Commit | Line | Data |
---|---|---|
920dae64 AT |
1 | package overload; |
2 | ||
3 | our $VERSION = '1.04'; | |
4 | ||
5 | $overload::hint_bits = 0x20000; # HINT_LOCALIZE_HH | |
6 | ||
7 | sub nil {} | |
8 | ||
9 | sub OVERLOAD { | |
10 | $package = shift; | |
11 | my %arg = @_; | |
12 | my ($sub, $fb); | |
13 | $ {$package . "::OVERLOAD"}{dummy}++; # Register with magic by touching. | |
14 | *{$package . "::()"} = \&nil; # Make it findable via fetchmethod. | |
15 | for (keys %arg) { | |
16 | if ($_ eq 'fallback') { | |
17 | $fb = $arg{$_}; | |
18 | } else { | |
19 | $sub = $arg{$_}; | |
20 | if (not ref $sub and $sub !~ /::/) { | |
21 | $ {$package . "::(" . $_} = $sub; | |
22 | $sub = \&nil; | |
23 | } | |
24 | #print STDERR "Setting `$ {'package'}::\cO$_' to \\&`$sub'.\n"; | |
25 | *{$package . "::(" . $_} = \&{ $sub }; | |
26 | } | |
27 | } | |
28 | ${$package . "::()"} = $fb; # Make it findable too (fallback only). | |
29 | } | |
30 | ||
31 | sub import { | |
32 | $package = (caller())[0]; | |
33 | # *{$package . "::OVERLOAD"} = \&OVERLOAD; | |
34 | shift; | |
35 | $package->overload::OVERLOAD(@_); | |
36 | } | |
37 | ||
38 | sub unimport { | |
39 | $package = (caller())[0]; | |
40 | ${$package . "::OVERLOAD"}{dummy}++; # Upgrade the table | |
41 | shift; | |
42 | for (@_) { | |
43 | if ($_ eq 'fallback') { | |
44 | undef $ {$package . "::()"}; | |
45 | } else { | |
46 | delete $ {$package . "::"}{"(" . $_}; | |
47 | } | |
48 | } | |
49 | } | |
50 | ||
51 | sub Overloaded { | |
52 | my $package = shift; | |
53 | $package = ref $package if ref $package; | |
54 | $package->can('()'); | |
55 | } | |
56 | ||
57 | sub ov_method { | |
58 | my $globref = shift; | |
59 | return undef unless $globref; | |
60 | my $sub = \&{*$globref}; | |
61 | return $sub if $sub ne \&nil; | |
62 | return shift->can($ {*$globref}); | |
63 | } | |
64 | ||
65 | sub OverloadedStringify { | |
66 | my $package = shift; | |
67 | $package = ref $package if ref $package; | |
68 | #$package->can('(""') | |
69 | ov_method mycan($package, '(""'), $package | |
70 | or ov_method mycan($package, '(0+'), $package | |
71 | or ov_method mycan($package, '(bool'), $package | |
72 | or ov_method mycan($package, '(nomethod'), $package; | |
73 | } | |
74 | ||
75 | sub Method { | |
76 | my $package = shift; | |
77 | $package = ref $package if ref $package; | |
78 | #my $meth = $package->can('(' . shift); | |
79 | ov_method mycan($package, '(' . shift), $package; | |
80 | #return $meth if $meth ne \&nil; | |
81 | #return $ {*{$meth}}; | |
82 | } | |
83 | ||
84 | sub AddrRef { | |
85 | my $package = ref $_[0]; | |
86 | return "$_[0]" unless $package; | |
87 | ||
88 | require Scalar::Util; | |
89 | my $class = Scalar::Util::blessed($_[0]); | |
90 | my $class_prefix = defined($class) ? "$class=" : ""; | |
91 | my $type = Scalar::Util::reftype($_[0]); | |
92 | my $addr = Scalar::Util::refaddr($_[0]); | |
93 | return sprintf("$class_prefix$type(0x%x)", $addr); | |
94 | } | |
95 | ||
96 | *StrVal = *AddrRef; | |
97 | ||
98 | sub mycan { # Real can would leave stubs. | |
99 | my ($package, $meth) = @_; | |
100 | return \*{$package . "::$meth"} if defined &{$package . "::$meth"}; | |
101 | my $p; | |
102 | foreach $p (@{$package . "::ISA"}) { | |
103 | my $out = mycan($p, $meth); | |
104 | return $out if $out; | |
105 | } | |
106 | return undef; | |
107 | } | |
108 | ||
109 | %constants = ( | |
110 | 'integer' => 0x1000, # HINT_NEW_INTEGER | |
111 | 'float' => 0x2000, # HINT_NEW_FLOAT | |
112 | 'binary' => 0x4000, # HINT_NEW_BINARY | |
113 | 'q' => 0x8000, # HINT_NEW_STRING | |
114 | 'qr' => 0x10000, # HINT_NEW_RE | |
115 | ); | |
116 | ||
117 | %ops = ( with_assign => "+ - * / % ** << >> x .", | |
118 | assign => "+= -= *= /= %= **= <<= >>= x= .=", | |
119 | num_comparison => "< <= > >= == !=", | |
120 | '3way_comparison'=> "<=> cmp", | |
121 | str_comparison => "lt le gt ge eq ne", | |
122 | binary => "& | ^", | |
123 | unary => "neg ! ~", | |
124 | mutators => '++ --', | |
125 | func => "atan2 cos sin exp abs log sqrt int", | |
126 | conversion => 'bool "" 0+', | |
127 | iterators => '<>', | |
128 | dereferencing => '${} @{} %{} &{} *{}', | |
129 | special => 'nomethod fallback ='); | |
130 | ||
131 | use warnings::register; | |
132 | sub constant { | |
133 | # Arguments: what, sub | |
134 | while (@_) { | |
135 | if (@_ == 1) { | |
136 | warnings::warnif ("Odd number of arguments for overload::constant"); | |
137 | last; | |
138 | } | |
139 | elsif (!exists $constants {$_ [0]}) { | |
140 | warnings::warnif ("`$_[0]' is not an overloadable type"); | |
141 | } | |
142 | elsif (!ref $_ [1] || "$_[1]" !~ /CODE\(0x[\da-f]+\)$/) { | |
143 | # Can't use C<ref $_[1] eq "CODE"> above as code references can be | |
144 | # blessed, and C<ref> would return the package the ref is blessed into. | |
145 | if (warnings::enabled) { | |
146 | $_ [1] = "undef" unless defined $_ [1]; | |
147 | warnings::warn ("`$_[1]' is not a code reference"); | |
148 | } | |
149 | } | |
150 | else { | |
151 | $^H{$_[0]} = $_[1]; | |
152 | $^H |= $constants{$_[0]} | $overload::hint_bits; | |
153 | } | |
154 | shift, shift; | |
155 | } | |
156 | } | |
157 | ||
158 | sub remove_constant { | |
159 | # Arguments: what, sub | |
160 | while (@_) { | |
161 | delete $^H{$_[0]}; | |
162 | $^H &= ~ $constants{$_[0]}; | |
163 | shift, shift; | |
164 | } | |
165 | } | |
166 | ||
167 | 1; | |
168 | ||
169 | __END__ | |
170 | ||
171 | =head1 NAME | |
172 | ||
173 | overload - Package for overloading Perl operations | |
174 | ||
175 | =head1 SYNOPSIS | |
176 | ||
177 | package SomeThing; | |
178 | ||
179 | use overload | |
180 | '+' => \&myadd, | |
181 | '-' => \&mysub; | |
182 | # etc | |
183 | ... | |
184 | ||
185 | package main; | |
186 | $a = new SomeThing 57; | |
187 | $b=5+$a; | |
188 | ... | |
189 | if (overload::Overloaded $b) {...} | |
190 | ... | |
191 | $strval = overload::StrVal $b; | |
192 | ||
193 | =head1 DESCRIPTION | |
194 | ||
195 | =head2 Declaration of overloaded functions | |
196 | ||
197 | The compilation directive | |
198 | ||
199 | package Number; | |
200 | use overload | |
201 | "+" => \&add, | |
202 | "*=" => "muas"; | |
203 | ||
204 | declares function Number::add() for addition, and method muas() in | |
205 | the "class" C<Number> (or one of its base classes) | |
206 | for the assignment form C<*=> of multiplication. | |
207 | ||
208 | Arguments of this directive come in (key, value) pairs. Legal values | |
209 | are values legal inside a C<&{ ... }> call, so the name of a | |
210 | subroutine, a reference to a subroutine, or an anonymous subroutine | |
211 | will all work. Note that values specified as strings are | |
212 | interpreted as methods, not subroutines. Legal keys are listed below. | |
213 | ||
214 | The subroutine C<add> will be called to execute C<$a+$b> if $a | |
215 | is a reference to an object blessed into the package C<Number>, or if $a is | |
216 | not an object from a package with defined mathemagic addition, but $b is a | |
217 | reference to a C<Number>. It can also be called in other situations, like | |
218 | C<$a+=7>, or C<$a++>. See L<MAGIC AUTOGENERATION>. (Mathemagical | |
219 | methods refer to methods triggered by an overloaded mathematical | |
220 | operator.) | |
221 | ||
222 | Since overloading respects inheritance via the @ISA hierarchy, the | |
223 | above declaration would also trigger overloading of C<+> and C<*=> in | |
224 | all the packages which inherit from C<Number>. | |
225 | ||
226 | =head2 Calling Conventions for Binary Operations | |
227 | ||
228 | The functions specified in the C<use overload ...> directive are called | |
229 | with three (in one particular case with four, see L<Last Resort>) | |
230 | arguments. If the corresponding operation is binary, then the first | |
231 | two arguments are the two arguments of the operation. However, due to | |
232 | general object calling conventions, the first argument should always be | |
233 | an object in the package, so in the situation of C<7+$a>, the | |
234 | order of the arguments is interchanged. It probably does not matter | |
235 | when implementing the addition method, but whether the arguments | |
236 | are reversed is vital to the subtraction method. The method can | |
237 | query this information by examining the third argument, which can take | |
238 | three different values: | |
239 | ||
240 | =over 7 | |
241 | ||
242 | =item FALSE | |
243 | ||
244 | the order of arguments is as in the current operation. | |
245 | ||
246 | =item TRUE | |
247 | ||
248 | the arguments are reversed. | |
249 | ||
250 | =item C<undef> | |
251 | ||
252 | the current operation is an assignment variant (as in | |
253 | C<$a+=7>), but the usual function is called instead. This additional | |
254 | information can be used to generate some optimizations. Compare | |
255 | L<Calling Conventions for Mutators>. | |
256 | ||
257 | =back | |
258 | ||
259 | =head2 Calling Conventions for Unary Operations | |
260 | ||
261 | Unary operation are considered binary operations with the second | |
262 | argument being C<undef>. Thus the functions that overloads C<{"++"}> | |
263 | is called with arguments C<($a,undef,'')> when $a++ is executed. | |
264 | ||
265 | =head2 Calling Conventions for Mutators | |
266 | ||
267 | Two types of mutators have different calling conventions: | |
268 | ||
269 | =over | |
270 | ||
271 | =item C<++> and C<--> | |
272 | ||
273 | The routines which implement these operators are expected to actually | |
274 | I<mutate> their arguments. So, assuming that $obj is a reference to a | |
275 | number, | |
276 | ||
277 | sub incr { my $n = $ {$_[0]}; ++$n; $_[0] = bless \$n} | |
278 | ||
279 | is an appropriate implementation of overloaded C<++>. Note that | |
280 | ||
281 | sub incr { ++$ {$_[0]} ; shift } | |
282 | ||
283 | is OK if used with preincrement and with postincrement. (In the case | |
284 | of postincrement a copying will be performed, see L<Copy Constructor>.) | |
285 | ||
286 | =item C<x=> and other assignment versions | |
287 | ||
288 | There is nothing special about these methods. They may change the | |
289 | value of their arguments, and may leave it as is. The result is going | |
290 | to be assigned to the value in the left-hand-side if different from | |
291 | this value. | |
292 | ||
293 | This allows for the same method to be used as overloaded C<+=> and | |
294 | C<+>. Note that this is I<allowed>, but not recommended, since by the | |
295 | semantic of L<"Fallback"> Perl will call the method for C<+> anyway, | |
296 | if C<+=> is not overloaded. | |
297 | ||
298 | =back | |
299 | ||
300 | B<Warning.> Due to the presence of assignment versions of operations, | |
301 | routines which may be called in assignment context may create | |
302 | self-referential structures. Currently Perl will not free self-referential | |
303 | structures until cycles are C<explicitly> broken. You may get problems | |
304 | when traversing your structures too. | |
305 | ||
306 | Say, | |
307 | ||
308 | use overload '+' => sub { bless [ \$_[0], \$_[1] ] }; | |
309 | ||
310 | is asking for trouble, since for code C<$obj += $foo> the subroutine | |
311 | is called as C<$obj = add($obj, $foo, undef)>, or C<$obj = [\$obj, | |
312 | \$foo]>. If using such a subroutine is an important optimization, one | |
313 | can overload C<+=> explicitly by a non-"optimized" version, or switch | |
314 | to non-optimized version if C<not defined $_[2]> (see | |
315 | L<Calling Conventions for Binary Operations>). | |
316 | ||
317 | Even if no I<explicit> assignment-variants of operators are present in | |
318 | the script, they may be generated by the optimizer. Say, C<",$obj,"> or | |
319 | C<',' . $obj . ','> may be both optimized to | |
320 | ||
321 | my $tmp = ',' . $obj; $tmp .= ','; | |
322 | ||
323 | =head2 Overloadable Operations | |
324 | ||
325 | The following symbols can be specified in C<use overload> directive: | |
326 | ||
327 | =over 5 | |
328 | ||
329 | =item * I<Arithmetic operations> | |
330 | ||
331 | "+", "+=", "-", "-=", "*", "*=", "/", "/=", "%", "%=", | |
332 | "**", "**=", "<<", "<<=", ">>", ">>=", "x", "x=", ".", ".=", | |
333 | ||
334 | For these operations a substituted non-assignment variant can be called if | |
335 | the assignment variant is not available. Methods for operations C<+>, | |
336 | C<->, C<+=>, and C<-=> can be called to automatically generate | |
337 | increment and decrement methods. The operation C<-> can be used to | |
338 | autogenerate missing methods for unary minus or C<abs>. | |
339 | ||
340 | See L<"MAGIC AUTOGENERATION">, L<"Calling Conventions for Mutators"> and | |
341 | L<"Calling Conventions for Binary Operations">) for details of these | |
342 | substitutions. | |
343 | ||
344 | =item * I<Comparison operations> | |
345 | ||
346 | "<", "<=", ">", ">=", "==", "!=", "<=>", | |
347 | "lt", "le", "gt", "ge", "eq", "ne", "cmp", | |
348 | ||
349 | If the corresponding "spaceship" variant is available, it can be | |
350 | used to substitute for the missing operation. During C<sort>ing | |
351 | arrays, C<cmp> is used to compare values subject to C<use overload>. | |
352 | ||
353 | =item * I<Bit operations> | |
354 | ||
355 | "&", "^", "|", "neg", "!", "~", | |
356 | ||
357 | C<neg> stands for unary minus. If the method for C<neg> is not | |
358 | specified, it can be autogenerated using the method for | |
359 | subtraction. If the method for C<!> is not specified, it can be | |
360 | autogenerated using the methods for C<bool>, or C<"">, or C<0+>. | |
361 | ||
362 | =item * I<Increment and decrement> | |
363 | ||
364 | "++", "--", | |
365 | ||
366 | If undefined, addition and subtraction methods can be | |
367 | used instead. These operations are called both in prefix and | |
368 | postfix form. | |
369 | ||
370 | =item * I<Transcendental functions> | |
371 | ||
372 | "atan2", "cos", "sin", "exp", "abs", "log", "sqrt", "int" | |
373 | ||
374 | If C<abs> is unavailable, it can be autogenerated using methods | |
375 | for "E<lt>" or "E<lt>=E<gt>" combined with either unary minus or subtraction. | |
376 | ||
377 | Note that traditionally the Perl function L<int> rounds to 0, thus for | |
378 | floating-point-like types one should follow the same semantic. If | |
379 | C<int> is unavailable, it can be autogenerated using the overloading of | |
380 | C<0+>. | |
381 | ||
382 | =item * I<Boolean, string and numeric conversion> | |
383 | ||
384 | 'bool', '""', '0+', | |
385 | ||
386 | If one or two of these operations are not overloaded, the remaining ones can | |
387 | be used instead. C<bool> is used in the flow control operators | |
388 | (like C<while>) and for the ternary C<?:> operation. These functions can | |
389 | return any arbitrary Perl value. If the corresponding operation for this value | |
390 | is overloaded too, that operation will be called again with this value. | |
391 | ||
392 | As a special case if the overload returns the object itself then it will | |
393 | be used directly. An overloaded conversion returning the object is | |
394 | probably a bug, because you're likely to get something that looks like | |
395 | C<YourPackage=HASH(0x8172b34)>. | |
396 | ||
397 | =item * I<Iteration> | |
398 | ||
399 | "<>" | |
400 | ||
401 | If not overloaded, the argument will be converted to a filehandle or | |
402 | glob (which may require a stringification). The same overloading | |
403 | happens both for the I<read-filehandle> syntax C<E<lt>$varE<gt>> and | |
404 | I<globbing> syntax C<E<lt>${var}E<gt>>. | |
405 | ||
406 | B<BUGS> Even in list context, the iterator is currently called only | |
407 | once and with scalar context. | |
408 | ||
409 | =item * I<Dereferencing> | |
410 | ||
411 | '${}', '@{}', '%{}', '&{}', '*{}'. | |
412 | ||
413 | If not overloaded, the argument will be dereferenced I<as is>, thus | |
414 | should be of correct type. These functions should return a reference | |
415 | of correct type, or another object with overloaded dereferencing. | |
416 | ||
417 | As a special case if the overload returns the object itself then it | |
418 | will be used directly (provided it is the correct type). | |
419 | ||
420 | The dereference operators must be specified explicitly they will not be passed to | |
421 | "nomethod". | |
422 | ||
423 | =item * I<Special> | |
424 | ||
425 | "nomethod", "fallback", "=", | |
426 | ||
427 | see L<SPECIAL SYMBOLS FOR C<use overload>>. | |
428 | ||
429 | =back | |
430 | ||
431 | See L<"Fallback"> for an explanation of when a missing method can be | |
432 | autogenerated. | |
433 | ||
434 | A computer-readable form of the above table is available in the hash | |
435 | %overload::ops, with values being space-separated lists of names: | |
436 | ||
437 | with_assign => '+ - * / % ** << >> x .', | |
438 | assign => '+= -= *= /= %= **= <<= >>= x= .=', | |
439 | num_comparison => '< <= > >= == !=', | |
440 | '3way_comparison'=> '<=> cmp', | |
441 | str_comparison => 'lt le gt ge eq ne', | |
442 | binary => '& | ^', | |
443 | unary => 'neg ! ~', | |
444 | mutators => '++ --', | |
445 | func => 'atan2 cos sin exp abs log sqrt', | |
446 | conversion => 'bool "" 0+', | |
447 | iterators => '<>', | |
448 | dereferencing => '${} @{} %{} &{} *{}', | |
449 | special => 'nomethod fallback =' | |
450 | ||
451 | =head2 Inheritance and overloading | |
452 | ||
453 | Inheritance interacts with overloading in two ways. | |
454 | ||
455 | =over | |
456 | ||
457 | =item Strings as values of C<use overload> directive | |
458 | ||
459 | If C<value> in | |
460 | ||
461 | use overload key => value; | |
462 | ||
463 | is a string, it is interpreted as a method name. | |
464 | ||
465 | =item Overloading of an operation is inherited by derived classes | |
466 | ||
467 | Any class derived from an overloaded class is also overloaded. The | |
468 | set of overloaded methods is the union of overloaded methods of all | |
469 | the ancestors. If some method is overloaded in several ancestor, then | |
470 | which description will be used is decided by the usual inheritance | |
471 | rules: | |
472 | ||
473 | If C<A> inherits from C<B> and C<C> (in this order), C<B> overloads | |
474 | C<+> with C<\&D::plus_sub>, and C<C> overloads C<+> by C<"plus_meth">, | |
475 | then the subroutine C<D::plus_sub> will be called to implement | |
476 | operation C<+> for an object in package C<A>. | |
477 | ||
478 | =back | |
479 | ||
480 | Note that since the value of the C<fallback> key is not a subroutine, | |
481 | its inheritance is not governed by the above rules. In the current | |
482 | implementation, the value of C<fallback> in the first overloaded | |
483 | ancestor is used, but this is accidental and subject to change. | |
484 | ||
485 | =head1 SPECIAL SYMBOLS FOR C<use overload> | |
486 | ||
487 | Three keys are recognized by Perl that are not covered by the above | |
488 | description. | |
489 | ||
490 | =head2 Last Resort | |
491 | ||
492 | C<"nomethod"> should be followed by a reference to a function of four | |
493 | parameters. If defined, it is called when the overloading mechanism | |
494 | cannot find a method for some operation. The first three arguments of | |
495 | this function coincide with the arguments for the corresponding method if | |
496 | it were found, the fourth argument is the symbol | |
497 | corresponding to the missing method. If several methods are tried, | |
498 | the last one is used. Say, C<1-$a> can be equivalent to | |
499 | ||
500 | &nomethodMethod($a,1,1,"-") | |
501 | ||
502 | if the pair C<"nomethod" =E<gt> "nomethodMethod"> was specified in the | |
503 | C<use overload> directive. | |
504 | ||
505 | The C<"nomethod"> mechanism is I<not> used for the dereference operators | |
506 | ( ${} @{} %{} &{} *{} ). | |
507 | ||
508 | ||
509 | If some operation cannot be resolved, and there is no function | |
510 | assigned to C<"nomethod">, then an exception will be raised via die()-- | |
511 | unless C<"fallback"> was specified as a key in C<use overload> directive. | |
512 | ||
513 | ||
514 | =head2 Fallback | |
515 | ||
516 | The key C<"fallback"> governs what to do if a method for a particular | |
517 | operation is not found. Three different cases are possible depending on | |
518 | the value of C<"fallback">: | |
519 | ||
520 | =over 16 | |
521 | ||
522 | =item * C<undef> | |
523 | ||
524 | Perl tries to use a | |
525 | substituted method (see L<MAGIC AUTOGENERATION>). If this fails, it | |
526 | then tries to calls C<"nomethod"> value; if missing, an exception | |
527 | will be raised. | |
528 | ||
529 | =item * TRUE | |
530 | ||
531 | The same as for the C<undef> value, but no exception is raised. Instead, | |
532 | it silently reverts to what it would have done were there no C<use overload> | |
533 | present. | |
534 | ||
535 | =item * defined, but FALSE | |
536 | ||
537 | No autogeneration is tried. Perl tries to call | |
538 | C<"nomethod"> value, and if this is missing, raises an exception. | |
539 | ||
540 | =back | |
541 | ||
542 | B<Note.> C<"fallback"> inheritance via @ISA is not carved in stone | |
543 | yet, see L<"Inheritance and overloading">. | |
544 | ||
545 | =head2 Copy Constructor | |
546 | ||
547 | The value for C<"="> is a reference to a function with three | |
548 | arguments, i.e., it looks like the other values in C<use | |
549 | overload>. However, it does not overload the Perl assignment | |
550 | operator. This would go against Camel hair. | |
551 | ||
552 | This operation is called in the situations when a mutator is applied | |
553 | to a reference that shares its object with some other reference, such | |
554 | as | |
555 | ||
556 | $a=$b; | |
557 | ++$a; | |
558 | ||
559 | To make this change $a and not change $b, a copy of C<$$a> is made, | |
560 | and $a is assigned a reference to this new object. This operation is | |
561 | done during execution of the C<++$a>, and not during the assignment, | |
562 | (so before the increment C<$$a> coincides with C<$$b>). This is only | |
563 | done if C<++> is expressed via a method for C<'++'> or C<'+='> (or | |
564 | C<nomethod>). Note that if this operation is expressed via C<'+'> | |
565 | a nonmutator, i.e., as in | |
566 | ||
567 | $a=$b; | |
568 | $a=$a+1; | |
569 | ||
570 | then C<$a> does not reference a new copy of C<$$a>, since $$a does not | |
571 | appear as lvalue when the above code is executed. | |
572 | ||
573 | If the copy constructor is required during the execution of some mutator, | |
574 | but a method for C<'='> was not specified, it can be autogenerated as a | |
575 | string copy if the object is a plain scalar. | |
576 | ||
577 | =over 5 | |
578 | ||
579 | =item B<Example> | |
580 | ||
581 | The actually executed code for | |
582 | ||
583 | $a=$b; | |
584 | Something else which does not modify $a or $b.... | |
585 | ++$a; | |
586 | ||
587 | may be | |
588 | ||
589 | $a=$b; | |
590 | Something else which does not modify $a or $b.... | |
591 | $a = $a->clone(undef,""); | |
592 | $a->incr(undef,""); | |
593 | ||
594 | if $b was mathemagical, and C<'++'> was overloaded with C<\&incr>, | |
595 | C<'='> was overloaded with C<\&clone>. | |
596 | ||
597 | =back | |
598 | ||
599 | Same behaviour is triggered by C<$b = $a++>, which is consider a synonym for | |
600 | C<$b = $a; ++$a>. | |
601 | ||
602 | =head1 MAGIC AUTOGENERATION | |
603 | ||
604 | If a method for an operation is not found, and the value for C<"fallback"> is | |
605 | TRUE or undefined, Perl tries to autogenerate a substitute method for | |
606 | the missing operation based on the defined operations. Autogenerated method | |
607 | substitutions are possible for the following operations: | |
608 | ||
609 | =over 16 | |
610 | ||
611 | =item I<Assignment forms of arithmetic operations> | |
612 | ||
613 | C<$a+=$b> can use the method for C<"+"> if the method for C<"+="> | |
614 | is not defined. | |
615 | ||
616 | =item I<Conversion operations> | |
617 | ||
618 | String, numeric, and boolean conversion are calculated in terms of one | |
619 | another if not all of them are defined. | |
620 | ||
621 | =item I<Increment and decrement> | |
622 | ||
623 | The C<++$a> operation can be expressed in terms of C<$a+=1> or C<$a+1>, | |
624 | and C<$a--> in terms of C<$a-=1> and C<$a-1>. | |
625 | ||
626 | =item C<abs($a)> | |
627 | ||
628 | can be expressed in terms of C<$aE<lt>0> and C<-$a> (or C<0-$a>). | |
629 | ||
630 | =item I<Unary minus> | |
631 | ||
632 | can be expressed in terms of subtraction. | |
633 | ||
634 | =item I<Negation> | |
635 | ||
636 | C<!> and C<not> can be expressed in terms of boolean conversion, or | |
637 | string or numerical conversion. | |
638 | ||
639 | =item I<Concatenation> | |
640 | ||
641 | can be expressed in terms of string conversion. | |
642 | ||
643 | =item I<Comparison operations> | |
644 | ||
645 | can be expressed in terms of its "spaceship" counterpart: either | |
646 | C<E<lt>=E<gt>> or C<cmp>: | |
647 | ||
648 | <, >, <=, >=, ==, != in terms of <=> | |
649 | lt, gt, le, ge, eq, ne in terms of cmp | |
650 | ||
651 | =item I<Iterator> | |
652 | ||
653 | <> in terms of builtin operations | |
654 | ||
655 | =item I<Dereferencing> | |
656 | ||
657 | ${} @{} %{} &{} *{} in terms of builtin operations | |
658 | ||
659 | =item I<Copy operator> | |
660 | ||
661 | can be expressed in terms of an assignment to the dereferenced value, if this | |
662 | value is a scalar and not a reference. | |
663 | ||
664 | =back | |
665 | ||
666 | =head1 Losing overloading | |
667 | ||
668 | The restriction for the comparison operation is that even if, for example, | |
669 | `C<cmp>' should return a blessed reference, the autogenerated `C<lt>' | |
670 | function will produce only a standard logical value based on the | |
671 | numerical value of the result of `C<cmp>'. In particular, a working | |
672 | numeric conversion is needed in this case (possibly expressed in terms of | |
673 | other conversions). | |
674 | ||
675 | Similarly, C<.=> and C<x=> operators lose their mathemagical properties | |
676 | if the string conversion substitution is applied. | |
677 | ||
678 | When you chop() a mathemagical object it is promoted to a string and its | |
679 | mathemagical properties are lost. The same can happen with other | |
680 | operations as well. | |
681 | ||
682 | =head1 Run-time Overloading | |
683 | ||
684 | Since all C<use> directives are executed at compile-time, the only way to | |
685 | change overloading during run-time is to | |
686 | ||
687 | eval 'use overload "+" => \&addmethod'; | |
688 | ||
689 | You can also use | |
690 | ||
691 | eval 'no overload "+", "--", "<="'; | |
692 | ||
693 | though the use of these constructs during run-time is questionable. | |
694 | ||
695 | =head1 Public functions | |
696 | ||
697 | Package C<overload.pm> provides the following public functions: | |
698 | ||
699 | =over 5 | |
700 | ||
701 | =item overload::StrVal(arg) | |
702 | ||
703 | Gives string value of C<arg> as in absence of stringify overloading. If you | |
704 | are using this to get the address of a reference (useful for checking if two | |
705 | references point to the same thing) then you may be better off using | |
706 | C<Scalar::Util::refaddr()>, which is faster. | |
707 | ||
708 | =item overload::Overloaded(arg) | |
709 | ||
710 | Returns true if C<arg> is subject to overloading of some operations. | |
711 | ||
712 | =item overload::Method(obj,op) | |
713 | ||
714 | Returns C<undef> or a reference to the method that implements C<op>. | |
715 | ||
716 | =back | |
717 | ||
718 | =head1 Overloading constants | |
719 | ||
720 | For some applications, the Perl parser mangles constants too much. | |
721 | It is possible to hook into this process via C<overload::constant()> | |
722 | and C<overload::remove_constant()> functions. | |
723 | ||
724 | These functions take a hash as an argument. The recognized keys of this hash | |
725 | are: | |
726 | ||
727 | =over 8 | |
728 | ||
729 | =item integer | |
730 | ||
731 | to overload integer constants, | |
732 | ||
733 | =item float | |
734 | ||
735 | to overload floating point constants, | |
736 | ||
737 | =item binary | |
738 | ||
739 | to overload octal and hexadecimal constants, | |
740 | ||
741 | =item q | |
742 | ||
743 | to overload C<q>-quoted strings, constant pieces of C<qq>- and C<qx>-quoted | |
744 | strings and here-documents, | |
745 | ||
746 | =item qr | |
747 | ||
748 | to overload constant pieces of regular expressions. | |
749 | ||
750 | =back | |
751 | ||
752 | The corresponding values are references to functions which take three arguments: | |
753 | the first one is the I<initial> string form of the constant, the second one | |
754 | is how Perl interprets this constant, the third one is how the constant is used. | |
755 | Note that the initial string form does not | |
756 | contain string delimiters, and has backslashes in backslash-delimiter | |
757 | combinations stripped (thus the value of delimiter is not relevant for | |
758 | processing of this string). The return value of this function is how this | |
759 | constant is going to be interpreted by Perl. The third argument is undefined | |
760 | unless for overloaded C<q>- and C<qr>- constants, it is C<q> in single-quote | |
761 | context (comes from strings, regular expressions, and single-quote HERE | |
762 | documents), it is C<tr> for arguments of C<tr>/C<y> operators, | |
763 | it is C<s> for right-hand side of C<s>-operator, and it is C<qq> otherwise. | |
764 | ||
765 | Since an expression C<"ab$cd,,"> is just a shortcut for C<'ab' . $cd . ',,'>, | |
766 | it is expected that overloaded constant strings are equipped with reasonable | |
767 | overloaded catenation operator, otherwise absurd results will result. | |
768 | Similarly, negative numbers are considered as negations of positive constants. | |
769 | ||
770 | Note that it is probably meaningless to call the functions overload::constant() | |
771 | and overload::remove_constant() from anywhere but import() and unimport() methods. | |
772 | From these methods they may be called as | |
773 | ||
774 | sub import { | |
775 | shift; | |
776 | return unless @_; | |
777 | die "unknown import: @_" unless @_ == 1 and $_[0] eq ':constant'; | |
778 | overload::constant integer => sub {Math::BigInt->new(shift)}; | |
779 | } | |
780 | ||
781 | B<BUGS> Currently overloaded-ness of constants does not propagate | |
782 | into C<eval '...'>. | |
783 | ||
784 | =head1 IMPLEMENTATION | |
785 | ||
786 | What follows is subject to change RSN. | |
787 | ||
788 | The table of methods for all operations is cached in magic for the | |
789 | symbol table hash for the package. The cache is invalidated during | |
790 | processing of C<use overload>, C<no overload>, new function | |
791 | definitions, and changes in @ISA. However, this invalidation remains | |
792 | unprocessed until the next C<bless>ing into the package. Hence if you | |
793 | want to change overloading structure dynamically, you'll need an | |
794 | additional (fake) C<bless>ing to update the table. | |
795 | ||
796 | (Every SVish thing has a magic queue, and magic is an entry in that | |
797 | queue. This is how a single variable may participate in multiple | |
798 | forms of magic simultaneously. For instance, environment variables | |
799 | regularly have two forms at once: their %ENV magic and their taint | |
800 | magic. However, the magic which implements overloading is applied to | |
801 | the stashes, which are rarely used directly, thus should not slow down | |
802 | Perl.) | |
803 | ||
804 | If an object belongs to a package using overload, it carries a special | |
805 | flag. Thus the only speed penalty during arithmetic operations without | |
806 | overloading is the checking of this flag. | |
807 | ||
808 | In fact, if C<use overload> is not present, there is almost no overhead | |
809 | for overloadable operations, so most programs should not suffer | |
810 | measurable performance penalties. A considerable effort was made to | |
811 | minimize the overhead when overload is used in some package, but the | |
812 | arguments in question do not belong to packages using overload. When | |
813 | in doubt, test your speed with C<use overload> and without it. So far | |
814 | there have been no reports of substantial speed degradation if Perl is | |
815 | compiled with optimization turned on. | |
816 | ||
817 | There is no size penalty for data if overload is not used. The only | |
818 | size penalty if overload is used in some package is that I<all> the | |
819 | packages acquire a magic during the next C<bless>ing into the | |
820 | package. This magic is three-words-long for packages without | |
821 | overloading, and carries the cache table if the package is overloaded. | |
822 | ||
823 | Copying (C<$a=$b>) is shallow; however, a one-level-deep copying is | |
824 | carried out before any operation that can imply an assignment to the | |
825 | object $a (or $b) refers to, like C<$a++>. You can override this | |
826 | behavior by defining your own copy constructor (see L<"Copy Constructor">). | |
827 | ||
828 | It is expected that arguments to methods that are not explicitly supposed | |
829 | to be changed are constant (but this is not enforced). | |
830 | ||
831 | =head1 Metaphor clash | |
832 | ||
833 | One may wonder why the semantic of overloaded C<=> is so counter intuitive. | |
834 | If it I<looks> counter intuitive to you, you are subject to a metaphor | |
835 | clash. | |
836 | ||
837 | Here is a Perl object metaphor: | |
838 | ||
839 | I< object is a reference to blessed data> | |
840 | ||
841 | and an arithmetic metaphor: | |
842 | ||
843 | I< object is a thing by itself>. | |
844 | ||
845 | The I<main> problem of overloading C<=> is the fact that these metaphors | |
846 | imply different actions on the assignment C<$a = $b> if $a and $b are | |
847 | objects. Perl-think implies that $a becomes a reference to whatever | |
848 | $b was referencing. Arithmetic-think implies that the value of "object" | |
849 | $a is changed to become the value of the object $b, preserving the fact | |
850 | that $a and $b are separate entities. | |
851 | ||
852 | The difference is not relevant in the absence of mutators. After | |
853 | a Perl-way assignment an operation which mutates the data referenced by $a | |
854 | would change the data referenced by $b too. Effectively, after | |
855 | C<$a = $b> values of $a and $b become I<indistinguishable>. | |
856 | ||
857 | On the other hand, anyone who has used algebraic notation knows the | |
858 | expressive power of the arithmetic metaphor. Overloading works hard | |
859 | to enable this metaphor while preserving the Perlian way as far as | |
860 | possible. Since it is not possible to freely mix two contradicting | |
861 | metaphors, overloading allows the arithmetic way to write things I<as | |
862 | far as all the mutators are called via overloaded access only>. The | |
863 | way it is done is described in L<Copy Constructor>. | |
864 | ||
865 | If some mutator methods are directly applied to the overloaded values, | |
866 | one may need to I<explicitly unlink> other values which references the | |
867 | same value: | |
868 | ||
869 | $a = new Data 23; | |
870 | ... | |
871 | $b = $a; # $b is "linked" to $a | |
872 | ... | |
873 | $a = $a->clone; # Unlink $b from $a | |
874 | $a->increment_by(4); | |
875 | ||
876 | Note that overloaded access makes this transparent: | |
877 | ||
878 | $a = new Data 23; | |
879 | $b = $a; # $b is "linked" to $a | |
880 | $a += 4; # would unlink $b automagically | |
881 | ||
882 | However, it would not make | |
883 | ||
884 | $a = new Data 23; | |
885 | $a = 4; # Now $a is a plain 4, not 'Data' | |
886 | ||
887 | preserve "objectness" of $a. But Perl I<has> a way to make assignments | |
888 | to an object do whatever you want. It is just not the overload, but | |
889 | tie()ing interface (see L<perlfunc/tie>). Adding a FETCH() method | |
890 | which returns the object itself, and STORE() method which changes the | |
891 | value of the object, one can reproduce the arithmetic metaphor in its | |
892 | completeness, at least for variables which were tie()d from the start. | |
893 | ||
894 | (Note that a workaround for a bug may be needed, see L<"BUGS">.) | |
895 | ||
896 | =head1 Cookbook | |
897 | ||
898 | Please add examples to what follows! | |
899 | ||
900 | =head2 Two-face scalars | |
901 | ||
902 | Put this in F<two_face.pm> in your Perl library directory: | |
903 | ||
904 | package two_face; # Scalars with separate string and | |
905 | # numeric values. | |
906 | sub new { my $p = shift; bless [@_], $p } | |
907 | use overload '""' => \&str, '0+' => \&num, fallback => 1; | |
908 | sub num {shift->[1]} | |
909 | sub str {shift->[0]} | |
910 | ||
911 | Use it as follows: | |
912 | ||
913 | require two_face; | |
914 | my $seven = new two_face ("vii", 7); | |
915 | printf "seven=$seven, seven=%d, eight=%d\n", $seven, $seven+1; | |
916 | print "seven contains `i'\n" if $seven =~ /i/; | |
917 | ||
918 | (The second line creates a scalar which has both a string value, and a | |
919 | numeric value.) This prints: | |
920 | ||
921 | seven=vii, seven=7, eight=8 | |
922 | seven contains `i' | |
923 | ||
924 | =head2 Two-face references | |
925 | ||
926 | Suppose you want to create an object which is accessible as both an | |
927 | array reference and a hash reference, similar to the | |
928 | L<pseudo-hash|perlref/"Pseudo-hashes: Using an array as a hash"> | |
929 | builtin Perl type. Let's make it better than a pseudo-hash by | |
930 | allowing index 0 to be treated as a normal element. | |
931 | ||
932 | package two_refs; | |
933 | use overload '%{}' => \&gethash, '@{}' => sub { $ {shift()} }; | |
934 | sub new { | |
935 | my $p = shift; | |
936 | bless \ [@_], $p; | |
937 | } | |
938 | sub gethash { | |
939 | my %h; | |
940 | my $self = shift; | |
941 | tie %h, ref $self, $self; | |
942 | \%h; | |
943 | } | |
944 | ||
945 | sub TIEHASH { my $p = shift; bless \ shift, $p } | |
946 | my %fields; | |
947 | my $i = 0; | |
948 | $fields{$_} = $i++ foreach qw{zero one two three}; | |
949 | sub STORE { | |
950 | my $self = ${shift()}; | |
951 | my $key = $fields{shift()}; | |
952 | defined $key or die "Out of band access"; | |
953 | $$self->[$key] = shift; | |
954 | } | |
955 | sub FETCH { | |
956 | my $self = ${shift()}; | |
957 | my $key = $fields{shift()}; | |
958 | defined $key or die "Out of band access"; | |
959 | $$self->[$key]; | |
960 | } | |
961 | ||
962 | Now one can access an object using both the array and hash syntax: | |
963 | ||
964 | my $bar = new two_refs 3,4,5,6; | |
965 | $bar->[2] = 11; | |
966 | $bar->{two} == 11 or die 'bad hash fetch'; | |
967 | ||
968 | Note several important features of this example. First of all, the | |
969 | I<actual> type of $bar is a scalar reference, and we do not overload | |
970 | the scalar dereference. Thus we can get the I<actual> non-overloaded | |
971 | contents of $bar by just using C<$$bar> (what we do in functions which | |
972 | overload dereference). Similarly, the object returned by the | |
973 | TIEHASH() method is a scalar reference. | |
974 | ||
975 | Second, we create a new tied hash each time the hash syntax is used. | |
976 | This allows us not to worry about a possibility of a reference loop, | |
977 | which would lead to a memory leak. | |
978 | ||
979 | Both these problems can be cured. Say, if we want to overload hash | |
980 | dereference on a reference to an object which is I<implemented> as a | |
981 | hash itself, the only problem one has to circumvent is how to access | |
982 | this I<actual> hash (as opposed to the I<virtual> hash exhibited by the | |
983 | overloaded dereference operator). Here is one possible fetching routine: | |
984 | ||
985 | sub access_hash { | |
986 | my ($self, $key) = (shift, shift); | |
987 | my $class = ref $self; | |
988 | bless $self, 'overload::dummy'; # Disable overloading of %{} | |
989 | my $out = $self->{$key}; | |
990 | bless $self, $class; # Restore overloading | |
991 | $out; | |
992 | } | |
993 | ||
994 | To remove creation of the tied hash on each access, one may an extra | |
995 | level of indirection which allows a non-circular structure of references: | |
996 | ||
997 | package two_refs1; | |
998 | use overload '%{}' => sub { ${shift()}->[1] }, | |
999 | '@{}' => sub { ${shift()}->[0] }; | |
1000 | sub new { | |
1001 | my $p = shift; | |
1002 | my $a = [@_]; | |
1003 | my %h; | |
1004 | tie %h, $p, $a; | |
1005 | bless \ [$a, \%h], $p; | |
1006 | } | |
1007 | sub gethash { | |
1008 | my %h; | |
1009 | my $self = shift; | |
1010 | tie %h, ref $self, $self; | |
1011 | \%h; | |
1012 | } | |
1013 | ||
1014 | sub TIEHASH { my $p = shift; bless \ shift, $p } | |
1015 | my %fields; | |
1016 | my $i = 0; | |
1017 | $fields{$_} = $i++ foreach qw{zero one two three}; | |
1018 | sub STORE { | |
1019 | my $a = ${shift()}; | |
1020 | my $key = $fields{shift()}; | |
1021 | defined $key or die "Out of band access"; | |
1022 | $a->[$key] = shift; | |
1023 | } | |
1024 | sub FETCH { | |
1025 | my $a = ${shift()}; | |
1026 | my $key = $fields{shift()}; | |
1027 | defined $key or die "Out of band access"; | |
1028 | $a->[$key]; | |
1029 | } | |
1030 | ||
1031 | Now if $baz is overloaded like this, then C<$baz> is a reference to a | |
1032 | reference to the intermediate array, which keeps a reference to an | |
1033 | actual array, and the access hash. The tie()ing object for the access | |
1034 | hash is a reference to a reference to the actual array, so | |
1035 | ||
1036 | =over | |
1037 | ||
1038 | =item * | |
1039 | ||
1040 | There are no loops of references. | |
1041 | ||
1042 | =item * | |
1043 | ||
1044 | Both "objects" which are blessed into the class C<two_refs1> are | |
1045 | references to a reference to an array, thus references to a I<scalar>. | |
1046 | Thus the accessor expression C<$$foo-E<gt>[$ind]> involves no | |
1047 | overloaded operations. | |
1048 | ||
1049 | =back | |
1050 | ||
1051 | =head2 Symbolic calculator | |
1052 | ||
1053 | Put this in F<symbolic.pm> in your Perl library directory: | |
1054 | ||
1055 | package symbolic; # Primitive symbolic calculator | |
1056 | use overload nomethod => \&wrap; | |
1057 | ||
1058 | sub new { shift; bless ['n', @_] } | |
1059 | sub wrap { | |
1060 | my ($obj, $other, $inv, $meth) = @_; | |
1061 | ($obj, $other) = ($other, $obj) if $inv; | |
1062 | bless [$meth, $obj, $other]; | |
1063 | } | |
1064 | ||
1065 | This module is very unusual as overloaded modules go: it does not | |
1066 | provide any usual overloaded operators, instead it provides the L<Last | |
1067 | Resort> operator C<nomethod>. In this example the corresponding | |
1068 | subroutine returns an object which encapsulates operations done over | |
1069 | the objects: C<new symbolic 3> contains C<['n', 3]>, C<2 + new | |
1070 | symbolic 3> contains C<['+', 2, ['n', 3]]>. | |
1071 | ||
1072 | Here is an example of the script which "calculates" the side of | |
1073 | circumscribed octagon using the above package: | |
1074 | ||
1075 | require symbolic; | |
1076 | my $iter = 1; # 2**($iter+2) = 8 | |
1077 | my $side = new symbolic 1; | |
1078 | my $cnt = $iter; | |
1079 | ||
1080 | while ($cnt--) { | |
1081 | $side = (sqrt(1 + $side**2) - 1)/$side; | |
1082 | } | |
1083 | print "OK\n"; | |
1084 | ||
1085 | The value of $side is | |
1086 | ||
1087 | ['/', ['-', ['sqrt', ['+', 1, ['**', ['n', 1], 2]], | |
1088 | undef], 1], ['n', 1]] | |
1089 | ||
1090 | Note that while we obtained this value using a nice little script, | |
1091 | there is no simple way to I<use> this value. In fact this value may | |
1092 | be inspected in debugger (see L<perldebug>), but ony if | |
1093 | C<bareStringify> B<O>ption is set, and not via C<p> command. | |
1094 | ||
1095 | If one attempts to print this value, then the overloaded operator | |
1096 | C<""> will be called, which will call C<nomethod> operator. The | |
1097 | result of this operator will be stringified again, but this result is | |
1098 | again of type C<symbolic>, which will lead to an infinite loop. | |
1099 | ||
1100 | Add a pretty-printer method to the module F<symbolic.pm>: | |
1101 | ||
1102 | sub pretty { | |
1103 | my ($meth, $a, $b) = @{+shift}; | |
1104 | $a = 'u' unless defined $a; | |
1105 | $b = 'u' unless defined $b; | |
1106 | $a = $a->pretty if ref $a; | |
1107 | $b = $b->pretty if ref $b; | |
1108 | "[$meth $a $b]"; | |
1109 | } | |
1110 | ||
1111 | Now one can finish the script by | |
1112 | ||
1113 | print "side = ", $side->pretty, "\n"; | |
1114 | ||
1115 | The method C<pretty> is doing object-to-string conversion, so it | |
1116 | is natural to overload the operator C<""> using this method. However, | |
1117 | inside such a method it is not necessary to pretty-print the | |
1118 | I<components> $a and $b of an object. In the above subroutine | |
1119 | C<"[$meth $a $b]"> is a catenation of some strings and components $a | |
1120 | and $b. If these components use overloading, the catenation operator | |
1121 | will look for an overloaded operator C<.>; if not present, it will | |
1122 | look for an overloaded operator C<"">. Thus it is enough to use | |
1123 | ||
1124 | use overload nomethod => \&wrap, '""' => \&str; | |
1125 | sub str { | |
1126 | my ($meth, $a, $b) = @{+shift}; | |
1127 | $a = 'u' unless defined $a; | |
1128 | $b = 'u' unless defined $b; | |
1129 | "[$meth $a $b]"; | |
1130 | } | |
1131 | ||
1132 | Now one can change the last line of the script to | |
1133 | ||
1134 | print "side = $side\n"; | |
1135 | ||
1136 | which outputs | |
1137 | ||
1138 | side = [/ [- [sqrt [+ 1 [** [n 1 u] 2]] u] 1] [n 1 u]] | |
1139 | ||
1140 | and one can inspect the value in debugger using all the possible | |
1141 | methods. | |
1142 | ||
1143 | Something is still amiss: consider the loop variable $cnt of the | |
1144 | script. It was a number, not an object. We cannot make this value of | |
1145 | type C<symbolic>, since then the loop will not terminate. | |
1146 | ||
1147 | Indeed, to terminate the cycle, the $cnt should become false. | |
1148 | However, the operator C<bool> for checking falsity is overloaded (this | |
1149 | time via overloaded C<"">), and returns a long string, thus any object | |
1150 | of type C<symbolic> is true. To overcome this, we need a way to | |
1151 | compare an object to 0. In fact, it is easier to write a numeric | |
1152 | conversion routine. | |
1153 | ||
1154 | Here is the text of F<symbolic.pm> with such a routine added (and | |
1155 | slightly modified str()): | |
1156 | ||
1157 | package symbolic; # Primitive symbolic calculator | |
1158 | use overload | |
1159 | nomethod => \&wrap, '""' => \&str, '0+' => \# | |
1160 | ||
1161 | sub new { shift; bless ['n', @_] } | |
1162 | sub wrap { | |
1163 | my ($obj, $other, $inv, $meth) = @_; | |
1164 | ($obj, $other) = ($other, $obj) if $inv; | |
1165 | bless [$meth, $obj, $other]; | |
1166 | } | |
1167 | sub str { | |
1168 | my ($meth, $a, $b) = @{+shift}; | |
1169 | $a = 'u' unless defined $a; | |
1170 | if (defined $b) { | |
1171 | "[$meth $a $b]"; | |
1172 | } else { | |
1173 | "[$meth $a]"; | |
1174 | } | |
1175 | } | |
1176 | my %subr = ( n => sub {$_[0]}, | |
1177 | sqrt => sub {sqrt $_[0]}, | |
1178 | '-' => sub {shift() - shift()}, | |
1179 | '+' => sub {shift() + shift()}, | |
1180 | '/' => sub {shift() / shift()}, | |
1181 | '*' => sub {shift() * shift()}, | |
1182 | '**' => sub {shift() ** shift()}, | |
1183 | ); | |
1184 | sub num { | |
1185 | my ($meth, $a, $b) = @{+shift}; | |
1186 | my $subr = $subr{$meth} | |
1187 | or die "Do not know how to ($meth) in symbolic"; | |
1188 | $a = $a->num if ref $a eq __PACKAGE__; | |
1189 | $b = $b->num if ref $b eq __PACKAGE__; | |
1190 | $subr->($a,$b); | |
1191 | } | |
1192 | ||
1193 | All the work of numeric conversion is done in %subr and num(). Of | |
1194 | course, %subr is not complete, it contains only operators used in the | |
1195 | example below. Here is the extra-credit question: why do we need an | |
1196 | explicit recursion in num()? (Answer is at the end of this section.) | |
1197 | ||
1198 | Use this module like this: | |
1199 | ||
1200 | require symbolic; | |
1201 | my $iter = new symbolic 2; # 16-gon | |
1202 | my $side = new symbolic 1; | |
1203 | my $cnt = $iter; | |
1204 | ||
1205 | while ($cnt) { | |
1206 | $cnt = $cnt - 1; # Mutator `--' not implemented | |
1207 | $side = (sqrt(1 + $side**2) - 1)/$side; | |
1208 | } | |
1209 | printf "%s=%f\n", $side, $side; | |
1210 | printf "pi=%f\n", $side*(2**($iter+2)); | |
1211 | ||
1212 | It prints (without so many line breaks) | |
1213 | ||
1214 | [/ [- [sqrt [+ 1 [** [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] | |
1215 | [n 1]] 2]]] 1] | |
1216 | [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] [n 1]]]=0.198912 | |
1217 | pi=3.182598 | |
1218 | ||
1219 | The above module is very primitive. It does not implement | |
1220 | mutator methods (C<++>, C<-=> and so on), does not do deep copying | |
1221 | (not required without mutators!), and implements only those arithmetic | |
1222 | operations which are used in the example. | |
1223 | ||
1224 | To implement most arithmetic operations is easy; one should just use | |
1225 | the tables of operations, and change the code which fills %subr to | |
1226 | ||
1227 | my %subr = ( 'n' => sub {$_[0]} ); | |
1228 | foreach my $op (split " ", $overload::ops{with_assign}) { | |
1229 | $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}"; | |
1230 | } | |
1231 | my @bins = qw(binary 3way_comparison num_comparison str_comparison); | |
1232 | foreach my $op (split " ", "@overload::ops{ @bins }") { | |
1233 | $subr{$op} = eval "sub {shift() $op shift()}"; | |
1234 | } | |
1235 | foreach my $op (split " ", "@overload::ops{qw(unary func)}") { | |
1236 | print "defining `$op'\n"; | |
1237 | $subr{$op} = eval "sub {$op shift()}"; | |
1238 | } | |
1239 | ||
1240 | Due to L<Calling Conventions for Mutators>, we do not need anything | |
1241 | special to make C<+=> and friends work, except filling C<+=> entry of | |
1242 | %subr, and defining a copy constructor (needed since Perl has no | |
1243 | way to know that the implementation of C<'+='> does not mutate | |
1244 | the argument, compare L<Copy Constructor>). | |
1245 | ||
1246 | To implement a copy constructor, add C<< '=' => \&cpy >> to C<use overload> | |
1247 | line, and code (this code assumes that mutators change things one level | |
1248 | deep only, so recursive copying is not needed): | |
1249 | ||
1250 | sub cpy { | |
1251 | my $self = shift; | |
1252 | bless [@$self], ref $self; | |
1253 | } | |
1254 | ||
1255 | To make C<++> and C<--> work, we need to implement actual mutators, | |
1256 | either directly, or in C<nomethod>. We continue to do things inside | |
1257 | C<nomethod>, thus add | |
1258 | ||
1259 | if ($meth eq '++' or $meth eq '--') { | |
1260 | @$obj = ($meth, (bless [@$obj]), 1); # Avoid circular reference | |
1261 | return $obj; | |
1262 | } | |
1263 | ||
1264 | after the first line of wrap(). This is not a most effective | |
1265 | implementation, one may consider | |
1266 | ||
1267 | sub inc { $_[0] = bless ['++', shift, 1]; } | |
1268 | ||
1269 | instead. | |
1270 | ||
1271 | As a final remark, note that one can fill %subr by | |
1272 | ||
1273 | my %subr = ( 'n' => sub {$_[0]} ); | |
1274 | foreach my $op (split " ", $overload::ops{with_assign}) { | |
1275 | $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}"; | |
1276 | } | |
1277 | my @bins = qw(binary 3way_comparison num_comparison str_comparison); | |
1278 | foreach my $op (split " ", "@overload::ops{ @bins }") { | |
1279 | $subr{$op} = eval "sub {shift() $op shift()}"; | |
1280 | } | |
1281 | foreach my $op (split " ", "@overload::ops{qw(unary func)}") { | |
1282 | $subr{$op} = eval "sub {$op shift()}"; | |
1283 | } | |
1284 | $subr{'++'} = $subr{'+'}; | |
1285 | $subr{'--'} = $subr{'-'}; | |
1286 | ||
1287 | This finishes implementation of a primitive symbolic calculator in | |
1288 | 50 lines of Perl code. Since the numeric values of subexpressions | |
1289 | are not cached, the calculator is very slow. | |
1290 | ||
1291 | Here is the answer for the exercise: In the case of str(), we need no | |
1292 | explicit recursion since the overloaded C<.>-operator will fall back | |
1293 | to an existing overloaded operator C<"">. Overloaded arithmetic | |
1294 | operators I<do not> fall back to numeric conversion if C<fallback> is | |
1295 | not explicitly requested. Thus without an explicit recursion num() | |
1296 | would convert C<['+', $a, $b]> to C<$a + $b>, which would just rebuild | |
1297 | the argument of num(). | |
1298 | ||
1299 | If you wonder why defaults for conversion are different for str() and | |
1300 | num(), note how easy it was to write the symbolic calculator. This | |
1301 | simplicity is due to an appropriate choice of defaults. One extra | |
1302 | note: due to the explicit recursion num() is more fragile than sym(): | |
1303 | we need to explicitly check for the type of $a and $b. If components | |
1304 | $a and $b happen to be of some related type, this may lead to problems. | |
1305 | ||
1306 | =head2 I<Really> symbolic calculator | |
1307 | ||
1308 | One may wonder why we call the above calculator symbolic. The reason | |
1309 | is that the actual calculation of the value of expression is postponed | |
1310 | until the value is I<used>. | |
1311 | ||
1312 | To see it in action, add a method | |
1313 | ||
1314 | sub STORE { | |
1315 | my $obj = shift; | |
1316 | $#$obj = 1; | |
1317 | @$obj->[0,1] = ('=', shift); | |
1318 | } | |
1319 | ||
1320 | to the package C<symbolic>. After this change one can do | |
1321 | ||
1322 | my $a = new symbolic 3; | |
1323 | my $b = new symbolic 4; | |
1324 | my $c = sqrt($a**2 + $b**2); | |
1325 | ||
1326 | and the numeric value of $c becomes 5. However, after calling | |
1327 | ||
1328 | $a->STORE(12); $b->STORE(5); | |
1329 | ||
1330 | the numeric value of $c becomes 13. There is no doubt now that the module | |
1331 | symbolic provides a I<symbolic> calculator indeed. | |
1332 | ||
1333 | To hide the rough edges under the hood, provide a tie()d interface to the | |
1334 | package C<symbolic> (compare with L<Metaphor clash>). Add methods | |
1335 | ||
1336 | sub TIESCALAR { my $pack = shift; $pack->new(@_) } | |
1337 | sub FETCH { shift } | |
1338 | sub nop { } # Around a bug | |
1339 | ||
1340 | (the bug is described in L<"BUGS">). One can use this new interface as | |
1341 | ||
1342 | tie $a, 'symbolic', 3; | |
1343 | tie $b, 'symbolic', 4; | |
1344 | $a->nop; $b->nop; # Around a bug | |
1345 | ||
1346 | my $c = sqrt($a**2 + $b**2); | |
1347 | ||
1348 | Now numeric value of $c is 5. After C<$a = 12; $b = 5> the numeric value | |
1349 | of $c becomes 13. To insulate the user of the module add a method | |
1350 | ||
1351 | sub vars { my $p = shift; tie($_, $p), $_->nop foreach @_; } | |
1352 | ||
1353 | Now | |
1354 | ||
1355 | my ($a, $b); | |
1356 | symbolic->vars($a, $b); | |
1357 | my $c = sqrt($a**2 + $b**2); | |
1358 | ||
1359 | $a = 3; $b = 4; | |
1360 | printf "c5 %s=%f\n", $c, $c; | |
1361 | ||
1362 | $a = 12; $b = 5; | |
1363 | printf "c13 %s=%f\n", $c, $c; | |
1364 | ||
1365 | shows that the numeric value of $c follows changes to the values of $a | |
1366 | and $b. | |
1367 | ||
1368 | =head1 AUTHOR | |
1369 | ||
1370 | Ilya Zakharevich E<lt>F<ilya@math.mps.ohio-state.edu>E<gt>. | |
1371 | ||
1372 | =head1 DIAGNOSTICS | |
1373 | ||
1374 | When Perl is run with the B<-Do> switch or its equivalent, overloading | |
1375 | induces diagnostic messages. | |
1376 | ||
1377 | Using the C<m> command of Perl debugger (see L<perldebug>) one can | |
1378 | deduce which operations are overloaded (and which ancestor triggers | |
1379 | this overloading). Say, if C<eq> is overloaded, then the method C<(eq> | |
1380 | is shown by debugger. The method C<()> corresponds to the C<fallback> | |
1381 | key (in fact a presence of this method shows that this package has | |
1382 | overloading enabled, and it is what is used by the C<Overloaded> | |
1383 | function of module C<overload>). | |
1384 | ||
1385 | The module might issue the following warnings: | |
1386 | ||
1387 | =over 4 | |
1388 | ||
1389 | =item Odd number of arguments for overload::constant | |
1390 | ||
1391 | (W) The call to overload::constant contained an odd number of arguments. | |
1392 | The arguments should come in pairs. | |
1393 | ||
1394 | =item `%s' is not an overloadable type | |
1395 | ||
1396 | (W) You tried to overload a constant type the overload package is unaware of. | |
1397 | ||
1398 | =item `%s' is not a code reference | |
1399 | ||
1400 | (W) The second (fourth, sixth, ...) argument of overload::constant needs | |
1401 | to be a code reference. Either an anonymous subroutine, or a reference | |
1402 | to a subroutine. | |
1403 | ||
1404 | =back | |
1405 | ||
1406 | =head1 BUGS | |
1407 | ||
1408 | Because it is used for overloading, the per-package hash %OVERLOAD now | |
1409 | has a special meaning in Perl. The symbol table is filled with names | |
1410 | looking like line-noise. | |
1411 | ||
1412 | For the purpose of inheritance every overloaded package behaves as if | |
1413 | C<fallback> is present (possibly undefined). This may create | |
1414 | interesting effects if some package is not overloaded, but inherits | |
1415 | from two overloaded packages. | |
1416 | ||
1417 | Relation between overloading and tie()ing is broken. Overloading is | |
1418 | triggered or not basing on the I<previous> class of tie()d value. | |
1419 | ||
1420 | This happens because the presence of overloading is checked too early, | |
1421 | before any tie()d access is attempted. If the FETCH()ed class of the | |
1422 | tie()d value does not change, a simple workaround is to access the value | |
1423 | immediately after tie()ing, so that after this call the I<previous> class | |
1424 | coincides with the current one. | |
1425 | ||
1426 | B<Needed:> a way to fix this without a speed penalty. | |
1427 | ||
1428 | Barewords are not covered by overloaded string constants. | |
1429 | ||
1430 | This document is confusing. There are grammos and misleading language | |
1431 | used in places. It would seem a total rewrite is needed. | |
1432 | ||
1433 | =cut | |
1434 |