<gt>> or C<cmp>: | |
1fef88e7 | 698 | |
4633a7c4 LW |
699 | <, >, <=, >=, ==, != in terms of <=> |
700 | lt, gt, le, ge, eq, ne in terms of cmp | |
701 | ||
f5284f61 IZ |
702 | =item I<Iterator> |
703 | ||
704 | <> in terms of builtin operations | |
705 | ||
706 | =item I<Dereferencing> | |
707 | ||
708 | ${} @{} %{} &{} *{} in terms of builtin operations | |
709 | ||
4633a7c4 LW |
710 | =item I<Copy operator> |
711 | ||
712 | can be expressed in terms of an assignment to the dereferenced value, if this | |
49c95d58 RD |
713 | value is a scalar and not a reference, or simply a reference assignment |
714 | otherwise. | |
4633a7c4 LW |
715 | |
716 | =back | |
717 | ||
84fc275b S |
718 | =head1 Minimal set of overloaded operations |
719 | ||
720 | Since some operations can be automatically generated from others, there is | |
721 | a minimal set of operations that need to be overloaded in order to have | |
299476e0 S |
722 | the complete set of overloaded operations at one's disposal. |
723 | Of course, the autogenerated operations may not do exactly what the user | |
724 | expects. See L<MAGIC AUTOGENERATION> above. The minimal set is: | |
84fc275b S |
725 | |
726 | + - * / % ** << >> x | |
727 | <=> cmp | |
728 | & | ^ ~ | |
729 | atan2 cos sin exp log sqrt int | |
730 | ||
731 | Additionally, you need to define at least one of string, boolean or | |
299476e0 S |
732 | numeric conversions because any one can be used to emulate the others. |
733 | The string conversion can also be used to emulate concatenation. | |
84fc275b | 734 | |
ee239bfe | 735 | =head1 Losing overloading |
4633a7c4 LW |
736 | |
737 | The restriction for the comparison operation is that even if, for example, | |
738 | `C<cmp>' should return a blessed reference, the autogenerated `C<lt>' | |
739 | function will produce only a standard logical value based on the | |
740 | numerical value of the result of `C<cmp>'. In particular, a working | |
741 | numeric conversion is needed in this case (possibly expressed in terms of | |
742 | other conversions). | |
743 | ||
744 | Similarly, C<.=> and C<x=> operators lose their mathemagical properties | |
745 | if the string conversion substitution is applied. | |
746 | ||
747 | When you chop() a mathemagical object it is promoted to a string and its | |
748 | mathemagical properties are lost. The same can happen with other | |
749 | operations as well. | |
750 | ||
751 | =head1 Run-time Overloading | |
752 | ||
753 | Since all C<use> directives are executed at compile-time, the only way to | |
754 | change overloading during run-time is to | |
755 | ||
756 | eval 'use overload "+" => \&addmethod'; | |
757 | ||
758 | You can also use | |
759 | ||
760 | eval 'no overload "+", "--", "<="'; | |
761 | ||
762 | though the use of these constructs during run-time is questionable. | |
763 | ||
764 | =head1 Public functions | |
765 | ||
766 | Package C<overload.pm> provides the following public functions: | |
767 | ||
768 | =over 5 | |
769 | ||
770 | =item overload::StrVal(arg) | |
771 | ||
6a0e9e72 FD |
772 | Gives string value of C<arg> as in absence of stringify overloading. If you |
773 | are using this to get the address of a reference (useful for checking if two | |
774 | references point to the same thing) then you may be better off using | |
775 | C<Scalar::Util::refaddr()>, which is faster. | |
4633a7c4 LW |
776 | |
777 | =item overload::Overloaded(arg) | |
778 | ||
779 | Returns true if C<arg> is subject to overloading of some operations. | |
780 | ||
781 | =item overload::Method(obj,op) | |
782 | ||
783 | Returns C<undef> or a reference to the method that implements C<op>. | |
784 | ||
785 | =back | |
786 | ||
b3ac6de7 IZ |
787 | =head1 Overloading constants |
788 | ||
7adf7a02 | 789 | For some applications, the Perl parser mangles constants too much. |
bfce84ec | 790 | It is possible to hook into this process via C<overload::constant()> |
7adf7a02 | 791 | and C<overload::remove_constant()> functions. |
b3ac6de7 IZ |
792 | |
793 | These functions take a hash as an argument. The recognized keys of this hash | |
7adf7a02 | 794 | are: |
b3ac6de7 IZ |
795 | |
796 | =over 8 | |
797 | ||
798 | =item integer | |
799 | ||
800 | to overload integer constants, | |
801 | ||
802 | =item float | |
803 | ||
804 | to overload floating point constants, | |
805 | ||
806 | =item binary | |
807 | ||
808 | to overload octal and hexadecimal constants, | |
809 | ||
810 | =item q | |
811 | ||
812 | to overload C<q>-quoted strings, constant pieces of C<qq>- and C<qx>-quoted | |
813 | strings and here-documents, | |
814 | ||
815 | =item qr | |
816 | ||
817 | to overload constant pieces of regular expressions. | |
818 | ||
819 | =back | |
820 | ||
821 | The corresponding values are references to functions which take three arguments: | |
822 | the first one is the I<initial> string form of the constant, the second one | |
b267980d | 823 | is how Perl interprets this constant, the third one is how the constant is used. |
b3ac6de7 | 824 | Note that the initial string form does not |
b267980d | 825 | contain string delimiters, and has backslashes in backslash-delimiter |
b3ac6de7 | 826 | combinations stripped (thus the value of delimiter is not relevant for |
b267980d | 827 | processing of this string). The return value of this function is how this |
b3ac6de7 IZ |
828 | constant is going to be interpreted by Perl. The third argument is undefined |
829 | unless for overloaded C<q>- and C<qr>- constants, it is C<q> in single-quote | |
830 | context (comes from strings, regular expressions, and single-quote HERE | |
b267980d | 831 | documents), it is C<tr> for arguments of C<tr>/C<y> operators, |
b3ac6de7 IZ |
832 | it is C<s> for right-hand side of C<s>-operator, and it is C<qq> otherwise. |
833 | ||
834 | Since an expression C<"ab$cd,,"> is just a shortcut for C<'ab' . $cd . ',,'>, | |
835 | it is expected that overloaded constant strings are equipped with reasonable | |
b267980d | 836 | overloaded catenation operator, otherwise absurd results will result. |
b3ac6de7 IZ |
837 | Similarly, negative numbers are considered as negations of positive constants. |
838 | ||
839 | Note that it is probably meaningless to call the functions overload::constant() | |
840 | and overload::remove_constant() from anywhere but import() and unimport() methods. | |
841 | From these methods they may be called as | |
842 | ||
843 | sub import { | |
844 | shift; | |
845 | return unless @_; | |
846 | die "unknown import: @_" unless @_ == 1 and $_[0] eq ':constant'; | |
847 | overload::constant integer => sub {Math::BigInt->new(shift)}; | |
848 | } | |
849 | ||
4633a7c4 LW |
850 | =head1 IMPLEMENTATION |
851 | ||
852 | What follows is subject to change RSN. | |
853 | ||
e7ea3e70 IZ |
854 | The table of methods for all operations is cached in magic for the |
855 | symbol table hash for the package. The cache is invalidated during | |
856 | processing of C<use overload>, C<no overload>, new function | |
857 | definitions, and changes in @ISA. However, this invalidation remains | |
858 | unprocessed until the next C<bless>ing into the package. Hence if you | |
859 | want to change overloading structure dynamically, you'll need an | |
860 | additional (fake) C<bless>ing to update the table. | |
861 | ||
862 | (Every SVish thing has a magic queue, and magic is an entry in that | |
863 | queue. This is how a single variable may participate in multiple | |
864 | forms of magic simultaneously. For instance, environment variables | |
865 | regularly have two forms at once: their %ENV magic and their taint | |
866 | magic. However, the magic which implements overloading is applied to | |
867 | the stashes, which are rarely used directly, thus should not slow down | |
868 | Perl.) | |
4633a7c4 LW |
869 | |
870 | If an object belongs to a package using overload, it carries a special | |
871 | flag. Thus the only speed penalty during arithmetic operations without | |
872 | overloading is the checking of this flag. | |
873 | ||
774d564b | 874 | In fact, if C<use overload> is not present, there is almost no overhead |
875 | for overloadable operations, so most programs should not suffer | |
876 | measurable performance penalties. A considerable effort was made to | |
877 | minimize the overhead when overload is used in some package, but the | |
878 | arguments in question do not belong to packages using overload. When | |
879 | in doubt, test your speed with C<use overload> and without it. So far | |
880 | there have been no reports of substantial speed degradation if Perl is | |
881 | compiled with optimization turned on. | |
4633a7c4 | 882 | |
e7ea3e70 IZ |
883 | There is no size penalty for data if overload is not used. The only |
884 | size penalty if overload is used in some package is that I<all> the | |
885 | packages acquire a magic during the next C<bless>ing into the | |
886 | package. This magic is three-words-long for packages without | |
f610777f | 887 | overloading, and carries the cache table if the package is overloaded. |
4633a7c4 | 888 | |
b267980d | 889 | Copying (C<$a=$b>) is shallow; however, a one-level-deep copying is |
4633a7c4 LW |
890 | carried out before any operation that can imply an assignment to the |
891 | object $a (or $b) refers to, like C<$a++>. You can override this | |
892 | behavior by defining your own copy constructor (see L<"Copy Constructor">). | |
893 | ||
894 | It is expected that arguments to methods that are not explicitly supposed | |
895 | to be changed are constant (but this is not enforced). | |
896 | ||
ee239bfe IZ |
897 | =head1 Metaphor clash |
898 | ||
f610777f | 899 | One may wonder why the semantic of overloaded C<=> is so counter intuitive. |
b267980d NIS |
900 | If it I<looks> counter intuitive to you, you are subject to a metaphor |
901 | clash. | |
ee239bfe IZ |
902 | |
903 | Here is a Perl object metaphor: | |
904 | ||
905 | I< object is a reference to blessed data> | |
906 | ||
907 | and an arithmetic metaphor: | |
908 | ||
909 | I< object is a thing by itself>. | |
910 | ||
911 | The I<main> problem of overloading C<=> is the fact that these metaphors | |
912 | imply different actions on the assignment C<$a = $b> if $a and $b are | |
913 | objects. Perl-think implies that $a becomes a reference to whatever | |
914 | $b was referencing. Arithmetic-think implies that the value of "object" | |
915 | $a is changed to become the value of the object $b, preserving the fact | |
916 | that $a and $b are separate entities. | |
917 | ||
918 | The difference is not relevant in the absence of mutators. After | |
919 | a Perl-way assignment an operation which mutates the data referenced by $a | |
b267980d | 920 | would change the data referenced by $b too. Effectively, after |
ee239bfe IZ |
921 | C<$a = $b> values of $a and $b become I<indistinguishable>. |
922 | ||
b267980d | 923 | On the other hand, anyone who has used algebraic notation knows the |
ee239bfe IZ |
924 | expressive power of the arithmetic metaphor. Overloading works hard |
925 | to enable this metaphor while preserving the Perlian way as far as | |
d1be9408 | 926 | possible. Since it is not possible to freely mix two contradicting |
ee239bfe IZ |
927 | metaphors, overloading allows the arithmetic way to write things I<as |
928 | far as all the mutators are called via overloaded access only>. The | |
929 | way it is done is described in L<Copy Constructor>. | |
930 | ||
931 | If some mutator methods are directly applied to the overloaded values, | |
b267980d | 932 | one may need to I<explicitly unlink> other values which references the |
ee239bfe IZ |
933 | same value: |
934 | ||
2b393bf4 | 935 | $a = Data->new(23); |
ee239bfe IZ |
936 | ... |
937 | $b = $a; # $b is "linked" to $a | |
938 | ... | |
939 | $a = $a->clone; # Unlink $b from $a | |
940 | $a->increment_by(4); | |
941 | ||
942 | Note that overloaded access makes this transparent: | |
943 | ||
2b393bf4 | 944 | $a = Data->new(23); |
ee239bfe IZ |
945 | $b = $a; # $b is "linked" to $a |
946 | $a += 4; # would unlink $b automagically | |
947 | ||
948 | However, it would not make | |
949 | ||
2b393bf4 | 950 | $a = Data->new(23); |
ee239bfe IZ |
951 | $a = 4; # Now $a is a plain 4, not 'Data' |
952 | ||
953 | preserve "objectness" of $a. But Perl I<has> a way to make assignments | |
954 | to an object do whatever you want. It is just not the overload, but | |
955 | tie()ing interface (see L<perlfunc/tie>). Adding a FETCH() method | |
b267980d | 956 | which returns the object itself, and STORE() method which changes the |
ee239bfe IZ |
957 | value of the object, one can reproduce the arithmetic metaphor in its |
958 | completeness, at least for variables which were tie()d from the start. | |
959 | ||
960 | (Note that a workaround for a bug may be needed, see L<"BUGS">.) | |
961 | ||
962 | =head1 Cookbook | |
963 | ||
964 | Please add examples to what follows! | |
965 | ||
966 | =head2 Two-face scalars | |
967 | ||
968 | Put this in F<two_face.pm> in your Perl library directory: | |
969 | ||
970 | package two_face; # Scalars with separate string and | |
971 | # numeric values. | |
972 | sub new { my $p = shift; bless [@_], $p } | |
973 | use overload '""' => \&str, '0+' => \&num, fallback => 1; | |
974 | sub num {shift->[1]} | |
975 | sub str {shift->[0]} | |
976 | ||
977 | Use it as follows: | |
978 | ||
979 | require two_face; | |
2b393bf4 | 980 | my $seven = two_face->new("vii", 7); |
ee239bfe IZ |
981 | printf "seven=$seven, seven=%d, eight=%d\n", $seven, $seven+1; |
982 | print "seven contains `i'\n" if $seven =~ /i/; | |
983 | ||
984 | (The second line creates a scalar which has both a string value, and a | |
985 | numeric value.) This prints: | |
986 | ||
987 | seven=vii, seven=7, eight=8 | |
988 | seven contains `i' | |
989 | ||
f5284f61 IZ |
990 | =head2 Two-face references |
991 | ||
992 | Suppose you want to create an object which is accessible as both an | |
6d822dc4 | 993 | array reference and a hash reference. |
f5284f61 IZ |
994 | |
995 | package two_refs; | |
996 | use overload '%{}' => \&gethash, '@{}' => sub { $ {shift()} }; | |
b267980d NIS |
997 | sub new { |
998 | my $p = shift; | |
f5284f61 IZ |
999 | bless \ [@_], $p; |
1000 | } | |
1001 | sub gethash { | |
1002 | my %h; | |
1003 | my $self = shift; | |
1004 | tie %h, ref $self, $self; | |
1005 | \%h; | |
1006 | } | |
1007 | ||
1008 | sub TIEHASH { my $p = shift; bless \ shift, $p } | |
1009 | my %fields; | |
1010 | my $i = 0; | |
1011 | $fields{$_} = $i++ foreach qw{zero one two three}; | |
b267980d | 1012 | sub STORE { |
f5284f61 IZ |
1013 | my $self = ${shift()}; |
1014 | my $key = $fields{shift()}; | |
1015 | defined $key or die "Out of band access"; | |
1016 | $$self->[$key] = shift; | |
1017 | } | |
b267980d | 1018 | sub FETCH { |
f5284f61 IZ |
1019 | my $self = ${shift()}; |
1020 | my $key = $fields{shift()}; | |
1021 | defined $key or die "Out of band access"; | |
1022 | $$self->[$key]; | |
1023 | } | |
1024 | ||
1025 | Now one can access an object using both the array and hash syntax: | |
1026 | ||
2b393bf4 | 1027 | my $bar = two_refs->new(3,4,5,6); |
f5284f61 IZ |
1028 | $bar->[2] = 11; |
1029 | $bar->{two} == 11 or die 'bad hash fetch'; | |
1030 | ||
1031 | Note several important features of this example. First of all, the | |
1032 | I<actual> type of $bar is a scalar reference, and we do not overload | |
1033 | the scalar dereference. Thus we can get the I<actual> non-overloaded | |
1034 | contents of $bar by just using C<$$bar> (what we do in functions which | |
1035 | overload dereference). Similarly, the object returned by the | |
1036 | TIEHASH() method is a scalar reference. | |
1037 | ||
1038 | Second, we create a new tied hash each time the hash syntax is used. | |
1039 | This allows us not to worry about a possibility of a reference loop, | |
d1be9408 | 1040 | which would lead to a memory leak. |
f5284f61 IZ |
1041 | |
1042 | Both these problems can be cured. Say, if we want to overload hash | |
1043 | dereference on a reference to an object which is I<implemented> as a | |
1044 | hash itself, the only problem one has to circumvent is how to access | |
1fd16925 | 1045 | this I<actual> hash (as opposed to the I<virtual> hash exhibited by the |
f5284f61 IZ |
1046 | overloaded dereference operator). Here is one possible fetching routine: |
1047 | ||
1048 | sub access_hash { | |
1049 | my ($self, $key) = (shift, shift); | |
1050 | my $class = ref $self; | |
b267980d | 1051 | bless $self, 'overload::dummy'; # Disable overloading of %{} |
f5284f61 IZ |
1052 | my $out = $self->{$key}; |
1053 | bless $self, $class; # Restore overloading | |
1054 | $out; | |
1055 | } | |
1056 | ||
1fd16925 | 1057 | To remove creation of the tied hash on each access, one may an extra |
f5284f61 IZ |
1058 | level of indirection which allows a non-circular structure of references: |
1059 | ||
1060 | package two_refs1; | |
1061 | use overload '%{}' => sub { ${shift()}->[1] }, | |
1062 | '@{}' => sub { ${shift()}->[0] }; | |
b267980d NIS |
1063 | sub new { |
1064 | my $p = shift; | |
f5284f61 IZ |
1065 | my $a = [@_]; |
1066 | my %h; | |
1067 | tie %h, $p, $a; | |
1068 | bless \ [$a, \%h], $p; | |
1069 | } | |
1070 | sub gethash { | |
1071 | my %h; | |
1072 | my $self = shift; | |
1073 | tie %h, ref $self, $self; | |
1074 | \%h; | |
1075 | } | |
1076 | ||
1077 | sub TIEHASH { my $p = shift; bless \ shift, $p } | |
1078 | my %fields; | |
1079 | my $i = 0; | |
1080 | $fields{$_} = $i++ foreach qw{zero one two three}; | |
b267980d | 1081 | sub STORE { |
f5284f61 IZ |
1082 | my $a = ${shift()}; |
1083 | my $key = $fields{shift()}; | |
1084 | defined $key or die "Out of band access"; | |
1085 | $a->[$key] = shift; | |
1086 | } | |
b267980d | 1087 | sub FETCH { |
f5284f61 IZ |
1088 | my $a = ${shift()}; |
1089 | my $key = $fields{shift()}; | |
1090 | defined $key or die "Out of band access"; | |
1091 | $a->[$key]; | |
1092 | } | |
1093 | ||
1fd16925 | 1094 | Now if $baz is overloaded like this, then C<$baz> is a reference to a |
f5284f61 IZ |
1095 | reference to the intermediate array, which keeps a reference to an |
1096 | actual array, and the access hash. The tie()ing object for the access | |
1fd16925 | 1097 | hash is a reference to a reference to the actual array, so |
f5284f61 | 1098 | |
88c28ceb | 1099 | =over |
f5284f61 IZ |
1100 | |
1101 | =item * | |
1102 | ||
1103 | There are no loops of references. | |
1104 | ||
1105 | =item * | |
1106 | ||
1107 | Both "objects" which are blessed into the class C<two_refs1> are | |
1108 | references to a reference to an array, thus references to a I<scalar>. | |
1109 | Thus the accessor expression C<$$foo-E<gt>[$ind]> involves no | |
1110 | overloaded operations. | |
1111 | ||
1112 | =back | |
1113 | ||
ee239bfe IZ |
1114 | =head2 Symbolic calculator |
1115 | ||
1116 | Put this in F<symbolic.pm> in your Perl library directory: | |
1117 | ||
1118 | package symbolic; # Primitive symbolic calculator | |
1119 | use overload nomethod => \&wrap; | |
1120 | ||
1121 | sub new { shift; bless ['n', @_] } | |
1122 | sub wrap { | |
1123 | my ($obj, $other, $inv, $meth) = @_; | |
1124 | ($obj, $other) = ($other, $obj) if $inv; | |
1125 | bless [$meth, $obj, $other]; | |
1126 | } | |
1127 | ||
1128 | This module is very unusual as overloaded modules go: it does not | |
88c28ceb JH |
1129 | provide any usual overloaded operators, instead it provides the L<Last |
1130 | Resort> operator C<nomethod>. In this example the corresponding | |
f610777f | 1131 | subroutine returns an object which encapsulates operations done over |
2b393bf4 RS |
1132 | the objects: C<< symbolic->new(3) >> contains C<['n', 3]>, C<< 2 + |
1133 | symbolic->new(3) >> contains C<['+', 2, ['n', 3]]>. | |
ee239bfe IZ |
1134 | |
1135 | Here is an example of the script which "calculates" the side of | |
1136 | circumscribed octagon using the above package: | |
1137 | ||
1138 | require symbolic; | |
1139 | my $iter = 1; # 2**($iter+2) = 8 | |
2b393bf4 | 1140 | my $side = symbolic->new(1); |
ee239bfe | 1141 | my $cnt = $iter; |
3cb6de81 | 1142 | |
ee239bfe IZ |
1143 | while ($cnt--) { |
1144 | $side = (sqrt(1 + $side**2) - 1)/$side; | |
1145 | } | |
1146 | print "OK\n"; | |
1147 | ||
1148 | The value of $side is | |
1149 | ||
1150 | ['/', ['-', ['sqrt', ['+', 1, ['**', ['n', 1], 2]], | |
1151 | undef], 1], ['n', 1]] | |
1152 | ||
1153 | Note that while we obtained this value using a nice little script, | |
1154 | there is no simple way to I<use> this value. In fact this value may | |
2d3232d7 | 1155 | be inspected in debugger (see L<perldebug>), but only if |
ee239bfe IZ |
1156 | C<bareStringify> B<O>ption is set, and not via C<p> command. |
1157 | ||
1158 | If one attempts to print this value, then the overloaded operator | |
1159 | C<""> will be called, which will call C<nomethod> operator. The | |
1160 | result of this operator will be stringified again, but this result is | |
1161 | again of type C<symbolic>, which will lead to an infinite loop. | |
1162 | ||
1163 | Add a pretty-printer method to the module F<symbolic.pm>: | |
1164 | ||
1165 | sub pretty { | |
1166 | my ($meth, $a, $b) = @{+shift}; | |
1167 | $a = 'u' unless defined $a; | |
1168 | $b = 'u' unless defined $b; | |
1169 | $a = $a->pretty if ref $a; | |
1170 | $b = $b->pretty if ref $b; | |
1171 | "[$meth $a $b]"; | |
b267980d | 1172 | } |
ee239bfe IZ |
1173 | |
1174 | Now one can finish the script by | |
1175 | ||
1176 | print "side = ", $side->pretty, "\n"; | |
1177 | ||
1178 | The method C<pretty> is doing object-to-string conversion, so it | |
1179 | is natural to overload the operator C<""> using this method. However, | |
1180 | inside such a method it is not necessary to pretty-print the | |
1181 | I<components> $a and $b of an object. In the above subroutine | |
1182 | C<"[$meth $a $b]"> is a catenation of some strings and components $a | |
1183 | and $b. If these components use overloading, the catenation operator | |
1fd16925 | 1184 | will look for an overloaded operator C<.>; if not present, it will |
ee239bfe IZ |
1185 | look for an overloaded operator C<"">. Thus it is enough to use |
1186 | ||
1187 | use overload nomethod => \&wrap, '""' => \&str; | |
1188 | sub str { | |
1189 | my ($meth, $a, $b) = @{+shift}; | |
1190 | $a = 'u' unless defined $a; | |
1191 | $b = 'u' unless defined $b; | |
1192 | "[$meth $a $b]"; | |
b267980d | 1193 | } |
ee239bfe IZ |
1194 | |
1195 | Now one can change the last line of the script to | |
1196 | ||
1197 | print "side = $side\n"; | |
1198 | ||
1199 | which outputs | |
1200 | ||
1201 | side = [/ [- [sqrt [+ 1 [** [n 1 u] 2]] u] 1] [n 1 u]] | |
1202 | ||
1203 | and one can inspect the value in debugger using all the possible | |
b267980d | 1204 | methods. |
ee239bfe | 1205 | |
d1be9408 | 1206 | Something is still amiss: consider the loop variable $cnt of the |
ee239bfe IZ |
1207 | script. It was a number, not an object. We cannot make this value of |
1208 | type C<symbolic>, since then the loop will not terminate. | |
1209 | ||
1210 | Indeed, to terminate the cycle, the $cnt should become false. | |
1211 | However, the operator C<bool> for checking falsity is overloaded (this | |
1212 | time via overloaded C<"">), and returns a long string, thus any object | |
1213 | of type C<symbolic> is true. To overcome this, we need a way to | |
1214 | compare an object to 0. In fact, it is easier to write a numeric | |
1215 | conversion routine. | |
1216 | ||
1217 | Here is the text of F<symbolic.pm> with such a routine added (and | |
f610777f | 1218 | slightly modified str()): |
ee239bfe IZ |
1219 | |
1220 | package symbolic; # Primitive symbolic calculator | |
1221 | use overload | |
1222 | nomethod => \&wrap, '""' => \&str, '0+' => \# | |
1223 | ||
1224 | sub new { shift; bless ['n', @_] } | |
1225 | sub wrap { | |
1226 | my ($obj, $other, $inv, $meth) = @_; | |
1227 | ($obj, $other) = ($other, $obj) if $inv; | |
1228 | bless [$meth, $obj, $other]; | |
1229 | } | |
1230 | sub str { | |
1231 | my ($meth, $a, $b) = @{+shift}; | |
1232 | $a = 'u' unless defined $a; | |
1233 | if (defined $b) { | |
1234 | "[$meth $a $b]"; | |
1235 | } else { | |
1236 | "[$meth $a]"; | |
1237 | } | |
b267980d NIS |
1238 | } |
1239 | my %subr = ( n => sub {$_[0]}, | |
1240 | sqrt => sub {sqrt $_[0]}, | |
ee239bfe IZ |
1241 | '-' => sub {shift() - shift()}, |
1242 | '+' => sub {shift() + shift()}, | |
1243 | '/' => sub {shift() / shift()}, | |
1244 | '*' => sub {shift() * shift()}, | |
1245 | '**' => sub {shift() ** shift()}, | |
1246 | ); | |
1247 | sub num { | |
1248 | my ($meth, $a, $b) = @{+shift}; | |
b267980d | 1249 | my $subr = $subr{$meth} |
ee239bfe IZ |
1250 | or die "Do not know how to ($meth) in symbolic"; |
1251 | $a = $a->num if ref $a eq __PACKAGE__; | |
1252 | $b = $b->num if ref $b eq __PACKAGE__; | |
1253 | $subr->($a,$b); | |
1254 | } | |
1255 | ||
1256 | All the work of numeric conversion is done in %subr and num(). Of | |
f610777f | 1257 | course, %subr is not complete, it contains only operators used in the |
ee239bfe IZ |
1258 | example below. Here is the extra-credit question: why do we need an |
1259 | explicit recursion in num()? (Answer is at the end of this section.) | |
1260 | ||
1261 | Use this module like this: | |
1262 | ||
1263 | require symbolic; | |
2b393bf4 RS |
1264 | my $iter = symbolic->new(2); # 16-gon |
1265 | my $side = symbolic->new(1); | |
ee239bfe | 1266 | my $cnt = $iter; |
3cb6de81 | 1267 | |
ee239bfe IZ |
1268 | while ($cnt) { |
1269 | $cnt = $cnt - 1; # Mutator `--' not implemented | |
1270 | $side = (sqrt(1 + $side**2) - 1)/$side; | |
1271 | } | |
1272 | printf "%s=%f\n", $side, $side; | |
1273 | printf "pi=%f\n", $side*(2**($iter+2)); | |
1274 | ||
1275 | It prints (without so many line breaks) | |
1276 | ||
1277 | [/ [- [sqrt [+ 1 [** [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] | |
1278 | [n 1]] 2]]] 1] | |
1279 | [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] [n 1]]]=0.198912 | |
1280 | pi=3.182598 | |
1281 | ||
1282 | The above module is very primitive. It does not implement | |
1283 | mutator methods (C<++>, C<-=> and so on), does not do deep copying | |
1284 | (not required without mutators!), and implements only those arithmetic | |
1285 | operations which are used in the example. | |
1286 | ||
1fd16925 | 1287 | To implement most arithmetic operations is easy; one should just use |
ee239bfe IZ |
1288 | the tables of operations, and change the code which fills %subr to |
1289 | ||
1290 | my %subr = ( 'n' => sub {$_[0]} ); | |
1291 | foreach my $op (split " ", $overload::ops{with_assign}) { | |
1292 | $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}"; | |
1293 | } | |
1294 | my @bins = qw(binary 3way_comparison num_comparison str_comparison); | |
1295 | foreach my $op (split " ", "@overload::ops{ @bins }") { | |
1296 | $subr{$op} = eval "sub {shift() $op shift()}"; | |
1297 | } | |
1298 | foreach my $op (split " ", "@overload::ops{qw(unary func)}") { | |
1299 | print "defining `$op'\n"; | |
1300 | $subr{$op} = eval "sub {$op shift()}"; | |
1301 | } | |
1302 | ||
1303 | Due to L<Calling Conventions for Mutators>, we do not need anything | |
1304 | special to make C<+=> and friends work, except filling C<+=> entry of | |
1305 | %subr, and defining a copy constructor (needed since Perl has no | |
1306 | way to know that the implementation of C<'+='> does not mutate | |
1307 | the argument, compare L<Copy Constructor>). | |
1308 | ||
1fd16925 | 1309 | To implement a copy constructor, add C<< '=' => \&cpy >> to C<use overload> |
ee239bfe IZ |
1310 | line, and code (this code assumes that mutators change things one level |
1311 | deep only, so recursive copying is not needed): | |
1312 | ||
1313 | sub cpy { | |
1314 | my $self = shift; | |
1315 | bless [@$self], ref $self; | |
1316 | } | |
1317 | ||
b267980d | 1318 | To make C<++> and C<--> work, we need to implement actual mutators, |
ee239bfe IZ |
1319 | either directly, or in C<nomethod>. We continue to do things inside |
1320 | C<nomethod>, thus add | |
1321 | ||
1322 | if ($meth eq '++' or $meth eq '--') { | |
1323 | @$obj = ($meth, (bless [@$obj]), 1); # Avoid circular reference | |
1324 | return $obj; | |
1325 | } | |
1326 | ||
b267980d | 1327 | after the first line of wrap(). This is not a most effective |
ee239bfe IZ |
1328 | implementation, one may consider |
1329 | ||
1330 | sub inc { $_[0] = bless ['++', shift, 1]; } | |
1331 | ||
1332 | instead. | |
1333 | ||
1334 | As a final remark, note that one can fill %subr by | |
1335 | ||
1336 | my %subr = ( 'n' => sub {$_[0]} ); | |
1337 | foreach my $op (split " ", $overload::ops{with_assign}) { | |
1338 | $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}"; | |
1339 | } | |
1340 | my @bins = qw(binary 3way_comparison num_comparison str_comparison); | |
1341 | foreach my $op (split " ", "@overload::ops{ @bins }") { | |
1342 | $subr{$op} = eval "sub {shift() $op shift()}"; | |
1343 | } | |
1344 | foreach my $op (split " ", "@overload::ops{qw(unary func)}") { | |
1345 | $subr{$op} = eval "sub {$op shift()}"; | |
1346 | } | |
1347 | $subr{'++'} = $subr{'+'}; | |
1348 | $subr{'--'} = $subr{'-'}; | |
1349 | ||
b267980d NIS |
1350 | This finishes implementation of a primitive symbolic calculator in |
1351 | 50 lines of Perl code. Since the numeric values of subexpressions | |
ee239bfe IZ |
1352 | are not cached, the calculator is very slow. |
1353 | ||
1354 | Here is the answer for the exercise: In the case of str(), we need no | |
1355 | explicit recursion since the overloaded C<.>-operator will fall back | |
1356 | to an existing overloaded operator C<"">. Overloaded arithmetic | |
1357 | operators I<do not> fall back to numeric conversion if C<fallback> is | |
1358 | not explicitly requested. Thus without an explicit recursion num() | |
1359 | would convert C<['+', $a, $b]> to C<$a + $b>, which would just rebuild | |
1360 | the argument of num(). | |
1361 | ||
1362 | If you wonder why defaults for conversion are different for str() and | |
1363 | num(), note how easy it was to write the symbolic calculator. This | |
1364 | simplicity is due to an appropriate choice of defaults. One extra | |
f610777f A |
1365 | note: due to the explicit recursion num() is more fragile than sym(): |
1366 | we need to explicitly check for the type of $a and $b. If components | |
ee239bfe IZ |
1367 | $a and $b happen to be of some related type, this may lead to problems. |
1368 | ||
1369 | =head2 I<Really> symbolic calculator | |
1370 | ||
1371 | One may wonder why we call the above calculator symbolic. The reason | |
1372 | is that the actual calculation of the value of expression is postponed | |
1373 | until the value is I<used>. | |
1374 | ||
1375 | To see it in action, add a method | |
1376 | ||
b267980d NIS |
1377 | sub STORE { |
1378 | my $obj = shift; | |
1379 | $#$obj = 1; | |
ee239bfe IZ |
1380 | @$obj->[0,1] = ('=', shift); |
1381 | } | |
1382 | ||
1383 | to the package C<symbolic>. After this change one can do | |
1384 | ||
2b393bf4 RS |
1385 | my $a = symbolic->new(3); |
1386 | my $b = symbolic->new(4); | |
ee239bfe IZ |
1387 | my $c = sqrt($a**2 + $b**2); |
1388 | ||
1389 | and the numeric value of $c becomes 5. However, after calling | |
1390 | ||
1391 | $a->STORE(12); $b->STORE(5); | |
1392 | ||
1393 | the numeric value of $c becomes 13. There is no doubt now that the module | |
1394 | symbolic provides a I<symbolic> calculator indeed. | |
1395 | ||
1396 | To hide the rough edges under the hood, provide a tie()d interface to the | |
1397 | package C<symbolic> (compare with L<Metaphor clash>). Add methods | |
1398 | ||
1399 | sub TIESCALAR { my $pack = shift; $pack->new(@_) } | |
1400 | sub FETCH { shift } | |
1401 | sub nop { } # Around a bug | |
1402 | ||
1403 | (the bug is described in L<"BUGS">). One can use this new interface as | |
1404 | ||
1405 | tie $a, 'symbolic', 3; | |
1406 | tie $b, 'symbolic', 4; | |
1407 | $a->nop; $b->nop; # Around a bug | |
1408 | ||
1409 | my $c = sqrt($a**2 + $b**2); | |
1410 | ||
1411 | Now numeric value of $c is 5. After C<$a = 12; $b = 5> the numeric value | |
1412 | of $c becomes 13. To insulate the user of the module add a method | |
1413 | ||
1414 | sub vars { my $p = shift; tie($_, $p), $_->nop foreach @_; } | |
1415 | ||
1416 | Now | |
1417 | ||
1418 | my ($a, $b); | |
1419 | symbolic->vars($a, $b); | |
1420 | my $c = sqrt($a**2 + $b**2); | |
1421 | ||
1422 | $a = 3; $b = 4; | |
1423 | printf "c5 %s=%f\n", $c, $c; | |
1424 | ||
1425 | $a = 12; $b = 5; | |
1426 | printf "c13 %s=%f\n", $c, $c; | |
1427 | ||
1428 | shows that the numeric value of $c follows changes to the values of $a | |
1429 | and $b. | |
1430 | ||
4633a7c4 LW |
1431 | =head1 AUTHOR |
1432 | ||
1fef88e7 | 1433 | Ilya Zakharevich E<lt>F<ilya@math.mps.ohio-state.edu>E<gt>. |
4633a7c4 | 1434 | |
7e494759 PF |
1435 | =head1 SEE ALSO |
1436 | ||
1437 | The L<overloading> pragma can be used to enable or disable overloaded | |
1438 | operations within a lexical scope. | |
1439 | ||
4633a7c4 LW |
1440 | =head1 DIAGNOSTICS |
1441 | ||
1442 | When Perl is run with the B<-Do> switch or its equivalent, overloading | |
1443 | induces diagnostic messages. | |
1444 | ||
e7ea3e70 IZ |
1445 | Using the C<m> command of Perl debugger (see L<perldebug>) one can |
1446 | deduce which operations are overloaded (and which ancestor triggers | |
1447 | this overloading). Say, if C<eq> is overloaded, then the method C<(eq> | |
1448 | is shown by debugger. The method C<()> corresponds to the C<fallback> | |
1449 | key (in fact a presence of this method shows that this package has | |
1450 | overloading enabled, and it is what is used by the C<Overloaded> | |
ee239bfe | 1451 | function of module C<overload>). |
e7ea3e70 | 1452 | |
6ad11d81 | 1453 | The module might issue the following warnings: |
6b82e2f5 A |
1454 | |
1455 | =over 4 | |
1456 | ||
1457 | =item Odd number of arguments for overload::constant | |
1458 | ||
1459 | (W) The call to overload::constant contained an odd number of arguments. | |
1460 | The arguments should come in pairs. | |
1461 | ||
1462 | =item `%s' is not an overloadable type | |
1463 | ||
1464 | (W) You tried to overload a constant type the overload package is unaware of. | |
1465 | ||
1466 | =item `%s' is not a code reference | |
1467 | ||
1468 | (W) The second (fourth, sixth, ...) argument of overload::constant needs | |
1469 | to be a code reference. Either an anonymous subroutine, or a reference | |
1470 | to a subroutine. | |
1471 | ||
1472 | =back | |
1473 | ||
4633a7c4 LW |
1474 | =head1 BUGS |
1475 | ||
aa689395 | 1476 | Because it is used for overloading, the per-package hash %OVERLOAD now |
1477 | has a special meaning in Perl. The symbol table is filled with names | |
1478 | looking like line-noise. | |
4633a7c4 | 1479 | |
a6006777 | 1480 | For the purpose of inheritance every overloaded package behaves as if |
1481 | C<fallback> is present (possibly undefined). This may create | |
1482 | interesting effects if some package is not overloaded, but inherits | |
1483 | from two overloaded packages. | |
4633a7c4 | 1484 | |
b267980d | 1485 | Relation between overloading and tie()ing is broken. Overloading is |
ee239bfe IZ |
1486 | triggered or not basing on the I<previous> class of tie()d value. |
1487 | ||
b267980d | 1488 | This happens because the presence of overloading is checked too early, |
ee239bfe | 1489 | before any tie()d access is attempted. If the FETCH()ed class of the |
b267980d | 1490 | tie()d value does not change, a simple workaround is to access the value |
ee239bfe IZ |
1491 | immediately after tie()ing, so that after this call the I<previous> class |
1492 | coincides with the current one. | |
1493 | ||
1494 | B<Needed:> a way to fix this without a speed penalty. | |
1495 | ||
b3ac6de7 IZ |
1496 | Barewords are not covered by overloaded string constants. |
1497 | ||
ee239bfe IZ |
1498 | This document is confusing. There are grammos and misleading language |
1499 | used in places. It would seem a total rewrite is needed. | |
4633a7c4 LW |
1500 | |
1501 | =cut | |
1502 |