Commit | Line | Data |
---|---|---|
4633a7c4 LW |
1 | package overload; |
2 | ||
98225a64 | 3 | our $VERSION = '1.04'; |
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; |
75 | $package = ref $package if ref $package; | |
44a8e56a | 76 | #my $meth = $package->can('(' . shift); |
77 | ov_method mycan($package, '(' . shift), $package; | |
78 | #return $meth if $meth ne \&nil; | |
79 | #return $ {*{$meth}}; | |
4633a7c4 LW |
80 | } |
81 | ||
82 | sub AddrRef { | |
a6006777 | 83 | my $package = ref $_[0]; |
84 | return "$_[0]" unless $package; | |
b3c0ec7c FD |
85 | |
86 | require Scalar::Util; | |
87 | my $class = Scalar::Util::blessed($_[0]); | |
88 | my $class_prefix = defined($class) ? "$class=" : ""; | |
89 | my $type = Scalar::Util::reftype($_[0]); | |
90 | my $addr = Scalar::Util::refaddr($_[0]); | |
91 | return sprintf("$class_prefix$type(0x%x)", $addr); | |
4633a7c4 LW |
92 | } |
93 | ||
1b1d102f | 94 | *StrVal = *AddrRef; |
4633a7c4 | 95 | |
44a8e56a | 96 | sub mycan { # Real can would leave stubs. |
97 | my ($package, $meth) = @_; | |
98 | return \*{$package . "::$meth"} if defined &{$package . "::$meth"}; | |
99 | my $p; | |
100 | foreach $p (@{$package . "::ISA"}) { | |
101 | my $out = mycan($p, $meth); | |
102 | return $out if $out; | |
103 | } | |
104 | return undef; | |
105 | } | |
106 | ||
b3ac6de7 | 107 | %constants = ( |
9cfe5470 RGS |
108 | 'integer' => 0x1000, # HINT_NEW_INTEGER |
109 | 'float' => 0x2000, # HINT_NEW_FLOAT | |
110 | 'binary' => 0x4000, # HINT_NEW_BINARY | |
111 | 'q' => 0x8000, # HINT_NEW_STRING | |
112 | 'qr' => 0x10000, # HINT_NEW_RE | |
b3ac6de7 IZ |
113 | ); |
114 | ||
ee239bfe IZ |
115 | %ops = ( with_assign => "+ - * / % ** << >> x .", |
116 | assign => "+= -= *= /= %= **= <<= >>= x= .=", | |
2877bd81 | 117 | num_comparison => "< <= > >= == !=", |
ee239bfe | 118 | '3way_comparison'=> "<=> cmp", |
2877bd81 | 119 | str_comparison => "lt le gt ge eq ne", |
ee239bfe IZ |
120 | binary => "& | ^", |
121 | unary => "neg ! ~", | |
122 | mutators => '++ --', | |
f216259d | 123 | func => "atan2 cos sin exp abs log sqrt int", |
ee239bfe | 124 | conversion => 'bool "" 0+', |
f5284f61 IZ |
125 | iterators => '<>', |
126 | dereferencing => '${} @{} %{} &{} *{}', | |
ee239bfe IZ |
127 | special => 'nomethod fallback ='); |
128 | ||
6b82e2f5 | 129 | use warnings::register; |
b3ac6de7 IZ |
130 | sub constant { |
131 | # Arguments: what, sub | |
132 | while (@_) { | |
6b82e2f5 | 133 | if (@_ == 1) { |
4498a751 | 134 | warnings::warnif ("Odd number of arguments for overload::constant"); |
6b82e2f5 A |
135 | last; |
136 | } | |
137 | elsif (!exists $constants {$_ [0]}) { | |
4498a751 | 138 | warnings::warnif ("`$_[0]' is not an overloadable type"); |
6b82e2f5 A |
139 | } |
140 | elsif (!ref $_ [1] || "$_[1]" !~ /CODE\(0x[\da-f]+\)$/) { | |
141 | # Can't use C<ref $_[1] eq "CODE"> above as code references can be | |
142 | # blessed, and C<ref> would return the package the ref is blessed into. | |
143 | if (warnings::enabled) { | |
6b82e2f5 | 144 | $_ [1] = "undef" unless defined $_ [1]; |
4498a751 | 145 | warnings::warn ("`$_[1]' is not a code reference"); |
6b82e2f5 A |
146 | } |
147 | } | |
148 | else { | |
149 | $^H{$_[0]} = $_[1]; | |
f22a2069 | 150 | $^H |= $constants{$_[0]}; |
6b82e2f5 | 151 | } |
b3ac6de7 IZ |
152 | shift, shift; |
153 | } | |
154 | } | |
155 | ||
156 | sub remove_constant { | |
157 | # Arguments: what, sub | |
158 | while (@_) { | |
159 | delete $^H{$_[0]}; | |
160 | $^H &= ~ $constants{$_[0]}; | |
161 | shift, shift; | |
162 | } | |
163 | } | |
164 | ||
4633a7c4 LW |
165 | 1; |
166 | ||
167 | __END__ | |
168 | ||
b267980d | 169 | =head1 NAME |
4633a7c4 | 170 | |
7adf7a02 | 171 | overload - Package for overloading Perl operations |
4633a7c4 LW |
172 | |
173 | =head1 SYNOPSIS | |
174 | ||
175 | package SomeThing; | |
176 | ||
b267980d | 177 | use overload |
4633a7c4 LW |
178 | '+' => \&myadd, |
179 | '-' => \&mysub; | |
180 | # etc | |
181 | ... | |
182 | ||
183 | package main; | |
184 | $a = new SomeThing 57; | |
185 | $b=5+$a; | |
186 | ... | |
187 | if (overload::Overloaded $b) {...} | |
188 | ... | |
189 | $strval = overload::StrVal $b; | |
190 | ||
4633a7c4 LW |
191 | =head1 DESCRIPTION |
192 | ||
193 | =head2 Declaration of overloaded functions | |
194 | ||
195 | The compilation directive | |
196 | ||
197 | package Number; | |
198 | use overload | |
b267980d | 199 | "+" => \&add, |
4633a7c4 LW |
200 | "*=" => "muas"; |
201 | ||
202 | declares function Number::add() for addition, and method muas() in | |
203 | the "class" C<Number> (or one of its base classes) | |
b267980d | 204 | for the assignment form C<*=> of multiplication. |
4633a7c4 LW |
205 | |
206 | Arguments of this directive come in (key, value) pairs. Legal values | |
e7ea3e70 IZ |
207 | are values legal inside a C<&{ ... }> call, so the name of a |
208 | subroutine, a reference to a subroutine, or an anonymous subroutine | |
209 | will all work. Note that values specified as strings are | |
210 | interpreted as methods, not subroutines. Legal keys are listed below. | |
4633a7c4 LW |
211 | |
212 | The subroutine C<add> will be called to execute C<$a+$b> if $a | |
213 | is a reference to an object blessed into the package C<Number>, or if $a is | |
214 | not an object from a package with defined mathemagic addition, but $b is a | |
215 | reference to a C<Number>. It can also be called in other situations, like | |
216 | C<$a+=7>, or C<$a++>. See L<MAGIC AUTOGENERATION>. (Mathemagical | |
217 | methods refer to methods triggered by an overloaded mathematical | |
218 | operator.) | |
219 | ||
774d564b | 220 | Since overloading respects inheritance via the @ISA hierarchy, the |
221 | above declaration would also trigger overloading of C<+> and C<*=> in | |
222 | all the packages which inherit from C<Number>. | |
e7ea3e70 | 223 | |
4633a7c4 LW |
224 | =head2 Calling Conventions for Binary Operations |
225 | ||
226 | The functions specified in the C<use overload ...> directive are called | |
227 | with three (in one particular case with four, see L<Last Resort>) | |
228 | arguments. If the corresponding operation is binary, then the first | |
229 | two arguments are the two arguments of the operation. However, due to | |
230 | general object calling conventions, the first argument should always be | |
231 | an object in the package, so in the situation of C<7+$a>, the | |
232 | order of the arguments is interchanged. It probably does not matter | |
233 | when implementing the addition method, but whether the arguments | |
234 | are reversed is vital to the subtraction method. The method can | |
235 | query this information by examining the third argument, which can take | |
236 | three different values: | |
237 | ||
238 | =over 7 | |
239 | ||
240 | =item FALSE | |
241 | ||
242 | the order of arguments is as in the current operation. | |
243 | ||
244 | =item TRUE | |
245 | ||
246 | the arguments are reversed. | |
247 | ||
248 | =item C<undef> | |
249 | ||
250 | the current operation is an assignment variant (as in | |
251 | C<$a+=7>), but the usual function is called instead. This additional | |
ee239bfe IZ |
252 | information can be used to generate some optimizations. Compare |
253 | L<Calling Conventions for Mutators>. | |
4633a7c4 LW |
254 | |
255 | =back | |
256 | ||
257 | =head2 Calling Conventions for Unary Operations | |
258 | ||
259 | Unary operation are considered binary operations with the second | |
260 | argument being C<undef>. Thus the functions that overloads C<{"++"}> | |
261 | is called with arguments C<($a,undef,'')> when $a++ is executed. | |
262 | ||
ee239bfe IZ |
263 | =head2 Calling Conventions for Mutators |
264 | ||
265 | Two types of mutators have different calling conventions: | |
266 | ||
88c28ceb | 267 | =over |
ee239bfe IZ |
268 | |
269 | =item C<++> and C<--> | |
270 | ||
271 | The routines which implement these operators are expected to actually | |
272 | I<mutate> their arguments. So, assuming that $obj is a reference to a | |
273 | number, | |
274 | ||
275 | sub incr { my $n = $ {$_[0]}; ++$n; $_[0] = bless \$n} | |
276 | ||
277 | is an appropriate implementation of overloaded C<++>. Note that | |
278 | ||
279 | sub incr { ++$ {$_[0]} ; shift } | |
280 | ||
281 | is OK if used with preincrement and with postincrement. (In the case | |
282 | of postincrement a copying will be performed, see L<Copy Constructor>.) | |
283 | ||
284 | =item C<x=> and other assignment versions | |
285 | ||
286 | There is nothing special about these methods. They may change the | |
287 | value of their arguments, and may leave it as is. The result is going | |
288 | to be assigned to the value in the left-hand-side if different from | |
289 | this value. | |
290 | ||
f610777f | 291 | This allows for the same method to be used as overloaded C<+=> and |
ee239bfe IZ |
292 | C<+>. Note that this is I<allowed>, but not recommended, since by the |
293 | semantic of L<"Fallback"> Perl will call the method for C<+> anyway, | |
294 | if C<+=> is not overloaded. | |
295 | ||
296 | =back | |
297 | ||
d1be9408 | 298 | B<Warning.> Due to the presence of assignment versions of operations, |
b267980d NIS |
299 | routines which may be called in assignment context may create |
300 | self-referential structures. Currently Perl will not free self-referential | |
ee239bfe IZ |
301 | structures until cycles are C<explicitly> broken. You may get problems |
302 | when traversing your structures too. | |
303 | ||
b267980d | 304 | Say, |
ee239bfe IZ |
305 | |
306 | use overload '+' => sub { bless [ \$_[0], \$_[1] ] }; | |
307 | ||
308 | is asking for trouble, since for code C<$obj += $foo> the subroutine | |
b267980d | 309 | is called as C<$obj = add($obj, $foo, undef)>, or C<$obj = [\$obj, |
ee239bfe IZ |
310 | \$foo]>. If using such a subroutine is an important optimization, one |
311 | can overload C<+=> explicitly by a non-"optimized" version, or switch | |
b267980d | 312 | to non-optimized version if C<not defined $_[2]> (see |
ee239bfe IZ |
313 | L<Calling Conventions for Binary Operations>). |
314 | ||
315 | Even if no I<explicit> assignment-variants of operators are present in | |
316 | the script, they may be generated by the optimizer. Say, C<",$obj,"> or | |
317 | C<',' . $obj . ','> may be both optimized to | |
318 | ||
319 | my $tmp = ',' . $obj; $tmp .= ','; | |
320 | ||
4633a7c4 LW |
321 | =head2 Overloadable Operations |
322 | ||
ee239bfe | 323 | The following symbols can be specified in C<use overload> directive: |
4633a7c4 LW |
324 | |
325 | =over 5 | |
326 | ||
327 | =item * I<Arithmetic operations> | |
328 | ||
329 | "+", "+=", "-", "-=", "*", "*=", "/", "/=", "%", "%=", | |
330 | "**", "**=", "<<", "<<=", ">>", ">>=", "x", "x=", ".", ".=", | |
331 | ||
332 | For these operations a substituted non-assignment variant can be called if | |
fa8a6580 MS |
333 | the assignment variant is not available. Methods for operations C<+>, |
334 | C<->, C<+=>, and C<-=> can be called to automatically generate | |
335 | increment and decrement methods. The operation C<-> can be used to | |
4633a7c4 LW |
336 | autogenerate missing methods for unary minus or C<abs>. |
337 | ||
ee239bfe IZ |
338 | See L<"MAGIC AUTOGENERATION">, L<"Calling Conventions for Mutators"> and |
339 | L<"Calling Conventions for Binary Operations">) for details of these | |
340 | substitutions. | |
341 | ||
4633a7c4 LW |
342 | =item * I<Comparison operations> |
343 | ||
344 | "<", "<=", ">", ">=", "==", "!=", "<=>", | |
345 | "lt", "le", "gt", "ge", "eq", "ne", "cmp", | |
346 | ||
347 | If the corresponding "spaceship" variant is available, it can be | |
348 | used to substitute for the missing operation. During C<sort>ing | |
349 | arrays, C<cmp> is used to compare values subject to C<use overload>. | |
350 | ||
351 | =item * I<Bit operations> | |
352 | ||
353 | "&", "^", "|", "neg", "!", "~", | |
354 | ||
fa8a6580 | 355 | C<neg> stands for unary minus. If the method for C<neg> is not |
3bc6ec80 | 356 | specified, it can be autogenerated using the method for |
fa8a6580 MS |
357 | subtraction. If the method for C<!> is not specified, it can be |
358 | autogenerated using the methods for C<bool>, or C<"">, or C<0+>. | |
4633a7c4 LW |
359 | |
360 | =item * I<Increment and decrement> | |
361 | ||
362 | "++", "--", | |
363 | ||
364 | If undefined, addition and subtraction methods can be | |
365 | used instead. These operations are called both in prefix and | |
366 | postfix form. | |
367 | ||
368 | =item * I<Transcendental functions> | |
369 | ||
f216259d | 370 | "atan2", "cos", "sin", "exp", "abs", "log", "sqrt", "int" |
4633a7c4 LW |
371 | |
372 | If C<abs> is unavailable, it can be autogenerated using methods | |
1fef88e7 | 373 | for "E<lt>" or "E<lt>=E<gt>" combined with either unary minus or subtraction. |
4633a7c4 | 374 | |
f216259d IZ |
375 | Note that traditionally the Perl function L<int> rounds to 0, thus for |
376 | floating-point-like types one should follow the same semantic. If | |
377 | C<int> is unavailable, it can be autogenerated using the overloading of | |
378 | C<0+>. | |
379 | ||
4633a7c4 LW |
380 | =item * I<Boolean, string and numeric conversion> |
381 | ||
fa8a6580 | 382 | 'bool', '""', '0+', |
4633a7c4 | 383 | |
f5284f61 | 384 | If one or two of these operations are not overloaded, the remaining ones can |
4633a7c4 | 385 | be used instead. C<bool> is used in the flow control operators |
fa8a6580 | 386 | (like C<while>) and for the ternary C<?:> operation. These functions can |
4633a7c4 LW |
387 | return any arbitrary Perl value. If the corresponding operation for this value |
388 | is overloaded too, that operation will be called again with this value. | |
389 | ||
1554e226 DC |
390 | As a special case if the overload returns the object itself then it will |
391 | be used directly. An overloaded conversion returning the object is | |
392 | probably a bug, because you're likely to get something that looks like | |
393 | C<YourPackage=HASH(0x8172b34)>. | |
394 | ||
f5284f61 IZ |
395 | =item * I<Iteration> |
396 | ||
397 | "<>" | |
398 | ||
399 | If not overloaded, the argument will be converted to a filehandle or | |
400 | glob (which may require a stringification). The same overloading | |
401 | happens both for the I<read-filehandle> syntax C<E<lt>$varE<gt>> and | |
402 | I<globbing> syntax C<E<lt>${var}E<gt>>. | |
403 | ||
54f8c773 YST |
404 | B<BUGS> Even in list context, the iterator is currently called only |
405 | once and with scalar context. | |
406 | ||
f5284f61 IZ |
407 | =item * I<Dereferencing> |
408 | ||
409 | '${}', '@{}', '%{}', '&{}', '*{}'. | |
410 | ||
411 | If not overloaded, the argument will be dereferenced I<as is>, thus | |
412 | should be of correct type. These functions should return a reference | |
413 | of correct type, or another object with overloaded dereferencing. | |
414 | ||
b267980d NIS |
415 | As a special case if the overload returns the object itself then it |
416 | will be used directly (provided it is the correct type). | |
417 | ||
418 | The dereference operators must be specified explicitly they will not be passed to | |
419 | "nomethod". | |
420 | ||
4633a7c4 LW |
421 | =item * I<Special> |
422 | ||
0d863452 | 423 | "nomethod", "fallback", "=", "~~", |
4633a7c4 LW |
424 | |
425 | see L<SPECIAL SYMBOLS FOR C<use overload>>. | |
426 | ||
427 | =back | |
428 | ||
ee239bfe IZ |
429 | See L<"Fallback"> for an explanation of when a missing method can be |
430 | autogenerated. | |
431 | ||
432 | A computer-readable form of the above table is available in the hash | |
433 | %overload::ops, with values being space-separated lists of names: | |
434 | ||
435 | with_assign => '+ - * / % ** << >> x .', | |
436 | assign => '+= -= *= /= %= **= <<= >>= x= .=', | |
2877bd81 | 437 | num_comparison => '< <= > >= == !=', |
ee239bfe | 438 | '3way_comparison'=> '<=> cmp', |
2877bd81 | 439 | str_comparison => 'lt le gt ge eq ne', |
ee239bfe IZ |
440 | binary => '& | ^', |
441 | unary => 'neg ! ~', | |
442 | mutators => '++ --', | |
443 | func => 'atan2 cos sin exp abs log sqrt', | |
444 | conversion => 'bool "" 0+', | |
f5284f61 IZ |
445 | iterators => '<>', |
446 | dereferencing => '${} @{} %{} &{} *{}', | |
ee239bfe | 447 | special => 'nomethod fallback =' |
4633a7c4 | 448 | |
e7ea3e70 IZ |
449 | =head2 Inheritance and overloading |
450 | ||
774d564b | 451 | Inheritance interacts with overloading in two ways. |
e7ea3e70 | 452 | |
88c28ceb | 453 | =over |
e7ea3e70 IZ |
454 | |
455 | =item Strings as values of C<use overload> directive | |
456 | ||
774d564b | 457 | If C<value> in |
e7ea3e70 IZ |
458 | |
459 | use overload key => value; | |
460 | ||
774d564b | 461 | is a string, it is interpreted as a method name. |
e7ea3e70 IZ |
462 | |
463 | =item Overloading of an operation is inherited by derived classes | |
464 | ||
774d564b | 465 | Any class derived from an overloaded class is also overloaded. The |
466 | set of overloaded methods is the union of overloaded methods of all | |
467 | the ancestors. If some method is overloaded in several ancestor, then | |
e7ea3e70 | 468 | which description will be used is decided by the usual inheritance |
774d564b | 469 | rules: |
e7ea3e70 | 470 | |
774d564b | 471 | If C<A> inherits from C<B> and C<C> (in this order), C<B> overloads |
472 | C<+> with C<\&D::plus_sub>, and C<C> overloads C<+> by C<"plus_meth">, | |
473 | then the subroutine C<D::plus_sub> will be called to implement | |
474 | operation C<+> for an object in package C<A>. | |
e7ea3e70 IZ |
475 | |
476 | =back | |
477 | ||
774d564b | 478 | Note that since the value of the C<fallback> key is not a subroutine, |
479 | its inheritance is not governed by the above rules. In the current | |
480 | implementation, the value of C<fallback> in the first overloaded | |
481 | ancestor is used, but this is accidental and subject to change. | |
e7ea3e70 | 482 | |
4633a7c4 LW |
483 | =head1 SPECIAL SYMBOLS FOR C<use overload> |
484 | ||
485 | Three keys are recognized by Perl that are not covered by the above | |
486 | description. | |
487 | ||
774d564b | 488 | =head2 Last Resort |
4633a7c4 LW |
489 | |
490 | C<"nomethod"> should be followed by a reference to a function of four | |
491 | parameters. If defined, it is called when the overloading mechanism | |
492 | cannot find a method for some operation. The first three arguments of | |
493 | this function coincide with the arguments for the corresponding method if | |
494 | it were found, the fourth argument is the symbol | |
495 | corresponding to the missing method. If several methods are tried, | |
496 | the last one is used. Say, C<1-$a> can be equivalent to | |
497 | ||
498 | &nomethodMethod($a,1,1,"-") | |
499 | ||
500 | if the pair C<"nomethod" =E<gt> "nomethodMethod"> was specified in the | |
501 | C<use overload> directive. | |
502 | ||
b267980d NIS |
503 | The C<"nomethod"> mechanism is I<not> used for the dereference operators |
504 | ( ${} @{} %{} &{} *{} ). | |
505 | ||
506 | ||
4633a7c4 LW |
507 | If some operation cannot be resolved, and there is no function |
508 | assigned to C<"nomethod">, then an exception will be raised via die()-- | |
509 | unless C<"fallback"> was specified as a key in C<use overload> directive. | |
510 | ||
b267980d NIS |
511 | |
512 | =head2 Fallback | |
4633a7c4 LW |
513 | |
514 | The key C<"fallback"> governs what to do if a method for a particular | |
515 | operation is not found. Three different cases are possible depending on | |
516 | the value of C<"fallback">: | |
517 | ||
518 | =over 16 | |
519 | ||
520 | =item * C<undef> | |
521 | ||
522 | Perl tries to use a | |
523 | substituted method (see L<MAGIC AUTOGENERATION>). If this fails, it | |
524 | then tries to calls C<"nomethod"> value; if missing, an exception | |
525 | will be raised. | |
526 | ||
527 | =item * TRUE | |
528 | ||
529 | The same as for the C<undef> value, but no exception is raised. Instead, | |
530 | it silently reverts to what it would have done were there no C<use overload> | |
531 | present. | |
532 | ||
533 | =item * defined, but FALSE | |
534 | ||
535 | No autogeneration is tried. Perl tries to call | |
b267980d | 536 | C<"nomethod"> value, and if this is missing, raises an exception. |
4633a7c4 LW |
537 | |
538 | =back | |
539 | ||
e7ea3e70 IZ |
540 | B<Note.> C<"fallback"> inheritance via @ISA is not carved in stone |
541 | yet, see L<"Inheritance and overloading">. | |
542 | ||
5e68dedd JD |
543 | =head2 Smart Match |
544 | ||
545 | The key C<"~~"> allows you to override the smart matching used by | |
546 | the switch construct. See L<feature>. | |
547 | ||
4633a7c4 LW |
548 | =head2 Copy Constructor |
549 | ||
550 | The value for C<"="> is a reference to a function with three | |
551 | arguments, i.e., it looks like the other values in C<use | |
552 | overload>. However, it does not overload the Perl assignment | |
553 | operator. This would go against Camel hair. | |
554 | ||
555 | This operation is called in the situations when a mutator is applied | |
556 | to a reference that shares its object with some other reference, such | |
557 | as | |
558 | ||
b267980d | 559 | $a=$b; |
ee239bfe | 560 | ++$a; |
4633a7c4 LW |
561 | |
562 | To make this change $a and not change $b, a copy of C<$$a> is made, | |
563 | and $a is assigned a reference to this new object. This operation is | |
ee239bfe | 564 | done during execution of the C<++$a>, and not during the assignment, |
4633a7c4 | 565 | (so before the increment C<$$a> coincides with C<$$b>). This is only |
ee239bfe IZ |
566 | done if C<++> is expressed via a method for C<'++'> or C<'+='> (or |
567 | C<nomethod>). Note that if this operation is expressed via C<'+'> | |
568 | a nonmutator, i.e., as in | |
4633a7c4 | 569 | |
b267980d | 570 | $a=$b; |
4633a7c4 LW |
571 | $a=$a+1; |
572 | ||
573 | then C<$a> does not reference a new copy of C<$$a>, since $$a does not | |
574 | appear as lvalue when the above code is executed. | |
575 | ||
576 | If the copy constructor is required during the execution of some mutator, | |
577 | but a method for C<'='> was not specified, it can be autogenerated as a | |
578 | string copy if the object is a plain scalar. | |
579 | ||
580 | =over 5 | |
581 | ||
582 | =item B<Example> | |
583 | ||
b267980d | 584 | The actually executed code for |
4633a7c4 | 585 | |
b267980d | 586 | $a=$b; |
4633a7c4 LW |
587 | Something else which does not modify $a or $b.... |
588 | ++$a; | |
589 | ||
590 | may be | |
591 | ||
b267980d | 592 | $a=$b; |
4633a7c4 LW |
593 | Something else which does not modify $a or $b.... |
594 | $a = $a->clone(undef,""); | |
595 | $a->incr(undef,""); | |
596 | ||
597 | if $b was mathemagical, and C<'++'> was overloaded with C<\&incr>, | |
598 | C<'='> was overloaded with C<\&clone>. | |
599 | ||
600 | =back | |
601 | ||
f610777f | 602 | Same behaviour is triggered by C<$b = $a++>, which is consider a synonym for |
ee239bfe IZ |
603 | C<$b = $a; ++$a>. |
604 | ||
4633a7c4 LW |
605 | =head1 MAGIC AUTOGENERATION |
606 | ||
607 | If a method for an operation is not found, and the value for C<"fallback"> is | |
608 | TRUE or undefined, Perl tries to autogenerate a substitute method for | |
609 | the missing operation based on the defined operations. Autogenerated method | |
610 | substitutions are possible for the following operations: | |
611 | ||
612 | =over 16 | |
613 | ||
614 | =item I<Assignment forms of arithmetic operations> | |
615 | ||
616 | C<$a+=$b> can use the method for C<"+"> if the method for C<"+="> | |
617 | is not defined. | |
618 | ||
b267980d | 619 | =item I<Conversion operations> |
4633a7c4 LW |
620 | |
621 | String, numeric, and boolean conversion are calculated in terms of one | |
622 | another if not all of them are defined. | |
623 | ||
624 | =item I<Increment and decrement> | |
625 | ||
626 | The C<++$a> operation can be expressed in terms of C<$a+=1> or C<$a+1>, | |
627 | and C<$a--> in terms of C<$a-=1> and C<$a-1>. | |
628 | ||
629 | =item C<abs($a)> | |
630 | ||
631 | can be expressed in terms of C<$aE<lt>0> and C<-$a> (or C<0-$a>). | |
632 | ||
633 | =item I<Unary minus> | |
634 | ||
635 | can be expressed in terms of subtraction. | |
636 | ||
3bc6ec80 | 637 | =item I<Negation> |
638 | ||
639 | C<!> and C<not> can be expressed in terms of boolean conversion, or | |
640 | string or numerical conversion. | |
641 | ||
4633a7c4 LW |
642 | =item I<Concatenation> |
643 | ||
644 | can be expressed in terms of string conversion. | |
645 | ||
b267980d | 646 | =item I<Comparison operations> |
4633a7c4 LW |
647 | |
648 | can be expressed in terms of its "spaceship" counterpart: either | |
649 | C<E<lt>=E<gt>> or C<cmp>: | |
1fef88e7 | 650 | |
4633a7c4 LW |
651 | <, >, <=, >=, ==, != in terms of <=> |
652 | lt, gt, le, ge, eq, ne in terms of cmp | |
653 | ||
f5284f61 IZ |
654 | =item I<Iterator> |
655 | ||
656 | <> in terms of builtin operations | |
657 | ||
658 | =item I<Dereferencing> | |
659 | ||
660 | ${} @{} %{} &{} *{} in terms of builtin operations | |
661 | ||
4633a7c4 LW |
662 | =item I<Copy operator> |
663 | ||
664 | can be expressed in terms of an assignment to the dereferenced value, if this | |
665 | value is a scalar and not a reference. | |
666 | ||
667 | =back | |
668 | ||
84fc275b S |
669 | =head1 Minimal set of overloaded operations |
670 | ||
671 | Since some operations can be automatically generated from others, there is | |
672 | a minimal set of operations that need to be overloaded in order to have | |
299476e0 S |
673 | the complete set of overloaded operations at one's disposal. |
674 | Of course, the autogenerated operations may not do exactly what the user | |
675 | expects. See L<MAGIC AUTOGENERATION> above. The minimal set is: | |
84fc275b S |
676 | |
677 | + - * / % ** << >> x | |
678 | <=> cmp | |
679 | & | ^ ~ | |
680 | atan2 cos sin exp log sqrt int | |
681 | ||
682 | Additionally, you need to define at least one of string, boolean or | |
299476e0 S |
683 | numeric conversions because any one can be used to emulate the others. |
684 | The string conversion can also be used to emulate concatenation. | |
84fc275b | 685 | |
ee239bfe | 686 | =head1 Losing overloading |
4633a7c4 LW |
687 | |
688 | The restriction for the comparison operation is that even if, for example, | |
689 | `C<cmp>' should return a blessed reference, the autogenerated `C<lt>' | |
690 | function will produce only a standard logical value based on the | |
691 | numerical value of the result of `C<cmp>'. In particular, a working | |
692 | numeric conversion is needed in this case (possibly expressed in terms of | |
693 | other conversions). | |
694 | ||
695 | Similarly, C<.=> and C<x=> operators lose their mathemagical properties | |
696 | if the string conversion substitution is applied. | |
697 | ||
698 | When you chop() a mathemagical object it is promoted to a string and its | |
699 | mathemagical properties are lost. The same can happen with other | |
700 | operations as well. | |
701 | ||
702 | =head1 Run-time Overloading | |
703 | ||
704 | Since all C<use> directives are executed at compile-time, the only way to | |
705 | change overloading during run-time is to | |
706 | ||
707 | eval 'use overload "+" => \&addmethod'; | |
708 | ||
709 | You can also use | |
710 | ||
711 | eval 'no overload "+", "--", "<="'; | |
712 | ||
713 | though the use of these constructs during run-time is questionable. | |
714 | ||
715 | =head1 Public functions | |
716 | ||
717 | Package C<overload.pm> provides the following public functions: | |
718 | ||
719 | =over 5 | |
720 | ||
721 | =item overload::StrVal(arg) | |
722 | ||
6a0e9e72 FD |
723 | Gives string value of C<arg> as in absence of stringify overloading. If you |
724 | are using this to get the address of a reference (useful for checking if two | |
725 | references point to the same thing) then you may be better off using | |
726 | C<Scalar::Util::refaddr()>, which is faster. | |
4633a7c4 LW |
727 | |
728 | =item overload::Overloaded(arg) | |
729 | ||
730 | Returns true if C<arg> is subject to overloading of some operations. | |
731 | ||
732 | =item overload::Method(obj,op) | |
733 | ||
734 | Returns C<undef> or a reference to the method that implements C<op>. | |
735 | ||
736 | =back | |
737 | ||
b3ac6de7 IZ |
738 | =head1 Overloading constants |
739 | ||
7adf7a02 | 740 | For some applications, the Perl parser mangles constants too much. |
bfce84ec | 741 | It is possible to hook into this process via C<overload::constant()> |
7adf7a02 | 742 | and C<overload::remove_constant()> functions. |
b3ac6de7 IZ |
743 | |
744 | These functions take a hash as an argument. The recognized keys of this hash | |
7adf7a02 | 745 | are: |
b3ac6de7 IZ |
746 | |
747 | =over 8 | |
748 | ||
749 | =item integer | |
750 | ||
751 | to overload integer constants, | |
752 | ||
753 | =item float | |
754 | ||
755 | to overload floating point constants, | |
756 | ||
757 | =item binary | |
758 | ||
759 | to overload octal and hexadecimal constants, | |
760 | ||
761 | =item q | |
762 | ||
763 | to overload C<q>-quoted strings, constant pieces of C<qq>- and C<qx>-quoted | |
764 | strings and here-documents, | |
765 | ||
766 | =item qr | |
767 | ||
768 | to overload constant pieces of regular expressions. | |
769 | ||
770 | =back | |
771 | ||
772 | The corresponding values are references to functions which take three arguments: | |
773 | the first one is the I<initial> string form of the constant, the second one | |
b267980d | 774 | is how Perl interprets this constant, the third one is how the constant is used. |
b3ac6de7 | 775 | Note that the initial string form does not |
b267980d | 776 | contain string delimiters, and has backslashes in backslash-delimiter |
b3ac6de7 | 777 | combinations stripped (thus the value of delimiter is not relevant for |
b267980d | 778 | processing of this string). The return value of this function is how this |
b3ac6de7 IZ |
779 | constant is going to be interpreted by Perl. The third argument is undefined |
780 | unless for overloaded C<q>- and C<qr>- constants, it is C<q> in single-quote | |
781 | context (comes from strings, regular expressions, and single-quote HERE | |
b267980d | 782 | documents), it is C<tr> for arguments of C<tr>/C<y> operators, |
b3ac6de7 IZ |
783 | it is C<s> for right-hand side of C<s>-operator, and it is C<qq> otherwise. |
784 | ||
785 | Since an expression C<"ab$cd,,"> is just a shortcut for C<'ab' . $cd . ',,'>, | |
786 | it is expected that overloaded constant strings are equipped with reasonable | |
b267980d | 787 | overloaded catenation operator, otherwise absurd results will result. |
b3ac6de7 IZ |
788 | Similarly, negative numbers are considered as negations of positive constants. |
789 | ||
790 | Note that it is probably meaningless to call the functions overload::constant() | |
791 | and overload::remove_constant() from anywhere but import() and unimport() methods. | |
792 | From these methods they may be called as | |
793 | ||
794 | sub import { | |
795 | shift; | |
796 | return unless @_; | |
797 | die "unknown import: @_" unless @_ == 1 and $_[0] eq ':constant'; | |
798 | overload::constant integer => sub {Math::BigInt->new(shift)}; | |
799 | } | |
800 | ||
4633a7c4 LW |
801 | =head1 IMPLEMENTATION |
802 | ||
803 | What follows is subject to change RSN. | |
804 | ||
e7ea3e70 IZ |
805 | The table of methods for all operations is cached in magic for the |
806 | symbol table hash for the package. The cache is invalidated during | |
807 | processing of C<use overload>, C<no overload>, new function | |
808 | definitions, and changes in @ISA. However, this invalidation remains | |
809 | unprocessed until the next C<bless>ing into the package. Hence if you | |
810 | want to change overloading structure dynamically, you'll need an | |
811 | additional (fake) C<bless>ing to update the table. | |
812 | ||
813 | (Every SVish thing has a magic queue, and magic is an entry in that | |
814 | queue. This is how a single variable may participate in multiple | |
815 | forms of magic simultaneously. For instance, environment variables | |
816 | regularly have two forms at once: their %ENV magic and their taint | |
817 | magic. However, the magic which implements overloading is applied to | |
818 | the stashes, which are rarely used directly, thus should not slow down | |
819 | Perl.) | |
4633a7c4 LW |
820 | |
821 | If an object belongs to a package using overload, it carries a special | |
822 | flag. Thus the only speed penalty during arithmetic operations without | |
823 | overloading is the checking of this flag. | |
824 | ||
774d564b | 825 | In fact, if C<use overload> is not present, there is almost no overhead |
826 | for overloadable operations, so most programs should not suffer | |
827 | measurable performance penalties. A considerable effort was made to | |
828 | minimize the overhead when overload is used in some package, but the | |
829 | arguments in question do not belong to packages using overload. When | |
830 | in doubt, test your speed with C<use overload> and without it. So far | |
831 | there have been no reports of substantial speed degradation if Perl is | |
832 | compiled with optimization turned on. | |
4633a7c4 | 833 | |
e7ea3e70 IZ |
834 | There is no size penalty for data if overload is not used. The only |
835 | size penalty if overload is used in some package is that I<all> the | |
836 | packages acquire a magic during the next C<bless>ing into the | |
837 | package. This magic is three-words-long for packages without | |
f610777f | 838 | overloading, and carries the cache table if the package is overloaded. |
4633a7c4 | 839 | |
b267980d | 840 | Copying (C<$a=$b>) is shallow; however, a one-level-deep copying is |
4633a7c4 LW |
841 | carried out before any operation that can imply an assignment to the |
842 | object $a (or $b) refers to, like C<$a++>. You can override this | |
843 | behavior by defining your own copy constructor (see L<"Copy Constructor">). | |
844 | ||
845 | It is expected that arguments to methods that are not explicitly supposed | |
846 | to be changed are constant (but this is not enforced). | |
847 | ||
ee239bfe IZ |
848 | =head1 Metaphor clash |
849 | ||
f610777f | 850 | One may wonder why the semantic of overloaded C<=> is so counter intuitive. |
b267980d NIS |
851 | If it I<looks> counter intuitive to you, you are subject to a metaphor |
852 | clash. | |
ee239bfe IZ |
853 | |
854 | Here is a Perl object metaphor: | |
855 | ||
856 | I< object is a reference to blessed data> | |
857 | ||
858 | and an arithmetic metaphor: | |
859 | ||
860 | I< object is a thing by itself>. | |
861 | ||
862 | The I<main> problem of overloading C<=> is the fact that these metaphors | |
863 | imply different actions on the assignment C<$a = $b> if $a and $b are | |
864 | objects. Perl-think implies that $a becomes a reference to whatever | |
865 | $b was referencing. Arithmetic-think implies that the value of "object" | |
866 | $a is changed to become the value of the object $b, preserving the fact | |
867 | that $a and $b are separate entities. | |
868 | ||
869 | The difference is not relevant in the absence of mutators. After | |
870 | a Perl-way assignment an operation which mutates the data referenced by $a | |
b267980d | 871 | would change the data referenced by $b too. Effectively, after |
ee239bfe IZ |
872 | C<$a = $b> values of $a and $b become I<indistinguishable>. |
873 | ||
b267980d | 874 | On the other hand, anyone who has used algebraic notation knows the |
ee239bfe IZ |
875 | expressive power of the arithmetic metaphor. Overloading works hard |
876 | to enable this metaphor while preserving the Perlian way as far as | |
d1be9408 | 877 | possible. Since it is not possible to freely mix two contradicting |
ee239bfe IZ |
878 | metaphors, overloading allows the arithmetic way to write things I<as |
879 | far as all the mutators are called via overloaded access only>. The | |
880 | way it is done is described in L<Copy Constructor>. | |
881 | ||
882 | If some mutator methods are directly applied to the overloaded values, | |
b267980d | 883 | one may need to I<explicitly unlink> other values which references the |
ee239bfe IZ |
884 | same value: |
885 | ||
886 | $a = new Data 23; | |
887 | ... | |
888 | $b = $a; # $b is "linked" to $a | |
889 | ... | |
890 | $a = $a->clone; # Unlink $b from $a | |
891 | $a->increment_by(4); | |
892 | ||
893 | Note that overloaded access makes this transparent: | |
894 | ||
895 | $a = new Data 23; | |
896 | $b = $a; # $b is "linked" to $a | |
897 | $a += 4; # would unlink $b automagically | |
898 | ||
899 | However, it would not make | |
900 | ||
901 | $a = new Data 23; | |
902 | $a = 4; # Now $a is a plain 4, not 'Data' | |
903 | ||
904 | preserve "objectness" of $a. But Perl I<has> a way to make assignments | |
905 | to an object do whatever you want. It is just not the overload, but | |
906 | tie()ing interface (see L<perlfunc/tie>). Adding a FETCH() method | |
b267980d | 907 | which returns the object itself, and STORE() method which changes the |
ee239bfe IZ |
908 | value of the object, one can reproduce the arithmetic metaphor in its |
909 | completeness, at least for variables which were tie()d from the start. | |
910 | ||
911 | (Note that a workaround for a bug may be needed, see L<"BUGS">.) | |
912 | ||
913 | =head1 Cookbook | |
914 | ||
915 | Please add examples to what follows! | |
916 | ||
917 | =head2 Two-face scalars | |
918 | ||
919 | Put this in F<two_face.pm> in your Perl library directory: | |
920 | ||
921 | package two_face; # Scalars with separate string and | |
922 | # numeric values. | |
923 | sub new { my $p = shift; bless [@_], $p } | |
924 | use overload '""' => \&str, '0+' => \&num, fallback => 1; | |
925 | sub num {shift->[1]} | |
926 | sub str {shift->[0]} | |
927 | ||
928 | Use it as follows: | |
929 | ||
930 | require two_face; | |
931 | my $seven = new two_face ("vii", 7); | |
932 | printf "seven=$seven, seven=%d, eight=%d\n", $seven, $seven+1; | |
933 | print "seven contains `i'\n" if $seven =~ /i/; | |
934 | ||
935 | (The second line creates a scalar which has both a string value, and a | |
936 | numeric value.) This prints: | |
937 | ||
938 | seven=vii, seven=7, eight=8 | |
939 | seven contains `i' | |
940 | ||
f5284f61 IZ |
941 | =head2 Two-face references |
942 | ||
943 | Suppose you want to create an object which is accessible as both an | |
6d822dc4 | 944 | array reference and a hash reference. |
f5284f61 IZ |
945 | |
946 | package two_refs; | |
947 | use overload '%{}' => \&gethash, '@{}' => sub { $ {shift()} }; | |
b267980d NIS |
948 | sub new { |
949 | my $p = shift; | |
f5284f61 IZ |
950 | bless \ [@_], $p; |
951 | } | |
952 | sub gethash { | |
953 | my %h; | |
954 | my $self = shift; | |
955 | tie %h, ref $self, $self; | |
956 | \%h; | |
957 | } | |
958 | ||
959 | sub TIEHASH { my $p = shift; bless \ shift, $p } | |
960 | my %fields; | |
961 | my $i = 0; | |
962 | $fields{$_} = $i++ foreach qw{zero one two three}; | |
b267980d | 963 | sub STORE { |
f5284f61 IZ |
964 | my $self = ${shift()}; |
965 | my $key = $fields{shift()}; | |
966 | defined $key or die "Out of band access"; | |
967 | $$self->[$key] = shift; | |
968 | } | |
b267980d | 969 | sub FETCH { |
f5284f61 IZ |
970 | my $self = ${shift()}; |
971 | my $key = $fields{shift()}; | |
972 | defined $key or die "Out of band access"; | |
973 | $$self->[$key]; | |
974 | } | |
975 | ||
976 | Now one can access an object using both the array and hash syntax: | |
977 | ||
978 | my $bar = new two_refs 3,4,5,6; | |
979 | $bar->[2] = 11; | |
980 | $bar->{two} == 11 or die 'bad hash fetch'; | |
981 | ||
982 | Note several important features of this example. First of all, the | |
983 | I<actual> type of $bar is a scalar reference, and we do not overload | |
984 | the scalar dereference. Thus we can get the I<actual> non-overloaded | |
985 | contents of $bar by just using C<$$bar> (what we do in functions which | |
986 | overload dereference). Similarly, the object returned by the | |
987 | TIEHASH() method is a scalar reference. | |
988 | ||
989 | Second, we create a new tied hash each time the hash syntax is used. | |
990 | This allows us not to worry about a possibility of a reference loop, | |
d1be9408 | 991 | which would lead to a memory leak. |
f5284f61 IZ |
992 | |
993 | Both these problems can be cured. Say, if we want to overload hash | |
994 | dereference on a reference to an object which is I<implemented> as a | |
995 | hash itself, the only problem one has to circumvent is how to access | |
1fd16925 | 996 | this I<actual> hash (as opposed to the I<virtual> hash exhibited by the |
f5284f61 IZ |
997 | overloaded dereference operator). Here is one possible fetching routine: |
998 | ||
999 | sub access_hash { | |
1000 | my ($self, $key) = (shift, shift); | |
1001 | my $class = ref $self; | |
b267980d | 1002 | bless $self, 'overload::dummy'; # Disable overloading of %{} |
f5284f61 IZ |
1003 | my $out = $self->{$key}; |
1004 | bless $self, $class; # Restore overloading | |
1005 | $out; | |
1006 | } | |
1007 | ||
1fd16925 | 1008 | To remove creation of the tied hash on each access, one may an extra |
f5284f61 IZ |
1009 | level of indirection which allows a non-circular structure of references: |
1010 | ||
1011 | package two_refs1; | |
1012 | use overload '%{}' => sub { ${shift()}->[1] }, | |
1013 | '@{}' => sub { ${shift()}->[0] }; | |
b267980d NIS |
1014 | sub new { |
1015 | my $p = shift; | |
f5284f61 IZ |
1016 | my $a = [@_]; |
1017 | my %h; | |
1018 | tie %h, $p, $a; | |
1019 | bless \ [$a, \%h], $p; | |
1020 | } | |
1021 | sub gethash { | |
1022 | my %h; | |
1023 | my $self = shift; | |
1024 | tie %h, ref $self, $self; | |
1025 | \%h; | |
1026 | } | |
1027 | ||
1028 | sub TIEHASH { my $p = shift; bless \ shift, $p } | |
1029 | my %fields; | |
1030 | my $i = 0; | |
1031 | $fields{$_} = $i++ foreach qw{zero one two three}; | |
b267980d | 1032 | sub STORE { |
f5284f61 IZ |
1033 | my $a = ${shift()}; |
1034 | my $key = $fields{shift()}; | |
1035 | defined $key or die "Out of band access"; | |
1036 | $a->[$key] = shift; | |
1037 | } | |
b267980d | 1038 | sub FETCH { |
f5284f61 IZ |
1039 | my $a = ${shift()}; |
1040 | my $key = $fields{shift()}; | |
1041 | defined $key or die "Out of band access"; | |
1042 | $a->[$key]; | |
1043 | } | |
1044 | ||
1fd16925 | 1045 | Now if $baz is overloaded like this, then C<$baz> is a reference to a |
f5284f61 IZ |
1046 | reference to the intermediate array, which keeps a reference to an |
1047 | actual array, and the access hash. The tie()ing object for the access | |
1fd16925 | 1048 | hash is a reference to a reference to the actual array, so |
f5284f61 | 1049 | |
88c28ceb | 1050 | =over |
f5284f61 IZ |
1051 | |
1052 | =item * | |
1053 | ||
1054 | There are no loops of references. | |
1055 | ||
1056 | =item * | |
1057 | ||
1058 | Both "objects" which are blessed into the class C<two_refs1> are | |
1059 | references to a reference to an array, thus references to a I<scalar>. | |
1060 | Thus the accessor expression C<$$foo-E<gt>[$ind]> involves no | |
1061 | overloaded operations. | |
1062 | ||
1063 | =back | |
1064 | ||
ee239bfe IZ |
1065 | =head2 Symbolic calculator |
1066 | ||
1067 | Put this in F<symbolic.pm> in your Perl library directory: | |
1068 | ||
1069 | package symbolic; # Primitive symbolic calculator | |
1070 | use overload nomethod => \&wrap; | |
1071 | ||
1072 | sub new { shift; bless ['n', @_] } | |
1073 | sub wrap { | |
1074 | my ($obj, $other, $inv, $meth) = @_; | |
1075 | ($obj, $other) = ($other, $obj) if $inv; | |
1076 | bless [$meth, $obj, $other]; | |
1077 | } | |
1078 | ||
1079 | This module is very unusual as overloaded modules go: it does not | |
88c28ceb JH |
1080 | provide any usual overloaded operators, instead it provides the L<Last |
1081 | Resort> operator C<nomethod>. In this example the corresponding | |
f610777f | 1082 | subroutine returns an object which encapsulates operations done over |
ee239bfe IZ |
1083 | the objects: C<new symbolic 3> contains C<['n', 3]>, C<2 + new |
1084 | symbolic 3> contains C<['+', 2, ['n', 3]]>. | |
1085 | ||
1086 | Here is an example of the script which "calculates" the side of | |
1087 | circumscribed octagon using the above package: | |
1088 | ||
1089 | require symbolic; | |
1090 | my $iter = 1; # 2**($iter+2) = 8 | |
1091 | my $side = new symbolic 1; | |
1092 | my $cnt = $iter; | |
3cb6de81 | 1093 | |
ee239bfe IZ |
1094 | while ($cnt--) { |
1095 | $side = (sqrt(1 + $side**2) - 1)/$side; | |
1096 | } | |
1097 | print "OK\n"; | |
1098 | ||
1099 | The value of $side is | |
1100 | ||
1101 | ['/', ['-', ['sqrt', ['+', 1, ['**', ['n', 1], 2]], | |
1102 | undef], 1], ['n', 1]] | |
1103 | ||
1104 | Note that while we obtained this value using a nice little script, | |
1105 | there is no simple way to I<use> this value. In fact this value may | |
2d3232d7 | 1106 | be inspected in debugger (see L<perldebug>), but only if |
ee239bfe IZ |
1107 | C<bareStringify> B<O>ption is set, and not via C<p> command. |
1108 | ||
1109 | If one attempts to print this value, then the overloaded operator | |
1110 | C<""> will be called, which will call C<nomethod> operator. The | |
1111 | result of this operator will be stringified again, but this result is | |
1112 | again of type C<symbolic>, which will lead to an infinite loop. | |
1113 | ||
1114 | Add a pretty-printer method to the module F<symbolic.pm>: | |
1115 | ||
1116 | sub pretty { | |
1117 | my ($meth, $a, $b) = @{+shift}; | |
1118 | $a = 'u' unless defined $a; | |
1119 | $b = 'u' unless defined $b; | |
1120 | $a = $a->pretty if ref $a; | |
1121 | $b = $b->pretty if ref $b; | |
1122 | "[$meth $a $b]"; | |
b267980d | 1123 | } |
ee239bfe IZ |
1124 | |
1125 | Now one can finish the script by | |
1126 | ||
1127 | print "side = ", $side->pretty, "\n"; | |
1128 | ||
1129 | The method C<pretty> is doing object-to-string conversion, so it | |
1130 | is natural to overload the operator C<""> using this method. However, | |
1131 | inside such a method it is not necessary to pretty-print the | |
1132 | I<components> $a and $b of an object. In the above subroutine | |
1133 | C<"[$meth $a $b]"> is a catenation of some strings and components $a | |
1134 | and $b. If these components use overloading, the catenation operator | |
1fd16925 | 1135 | will look for an overloaded operator C<.>; if not present, it will |
ee239bfe IZ |
1136 | look for an overloaded operator C<"">. Thus it is enough to use |
1137 | ||
1138 | use overload nomethod => \&wrap, '""' => \&str; | |
1139 | sub str { | |
1140 | my ($meth, $a, $b) = @{+shift}; | |
1141 | $a = 'u' unless defined $a; | |
1142 | $b = 'u' unless defined $b; | |
1143 | "[$meth $a $b]"; | |
b267980d | 1144 | } |
ee239bfe IZ |
1145 | |
1146 | Now one can change the last line of the script to | |
1147 | ||
1148 | print "side = $side\n"; | |
1149 | ||
1150 | which outputs | |
1151 | ||
1152 | side = [/ [- [sqrt [+ 1 [** [n 1 u] 2]] u] 1] [n 1 u]] | |
1153 | ||
1154 | and one can inspect the value in debugger using all the possible | |
b267980d | 1155 | methods. |
ee239bfe | 1156 | |
d1be9408 | 1157 | Something is still amiss: consider the loop variable $cnt of the |
ee239bfe IZ |
1158 | script. It was a number, not an object. We cannot make this value of |
1159 | type C<symbolic>, since then the loop will not terminate. | |
1160 | ||
1161 | Indeed, to terminate the cycle, the $cnt should become false. | |
1162 | However, the operator C<bool> for checking falsity is overloaded (this | |
1163 | time via overloaded C<"">), and returns a long string, thus any object | |
1164 | of type C<symbolic> is true. To overcome this, we need a way to | |
1165 | compare an object to 0. In fact, it is easier to write a numeric | |
1166 | conversion routine. | |
1167 | ||
1168 | Here is the text of F<symbolic.pm> with such a routine added (and | |
f610777f | 1169 | slightly modified str()): |
ee239bfe IZ |
1170 | |
1171 | package symbolic; # Primitive symbolic calculator | |
1172 | use overload | |
1173 | nomethod => \&wrap, '""' => \&str, '0+' => \# | |
1174 | ||
1175 | sub new { shift; bless ['n', @_] } | |
1176 | sub wrap { | |
1177 | my ($obj, $other, $inv, $meth) = @_; | |
1178 | ($obj, $other) = ($other, $obj) if $inv; | |
1179 | bless [$meth, $obj, $other]; | |
1180 | } | |
1181 | sub str { | |
1182 | my ($meth, $a, $b) = @{+shift}; | |
1183 | $a = 'u' unless defined $a; | |
1184 | if (defined $b) { | |
1185 | "[$meth $a $b]"; | |
1186 | } else { | |
1187 | "[$meth $a]"; | |
1188 | } | |
b267980d NIS |
1189 | } |
1190 | my %subr = ( n => sub {$_[0]}, | |
1191 | sqrt => sub {sqrt $_[0]}, | |
ee239bfe IZ |
1192 | '-' => sub {shift() - shift()}, |
1193 | '+' => sub {shift() + shift()}, | |
1194 | '/' => sub {shift() / shift()}, | |
1195 | '*' => sub {shift() * shift()}, | |
1196 | '**' => sub {shift() ** shift()}, | |
1197 | ); | |
1198 | sub num { | |
1199 | my ($meth, $a, $b) = @{+shift}; | |
b267980d | 1200 | my $subr = $subr{$meth} |
ee239bfe IZ |
1201 | or die "Do not know how to ($meth) in symbolic"; |
1202 | $a = $a->num if ref $a eq __PACKAGE__; | |
1203 | $b = $b->num if ref $b eq __PACKAGE__; | |
1204 | $subr->($a,$b); | |
1205 | } | |
1206 | ||
1207 | All the work of numeric conversion is done in %subr and num(). Of | |
f610777f | 1208 | course, %subr is not complete, it contains only operators used in the |
ee239bfe IZ |
1209 | example below. Here is the extra-credit question: why do we need an |
1210 | explicit recursion in num()? (Answer is at the end of this section.) | |
1211 | ||
1212 | Use this module like this: | |
1213 | ||
1214 | require symbolic; | |
1215 | my $iter = new symbolic 2; # 16-gon | |
1216 | my $side = new symbolic 1; | |
1217 | my $cnt = $iter; | |
3cb6de81 | 1218 | |
ee239bfe IZ |
1219 | while ($cnt) { |
1220 | $cnt = $cnt - 1; # Mutator `--' not implemented | |
1221 | $side = (sqrt(1 + $side**2) - 1)/$side; | |
1222 | } | |
1223 | printf "%s=%f\n", $side, $side; | |
1224 | printf "pi=%f\n", $side*(2**($iter+2)); | |
1225 | ||
1226 | It prints (without so many line breaks) | |
1227 | ||
1228 | [/ [- [sqrt [+ 1 [** [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] | |
1229 | [n 1]] 2]]] 1] | |
1230 | [/ [- [sqrt [+ 1 [** [n 1] 2]]] 1] [n 1]]]=0.198912 | |
1231 | pi=3.182598 | |
1232 | ||
1233 | The above module is very primitive. It does not implement | |
1234 | mutator methods (C<++>, C<-=> and so on), does not do deep copying | |
1235 | (not required without mutators!), and implements only those arithmetic | |
1236 | operations which are used in the example. | |
1237 | ||
1fd16925 | 1238 | To implement most arithmetic operations is easy; one should just use |
ee239bfe IZ |
1239 | the tables of operations, and change the code which fills %subr to |
1240 | ||
1241 | my %subr = ( 'n' => sub {$_[0]} ); | |
1242 | foreach my $op (split " ", $overload::ops{with_assign}) { | |
1243 | $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}"; | |
1244 | } | |
1245 | my @bins = qw(binary 3way_comparison num_comparison str_comparison); | |
1246 | foreach my $op (split " ", "@overload::ops{ @bins }") { | |
1247 | $subr{$op} = eval "sub {shift() $op shift()}"; | |
1248 | } | |
1249 | foreach my $op (split " ", "@overload::ops{qw(unary func)}") { | |
1250 | print "defining `$op'\n"; | |
1251 | $subr{$op} = eval "sub {$op shift()}"; | |
1252 | } | |
1253 | ||
1254 | Due to L<Calling Conventions for Mutators>, we do not need anything | |
1255 | special to make C<+=> and friends work, except filling C<+=> entry of | |
1256 | %subr, and defining a copy constructor (needed since Perl has no | |
1257 | way to know that the implementation of C<'+='> does not mutate | |
1258 | the argument, compare L<Copy Constructor>). | |
1259 | ||
1fd16925 | 1260 | To implement a copy constructor, add C<< '=' => \&cpy >> to C<use overload> |
ee239bfe IZ |
1261 | line, and code (this code assumes that mutators change things one level |
1262 | deep only, so recursive copying is not needed): | |
1263 | ||
1264 | sub cpy { | |
1265 | my $self = shift; | |
1266 | bless [@$self], ref $self; | |
1267 | } | |
1268 | ||
b267980d | 1269 | To make C<++> and C<--> work, we need to implement actual mutators, |
ee239bfe IZ |
1270 | either directly, or in C<nomethod>. We continue to do things inside |
1271 | C<nomethod>, thus add | |
1272 | ||
1273 | if ($meth eq '++' or $meth eq '--') { | |
1274 | @$obj = ($meth, (bless [@$obj]), 1); # Avoid circular reference | |
1275 | return $obj; | |
1276 | } | |
1277 | ||
b267980d | 1278 | after the first line of wrap(). This is not a most effective |
ee239bfe IZ |
1279 | implementation, one may consider |
1280 | ||
1281 | sub inc { $_[0] = bless ['++', shift, 1]; } | |
1282 | ||
1283 | instead. | |
1284 | ||
1285 | As a final remark, note that one can fill %subr by | |
1286 | ||
1287 | my %subr = ( 'n' => sub {$_[0]} ); | |
1288 | foreach my $op (split " ", $overload::ops{with_assign}) { | |
1289 | $subr{$op} = $subr{"$op="} = eval "sub {shift() $op shift()}"; | |
1290 | } | |
1291 | my @bins = qw(binary 3way_comparison num_comparison str_comparison); | |
1292 | foreach my $op (split " ", "@overload::ops{ @bins }") { | |
1293 | $subr{$op} = eval "sub {shift() $op shift()}"; | |
1294 | } | |
1295 | foreach my $op (split " ", "@overload::ops{qw(unary func)}") { | |
1296 | $subr{$op} = eval "sub {$op shift()}"; | |
1297 | } | |
1298 | $subr{'++'} = $subr{'+'}; | |
1299 | $subr{'--'} = $subr{'-'}; | |
1300 | ||
b267980d NIS |
1301 | This finishes implementation of a primitive symbolic calculator in |
1302 | 50 lines of Perl code. Since the numeric values of subexpressions | |
ee239bfe IZ |
1303 | are not cached, the calculator is very slow. |
1304 | ||
1305 | Here is the answer for the exercise: In the case of str(), we need no | |
1306 | explicit recursion since the overloaded C<.>-operator will fall back | |
1307 | to an existing overloaded operator C<"">. Overloaded arithmetic | |
1308 | operators I<do not> fall back to numeric conversion if C<fallback> is | |
1309 | not explicitly requested. Thus without an explicit recursion num() | |
1310 | would convert C<['+', $a, $b]> to C<$a + $b>, which would just rebuild | |
1311 | the argument of num(). | |
1312 | ||
1313 | If you wonder why defaults for conversion are different for str() and | |
1314 | num(), note how easy it was to write the symbolic calculator. This | |
1315 | simplicity is due to an appropriate choice of defaults. One extra | |
f610777f A |
1316 | note: due to the explicit recursion num() is more fragile than sym(): |
1317 | we need to explicitly check for the type of $a and $b. If components | |
ee239bfe IZ |
1318 | $a and $b happen to be of some related type, this may lead to problems. |
1319 | ||
1320 | =head2 I<Really> symbolic calculator | |
1321 | ||
1322 | One may wonder why we call the above calculator symbolic. The reason | |
1323 | is that the actual calculation of the value of expression is postponed | |
1324 | until the value is I<used>. | |
1325 | ||
1326 | To see it in action, add a method | |
1327 | ||
b267980d NIS |
1328 | sub STORE { |
1329 | my $obj = shift; | |
1330 | $#$obj = 1; | |
ee239bfe IZ |
1331 | @$obj->[0,1] = ('=', shift); |
1332 | } | |
1333 | ||
1334 | to the package C<symbolic>. After this change one can do | |
1335 | ||
1336 | my $a = new symbolic 3; | |
1337 | my $b = new symbolic 4; | |
1338 | my $c = sqrt($a**2 + $b**2); | |
1339 | ||
1340 | and the numeric value of $c becomes 5. However, after calling | |
1341 | ||
1342 | $a->STORE(12); $b->STORE(5); | |
1343 | ||
1344 | the numeric value of $c becomes 13. There is no doubt now that the module | |
1345 | symbolic provides a I<symbolic> calculator indeed. | |
1346 | ||
1347 | To hide the rough edges under the hood, provide a tie()d interface to the | |
1348 | package C<symbolic> (compare with L<Metaphor clash>). Add methods | |
1349 | ||
1350 | sub TIESCALAR { my $pack = shift; $pack->new(@_) } | |
1351 | sub FETCH { shift } | |
1352 | sub nop { } # Around a bug | |
1353 | ||
1354 | (the bug is described in L<"BUGS">). One can use this new interface as | |
1355 | ||
1356 | tie $a, 'symbolic', 3; | |
1357 | tie $b, 'symbolic', 4; | |
1358 | $a->nop; $b->nop; # Around a bug | |
1359 | ||
1360 | my $c = sqrt($a**2 + $b**2); | |
1361 | ||
1362 | Now numeric value of $c is 5. After C<$a = 12; $b = 5> the numeric value | |
1363 | of $c becomes 13. To insulate the user of the module add a method | |
1364 | ||
1365 | sub vars { my $p = shift; tie($_, $p), $_->nop foreach @_; } | |
1366 | ||
1367 | Now | |
1368 | ||
1369 | my ($a, $b); | |
1370 | symbolic->vars($a, $b); | |
1371 | my $c = sqrt($a**2 + $b**2); | |
1372 | ||
1373 | $a = 3; $b = 4; | |
1374 | printf "c5 %s=%f\n", $c, $c; | |
1375 | ||
1376 | $a = 12; $b = 5; | |
1377 | printf "c13 %s=%f\n", $c, $c; | |
1378 | ||
1379 | shows that the numeric value of $c follows changes to the values of $a | |
1380 | and $b. | |
1381 | ||
4633a7c4 LW |
1382 | =head1 AUTHOR |
1383 | ||
1fef88e7 | 1384 | Ilya Zakharevich E<lt>F<ilya@math.mps.ohio-state.edu>E<gt>. |
4633a7c4 LW |
1385 | |
1386 | =head1 DIAGNOSTICS | |
1387 | ||
1388 | When Perl is run with the B<-Do> switch or its equivalent, overloading | |
1389 | induces diagnostic messages. | |
1390 | ||
e7ea3e70 IZ |
1391 | Using the C<m> command of Perl debugger (see L<perldebug>) one can |
1392 | deduce which operations are overloaded (and which ancestor triggers | |
1393 | this overloading). Say, if C<eq> is overloaded, then the method C<(eq> | |
1394 | is shown by debugger. The method C<()> corresponds to the C<fallback> | |
1395 | key (in fact a presence of this method shows that this package has | |
1396 | overloading enabled, and it is what is used by the C<Overloaded> | |
ee239bfe | 1397 | function of module C<overload>). |
e7ea3e70 | 1398 | |
6ad11d81 | 1399 | The module might issue the following warnings: |
6b82e2f5 A |
1400 | |
1401 | =over 4 | |
1402 | ||
1403 | =item Odd number of arguments for overload::constant | |
1404 | ||
1405 | (W) The call to overload::constant contained an odd number of arguments. | |
1406 | The arguments should come in pairs. | |
1407 | ||
1408 | =item `%s' is not an overloadable type | |
1409 | ||
1410 | (W) You tried to overload a constant type the overload package is unaware of. | |
1411 | ||
1412 | =item `%s' is not a code reference | |
1413 | ||
1414 | (W) The second (fourth, sixth, ...) argument of overload::constant needs | |
1415 | to be a code reference. Either an anonymous subroutine, or a reference | |
1416 | to a subroutine. | |
1417 | ||
1418 | =back | |
1419 | ||
4633a7c4 LW |
1420 | =head1 BUGS |
1421 | ||
aa689395 | 1422 | Because it is used for overloading, the per-package hash %OVERLOAD now |
1423 | has a special meaning in Perl. The symbol table is filled with names | |
1424 | looking like line-noise. | |
4633a7c4 | 1425 | |
a6006777 | 1426 | For the purpose of inheritance every overloaded package behaves as if |
1427 | C<fallback> is present (possibly undefined). This may create | |
1428 | interesting effects if some package is not overloaded, but inherits | |
1429 | from two overloaded packages. | |
4633a7c4 | 1430 | |
b267980d | 1431 | Relation between overloading and tie()ing is broken. Overloading is |
ee239bfe IZ |
1432 | triggered or not basing on the I<previous> class of tie()d value. |
1433 | ||
b267980d | 1434 | This happens because the presence of overloading is checked too early, |
ee239bfe | 1435 | before any tie()d access is attempted. If the FETCH()ed class of the |
b267980d | 1436 | tie()d value does not change, a simple workaround is to access the value |
ee239bfe IZ |
1437 | immediately after tie()ing, so that after this call the I<previous> class |
1438 | coincides with the current one. | |
1439 | ||
1440 | B<Needed:> a way to fix this without a speed penalty. | |
1441 | ||
b3ac6de7 IZ |
1442 | Barewords are not covered by overloaded string constants. |
1443 | ||
ee239bfe IZ |
1444 | This document is confusing. There are grammos and misleading language |
1445 | used in places. It would seem a total rewrite is needed. | |
4633a7c4 LW |
1446 | |
1447 | =cut | |
1448 |