Commit | Line | Data |
---|---|---|
07d83abe MV |
1 | @c -*-texinfo-*- |
2 | @c This is part of the GNU Guile Reference Manual. | |
7545ddd4 | 3 | @c Copyright (C) 1996, 1997, 2000, 2001, 2002, 2003, 2004, 2009, 2010 |
07d83abe MV |
4 | @c Free Software Foundation, Inc. |
5 | @c See the file guile.texi for copying conditions. | |
6 | ||
07d83abe MV |
7 | @node Memory Management |
8 | @section Memory Management and Garbage Collection | |
9 | ||
10 | Guile uses a @emph{garbage collector} to manage most of its objects. | |
11 | While the garbage collector is designed to be mostly invisible, you | |
877f06c3 | 12 | sometimes need to interact with it explicitly. |
07d83abe MV |
13 | |
14 | See @ref{Garbage Collection} for a general discussion of how garbage | |
15 | collection relates to using Guile from C. | |
16 | ||
17 | @menu | |
18 | * Garbage Collection Functions:: | |
19 | * Memory Blocks:: | |
20 | * Weak References:: | |
21 | * Guardians:: | |
22 | @end menu | |
23 | ||
24 | ||
25 | @node Garbage Collection Functions | |
26 | @subsection Function related to Garbage Collection | |
27 | ||
28 | @deffn {Scheme Procedure} gc | |
29 | @deffnx {C Function} scm_gc () | |
30 | Scans all of SCM objects and reclaims for further use those that are | |
31 | no longer accessible. You normally don't need to call this function | |
32 | explicitly. It is called automatically when appropriate. | |
33 | @end deffn | |
34 | ||
35 | @deftypefn {C Function} SCM scm_gc_protect_object (SCM @var{obj}) | |
36 | Protects @var{obj} from being freed by the garbage collector, when it | |
37 | otherwise might be. When you are done with the object, call | |
38 | @code{scm_gc_unprotect_object} on the object. Calls to | |
39 | @code{scm_gc_protect}/@code{scm_gc_unprotect_object} can be nested, and | |
40 | the object remains protected until it has been unprotected as many times | |
41 | as it was protected. It is an error to unprotect an object more times | |
42 | than it has been protected. Returns the SCM object it was passed. | |
f07c349e LC |
43 | |
44 | Note that storing @var{obj} in a C global variable has the same | |
45 | effect@footnote{In Guile up to version 1.8, C global variables were not | |
46 | scanned by the garbage collector; hence, @code{scm_gc_protect_object} | |
47 | was the only way in C to prevent a Scheme object from being freed.}. | |
07d83abe MV |
48 | @end deftypefn |
49 | ||
50 | @deftypefn {C Function} SCM scm_gc_unprotect_object (SCM @var{obj}) | |
51 | ||
52 | Unprotects an object from the garbage collector which was protected by | |
53 | @code{scm_gc_unprotect_object}. Returns the SCM object it was passed. | |
54 | @end deftypefn | |
55 | ||
56 | @deftypefn {C Function} SCM scm_permanent_object (SCM @var{obj}) | |
57 | ||
58 | Similar to @code{scm_gc_protect_object} in that it causes the | |
59 | collector to always mark the object, except that it should not be | |
60 | nested (only call @code{scm_permanent_object} on an object once), and | |
61 | it has no corresponding unpermanent function. Once an object is | |
62 | declared permanent, it will never be freed. Returns the SCM object it | |
63 | was passed. | |
64 | @end deftypefn | |
65 | ||
66 | @c NOTE: The varargs scm_remember_upto_here is deliberately not | |
67 | @c documented, because we don't think it can be implemented as a nice | |
68 | @c inline compiler directive or asm block. New _3, _4 or whatever | |
69 | @c forms could certainly be added though, if needed. | |
70 | ||
71 | @deftypefn {C Macro} void scm_remember_upto_here_1 (SCM obj) | |
72 | @deftypefnx {C Macro} void scm_remember_upto_here_2 (SCM obj1, SCM obj2) | |
73 | Create a reference to the given object or objects, so they're certain | |
74 | to be present on the stack or in a register and hence will not be | |
75 | freed by the garbage collector before this point. | |
76 | ||
77 | Note that these functions can only be applied to ordinary C local | |
78 | variables (ie.@: ``automatics''). Objects held in global or static | |
79 | variables or some malloced block or the like cannot be protected with | |
80 | this mechanism. | |
81 | @end deftypefn | |
82 | ||
83 | @deffn {Scheme Procedure} gc-stats | |
84 | @deffnx {C Function} scm_gc_stats () | |
85 | Return an association list of statistics about Guile's current | |
86 | use of storage. | |
c93557e7 | 87 | @end deffn |
07d83abe | 88 | |
673ba2da MV |
89 | @deffn {Scheme Procedure} gc-live-object-stats |
90 | @deffnx {C Function} scm_gc_live_object_stats () | |
91 | Return an alist of statistics of the current live objects. | |
92 | @end deffn | |
93 | ||
07d83abe MV |
94 | @deftypefun void scm_gc_mark (SCM @var{x}) |
95 | Mark the object @var{x}, and recurse on any objects @var{x} refers to. | |
96 | If @var{x}'s mark bit is already set, return immediately. This function | |
97 | must only be called during the mark-phase of garbage collection, | |
98 | typically from a smob @emph{mark} function. | |
99 | @end deftypefun | |
100 | ||
101 | ||
07d83abe MV |
102 | @node Memory Blocks |
103 | @subsection Memory Blocks | |
104 | ||
56273dea LC |
105 | @cindex automatically-managed memory |
106 | @cindex GC-managed memory | |
107 | @cindex conservative garbage collection | |
108 | ||
07d83abe MV |
109 | In C programs, dynamic management of memory blocks is normally done |
110 | with the functions malloc, realloc, and free. Guile has additional | |
111 | functions for dynamic memory allocation that are integrated into the | |
112 | garbage collector and the error reporting system. | |
113 | ||
114 | Memory blocks that are associated with Scheme objects (for example a | |
c5923112 | 115 | smob) should be allocated with @code{scm_gc_malloc} or |
56273dea LC |
116 | @code{scm_gc_malloc_pointerless}. These two functions will either |
117 | return a valid pointer or signal an error. Memory blocks allocated this | |
118 | way can be freed with @code{scm_gc_free}; however, this is not strictly | |
119 | needed: memory allocated with @code{scm_gc_malloc} or | |
120 | @code{scm_gc_malloc_pointerless} is automatically reclaimed when the | |
121 | garbage collector no longer sees any live reference to it@footnote{In | |
122 | Guile up to version 1.8, memory allocated with @code{scm_gc_malloc} | |
123 | @emph{had} to be freed with @code{scm_gc_free}.}. | |
124 | ||
125 | Memory allocated with @code{scm_gc_malloc} is scanned for live pointers. | |
126 | This means that if @code{scm_gc_malloc}-allocated memory contains a | |
127 | pointer to some other part of the memory, the garbage collector notices | |
128 | it and prevents it from being reclaimed@footnote{In Guile up to 1.8, | |
129 | memory allocated with @code{scm_gc_malloc} was @emph{not} scanned. | |
130 | Consequently, the GC had to be told explicitly about pointers to live | |
131 | objects contained in the memory block, e.g., @i{via} SMOB mark functions | |
132 | (@pxref{Smobs, @code{scm_set_smob_mark}})}. Conversely, memory | |
133 | allocated with @code{scm_gc_malloc_pointerless} is assumed to be | |
134 | ``pointer-less'' and is not scanned. | |
07d83abe MV |
135 | |
136 | For memory that is not associated with a Scheme object, you can use | |
137 | @code{scm_malloc} instead of @code{malloc}. Like | |
138 | @code{scm_gc_malloc}, it will either return a valid pointer or signal | |
139 | an error. However, it will not assume that the new memory block can | |
56273dea LC |
140 | be freed by a garbage collection. The memory must be explicitly freed |
141 | with @code{free}. | |
07d83abe MV |
142 | |
143 | There is also @code{scm_gc_realloc} and @code{scm_realloc}, to be used | |
ad0c2091 MV |
144 | in place of @code{realloc} when appropriate, and @code{scm_gc_calloc} |
145 | and @code{scm_calloc}, to be used in place of @code{calloc} when | |
07d83abe MV |
146 | appropriate. |
147 | ||
56273dea LC |
148 | The function @code{scm_dynwind_free} can be useful when memory should be |
149 | freed with libc's @code{free} when leaving a dynwind context, | |
150 | @xref{Dynamic Wind}. | |
07d83abe MV |
151 | |
152 | @deftypefn {C Function} {void *} scm_malloc (size_t @var{size}) | |
153 | @deftypefnx {C Function} {void *} scm_calloc (size_t @var{size}) | |
154 | Allocate @var{size} bytes of memory and return a pointer to it. When | |
155 | @var{size} is 0, return @code{NULL}. When not enough memory is | |
156 | available, signal an error. This function runs the GC to free up some | |
157 | memory when it deems it appropriate. | |
158 | ||
159 | The memory is allocated by the libc @code{malloc} function and can be | |
160 | freed with @code{free}. There is no @code{scm_free} function to go | |
161 | with @code{scm_malloc} to make it easier to pass memory back and forth | |
a90968fa | 162 | between different modules. |
07d83abe MV |
163 | |
164 | The function @code{scm_calloc} is similar to @code{scm_malloc}, but | |
165 | initializes the block of memory to zero as well. | |
166 | @end deftypefn | |
167 | ||
168 | @deftypefn {C Function} {void *} scm_realloc (void *@var{mem}, size_t @var{new_size}) | |
169 | Change the size of the memory block at @var{mem} to @var{new_size} and | |
170 | return its new location. When @var{new_size} is 0, this is the same | |
171 | as calling @code{free} on @var{mem} and @code{NULL} is returned. When | |
172 | @var{mem} is @code{NULL}, this function behaves like @code{scm_malloc} | |
173 | and allocates a new block of size @var{new_size}. | |
174 | ||
175 | When not enough memory is available, signal an error. This function | |
176 | runs the GC to free up some memory when it deems it appropriate. | |
177 | @end deftypefn | |
178 | ||
179 | ||
180 | ||
181 | ||
56273dea LC |
182 | @deftypefn {C Function} {void *} scm_gc_malloc (size_t @var{size}, const char *@var{what}) |
183 | @deftypefnx {C Function} {void *} scm_gc_malloc_pointerless (size_t @var{size}, const char *@var{what}) | |
184 | @deftypefnx {C Function} {void *} scm_gc_realloc (void *@var{mem}, size_t @var{old_size}, size_t @var{new_size}, const char *@var{what}); | |
185 | @deftypefnx {C Function} {void *} scm_gc_calloc (size_t @var{size}, const char *@var{what}) | |
186 | Allocate @var{size} bytes of automatically-managed memory. The memory | |
187 | is automatically freed when no longer referenced from any live memory | |
188 | block. | |
189 | ||
190 | Memory allocated with @code{scm_gc_malloc} or @code{scm_gc_calloc} is | |
191 | scanned for pointers. Memory allocated by | |
192 | @code{scm_gc_malloc_pointerless} is not scanned. | |
193 | ||
194 | The @code{scm_gc_realloc} call preserves the ``pointerlessness'' of the | |
195 | memory area pointed to by @var{mem}. Note that you need to pass the old | |
196 | size of a reallocated memory block as well. See below for a motivation. | |
197 | @end deftypefn | |
198 | ||
199 | ||
200 | @deftypefn {C Function} void scm_gc_free (void *@var{mem}, size_t @var{size}, const char *@var{what}) | |
201 | Explicitly free the memory block pointed to by @var{mem}, which was | |
202 | previously allocated by one of the above @code{scm_gc} functions. | |
203 | ||
204 | Note that you need to explicitly pass the @var{size} parameter. This | |
205 | is done since it should normally be easy to provide this parameter | |
206 | (for memory that is associated with GC controlled objects) and help keep | |
207 | the memory management overhead very low. However, in Guile 2.x, | |
208 | @var{size} is always ignored. | |
209 | @end deftypefn | |
210 | ||
211 | ||
07d83abe MV |
212 | @deftypefn {C Function} void scm_gc_register_collectable_memory (void *@var{mem}, size_t @var{size}, const char *@var{what}) |
213 | Informs the GC that the memory at @var{mem} of size @var{size} can | |
214 | potentially be freed during a GC. That is, announce that @var{mem} is | |
215 | part of a GC controlled object and when the GC happens to free that | |
216 | object, @var{size} bytes will be freed along with it. The GC will | |
217 | @strong{not} free the memory itself, it will just know that so-and-so | |
218 | much bytes of memory are associated with GC controlled objects and the | |
219 | memory system figures this into its decisions when to run a GC. | |
220 | ||
07d83abe MV |
221 | The @var{what} argument is used for statistical purposes. It should |
222 | describe the type of object that the memory will be used for so that | |
223 | users can identify just what strange objects are eating up their | |
224 | memory. | |
56273dea LC |
225 | |
226 | In Guile 2.x, this function has no effect. | |
07d83abe MV |
227 | @end deftypefn |
228 | ||
229 | @deftypefn {C Function} void scm_gc_unregister_collectable_memory (void *@var{mem}, size_t @var{size}) | |
230 | Informs the GC that the memory at @var{mem} of size @var{size} is no | |
231 | longer associated with a GC controlled object. You must take care to | |
232 | match up every call to @code{scm_gc_register_collectable_memory} with | |
233 | a call to @code{scm_gc_unregister_collectable_memory}. If you don't do | |
234 | this, the GC might have a wrong impression of what is going on and run | |
235 | much less efficiently than it could. | |
07d83abe | 236 | |
56273dea | 237 | In Guile 2.x, this function has no effect. |
07d83abe MV |
238 | @end deftypefn |
239 | ||
240 | ||
a90968fa MV |
241 | @deftypefn {C Function} void scm_frame_free (void *mem) |
242 | Equivalent to @code{scm_frame_unwind_handler (free, @var{mem}, | |
243 | SCM_F_WIND_EXPLICITLY)}. That is, the memory block at @var{mem} will | |
244 | be freed when the current frame is left. | |
245 | @end deftypefn | |
246 | ||
07d83abe MV |
247 | @deffn {Scheme Procedure} malloc-stats |
248 | Return an alist ((@var{what} . @var{n}) ...) describing number | |
249 | of malloced objects. | |
250 | @var{what} is the second argument to @code{scm_gc_malloc}, | |
251 | @var{n} is the number of objects of that type currently | |
252 | allocated. | |
56273dea LC |
253 | |
254 | This function is only available if the @code{GUILE_DEBUG_MALLOC} | |
255 | preprocessor macro was defined when Guile was compiled. | |
07d83abe MV |
256 | @end deffn |
257 | ||
258 | ||
259 | @subsubsection Upgrading from scm_must_malloc et al. | |
260 | ||
261 | Version 1.6 of Guile and earlier did not have the functions from the | |
262 | previous section. In their place, it had the functions | |
263 | @code{scm_must_malloc}, @code{scm_must_realloc} and | |
264 | @code{scm_must_free}. This section explains why we want you to stop | |
265 | using them, and how to do this. | |
266 | ||
267 | @findex scm_must_malloc | |
268 | @findex scm_must_realloc | |
269 | @findex scm_must_calloc | |
270 | @findex scm_must_free | |
271 | The functions @code{scm_must_malloc} and @code{scm_must_realloc} | |
272 | behaved like @code{scm_gc_malloc} and @code{scm_gc_realloc} do now, | |
273 | respectively. They would inform the GC about the newly allocated | |
274 | memory via the internal equivalent of | |
275 | @code{scm_gc_register_collectable_memory}. However, | |
276 | @code{scm_must_free} did not unregister the memory it was about to | |
277 | free. The usual way to unregister memory was to return its size from | |
278 | a smob free function. | |
279 | ||
280 | This disconnectedness of the actual freeing of memory and reporting | |
281 | this to the GC proved to be bad in practice. It was easy to make | |
282 | mistakes and report the wrong size because allocating and freeing was | |
283 | not done with symmetric code, and because it is cumbersome to compute | |
284 | the total size of nested data structures that were freed with multiple | |
285 | calls to @code{scm_must_free}. Additionally, there was no equivalent | |
286 | to @code{scm_malloc}, and it was tempting to just use | |
287 | @code{scm_must_malloc} and never to tell the GC that the memory has | |
288 | been freed. | |
289 | ||
290 | The effect was that the internal statistics kept by the GC drifted out | |
291 | of sync with reality and could even overflow in long running programs. | |
292 | When this happened, the result was a dramatic increase in (senseless) | |
293 | GC activity which would effectively stop the program dead. | |
294 | ||
295 | @findex scm_done_malloc | |
296 | @findex scm_done_free | |
297 | The functions @code{scm_done_malloc} and @code{scm_done_free} were | |
298 | introduced to help restore balance to the force, but existing bugs did | |
299 | not magically disappear, of course. | |
300 | ||
301 | Therefore we decided to force everybody to review their code by | |
302 | deprecating the existing functions and introducing new ones in their | |
303 | place that are hopefully easier to use correctly. | |
304 | ||
305 | For every use of @code{scm_must_malloc} you need to decide whether to | |
306 | use @code{scm_malloc} or @code{scm_gc_malloc} in its place. When the | |
307 | memory block is not part of a smob or some other Scheme object whose | |
308 | lifetime is ultimately managed by the garbage collector, use | |
309 | @code{scm_malloc} and @code{free}. When it is part of a smob, use | |
310 | @code{scm_gc_malloc} and change the smob free function to use | |
311 | @code{scm_gc_free} instead of @code{scm_must_free} or @code{free} and | |
312 | make it return zero. | |
313 | ||
314 | The important thing is to always pair @code{scm_malloc} with | |
315 | @code{free}; and to always pair @code{scm_gc_malloc} with | |
316 | @code{scm_gc_free}. | |
317 | ||
318 | The same reasoning applies to @code{scm_must_realloc} and | |
319 | @code{scm_realloc} versus @code{scm_gc_realloc}. | |
320 | ||
321 | ||
322 | @node Weak References | |
323 | @subsection Weak References | |
324 | ||
325 | [FIXME: This chapter is based on Mikael Djurfeldt's answer to a | |
326 | question by Michael Livshin. Any mistakes are not theirs, of course. ] | |
327 | ||
328 | Weak references let you attach bookkeeping information to data so that | |
329 | the additional information automatically disappears when the original | |
330 | data is no longer in use and gets garbage collected. In a weak key hash, | |
331 | the hash entry for that key disappears as soon as the key is no longer | |
332 | referenced from anywhere else. For weak value hashes, the same happens | |
333 | as soon as the value is no longer in use. Entries in a doubly weak hash | |
334 | disappear when either the key or the value are not used anywhere else | |
335 | anymore. | |
336 | ||
337 | Object properties offer the same kind of functionality as weak key | |
338 | hashes in many situations. (@pxref{Object Properties}) | |
339 | ||
340 | Here's an example (a little bit strained perhaps, but one of the | |
341 | examples is actually used in Guile): | |
342 | ||
343 | Assume that you're implementing a debugging system where you want to | |
344 | associate information about filename and position of source code | |
345 | expressions with the expressions themselves. | |
346 | ||
347 | Hashtables can be used for that, but if you use ordinary hash tables | |
348 | it will be impossible for the scheme interpreter to "forget" old | |
349 | source when, for example, a file is reloaded. | |
350 | ||
351 | To implement the mapping from source code expressions to positional | |
352 | information it is necessary to use weak-key tables since we don't want | |
353 | the expressions to be remembered just because they are in our table. | |
354 | ||
355 | To implement a mapping from source file line numbers to source code | |
356 | expressions you would use a weak-value table. | |
357 | ||
358 | To implement a mapping from source code expressions to the procedures | |
359 | they constitute a doubly-weak table has to be used. | |
360 | ||
361 | @menu | |
cdf1ad3b | 362 | * Weak hash tables:: |
07d83abe MV |
363 | * Weak vectors:: |
364 | @end menu | |
365 | ||
366 | ||
cdf1ad3b MV |
367 | @node Weak hash tables |
368 | @subsubsection Weak hash tables | |
07d83abe MV |
369 | |
370 | @deffn {Scheme Procedure} make-weak-key-hash-table size | |
371 | @deffnx {Scheme Procedure} make-weak-value-hash-table size | |
372 | @deffnx {Scheme Procedure} make-doubly-weak-hash-table size | |
373 | @deffnx {C Function} scm_make_weak_key_hash_table (size) | |
374 | @deffnx {C Function} scm_make_weak_value_hash_table (size) | |
375 | @deffnx {C Function} scm_make_doubly_weak_hash_table (size) | |
376 | Return a weak hash table with @var{size} buckets. As with any | |
377 | hash table, choosing a good size for the table requires some | |
378 | caution. | |
379 | ||
380 | You can modify weak hash tables in exactly the same way you | |
381 | would modify regular hash tables. (@pxref{Hash Tables}) | |
382 | @end deffn | |
383 | ||
384 | @deffn {Scheme Procedure} weak-key-hash-table? obj | |
385 | @deffnx {Scheme Procedure} weak-value-hash-table? obj | |
386 | @deffnx {Scheme Procedure} doubly-weak-hash-table? obj | |
387 | @deffnx {C Function} scm_weak_key_hash_table_p (obj) | |
388 | @deffnx {C Function} scm_weak_value_hash_table_p (obj) | |
389 | @deffnx {C Function} scm_doubly_weak_hash_table_p (obj) | |
390 | Return @code{#t} if @var{obj} is the specified weak hash | |
391 | table. Note that a doubly weak hash table is neither a weak key | |
392 | nor a weak value hash table. | |
393 | @end deffn | |
394 | ||
07d83abe MV |
395 | @node Weak vectors |
396 | @subsubsection Weak vectors | |
397 | ||
398 | Weak vectors are mainly useful in Guile's implementation of weak hash | |
399 | tables. | |
400 | ||
401 | @deffn {Scheme Procedure} make-weak-vector size [fill] | |
402 | @deffnx {C Function} scm_make_weak_vector (size, fill) | |
403 | Return a weak vector with @var{size} elements. If the optional | |
404 | argument @var{fill} is given, all entries in the vector will be | |
405 | set to @var{fill}. The default value for @var{fill} is the | |
406 | empty list. | |
407 | @end deffn | |
408 | ||
409 | @deffn {Scheme Procedure} weak-vector . l | |
410 | @deffnx {Scheme Procedure} list->weak-vector l | |
411 | @deffnx {C Function} scm_weak_vector (l) | |
412 | Construct a weak vector from a list: @code{weak-vector} uses | |
413 | the list of its arguments while @code{list->weak-vector} uses | |
414 | its only argument @var{l} (a list) to construct a weak vector | |
415 | the same way @code{list->vector} would. | |
416 | @end deffn | |
417 | ||
418 | @deffn {Scheme Procedure} weak-vector? obj | |
419 | @deffnx {C Function} scm_weak_vector_p (obj) | |
420 | Return @code{#t} if @var{obj} is a weak vector. Note that all | |
421 | weak hashes are also weak vectors. | |
422 | @end deffn | |
423 | ||
424 | ||
425 | @node Guardians | |
426 | @subsection Guardians | |
427 | ||
930888e8 MV |
428 | Guardians provide a way to be notified about objects that would |
429 | otherwise be collected as garbage. Guarding them prevents the objects | |
430 | from being collected and cleanup actions can be performed on them, for | |
431 | example. | |
07d83abe | 432 | |
930888e8 MV |
433 | See R. Kent Dybvig, Carl Bruggeman, and David Eby (1993) "Guardians in |
434 | a Generation-Based Garbage Collector". ACM SIGPLAN Conference on | |
435 | Programming Language Design and Implementation, June 1993. | |
07d83abe | 436 | |
930888e8 MV |
437 | @deffn {Scheme Procedure} make-guardian |
438 | @deffnx {C Function} scm_make_guardian () | |
439 | Create a new guardian. A guardian protects a set of objects from | |
440 | garbage collection, allowing a program to apply cleanup or other | |
441 | actions. | |
07d83abe | 442 | |
930888e8 MV |
443 | @code{make-guardian} returns a procedure representing the guardian. |
444 | Calling the guardian procedure with an argument adds the argument to | |
445 | the guardian's set of protected objects. Calling the guardian | |
446 | procedure without an argument returns one of the protected objects | |
447 | which are ready for garbage collection, or @code{#f} if no such object | |
448 | is available. Objects which are returned in this way are removed from | |
449 | the guardian. | |
07d83abe | 450 | |
930888e8 MV |
451 | You can put a single object into a guardian more than once and you can |
452 | put a single object into more than one guardian. The object will then | |
453 | be returned multiple times by the guardian procedures. | |
454 | ||
455 | An object is eligible to be returned from a guardian when it is no | |
456 | longer referenced from outside any guardian. | |
457 | ||
458 | There is no guarantee about the order in which objects are returned | |
459 | from a guardian. If you want to impose an order on finalization | |
460 | actions, for example, you can do that by keeping objects alive in some | |
461 | global data structure until they are no longer needed for finalizing | |
462 | other objects. | |
463 | ||
464 | Being an element in a weak vector, a key in a hash table with weak | |
e5547d5f | 465 | keys, or a value in a hash table with weak values does not prevent an |
930888e8 MV |
466 | object from being returned by a guardian. But as long as an object |
467 | can be returned from a guardian it will not be removed from such a | |
468 | weak vector or hash table. In other words, a weak link does not | |
469 | prevent an object from being considered collectable, but being inside | |
470 | a guardian prevents a weak link from being broken. | |
471 | ||
e5547d5f | 472 | A key in a weak key hash table can be thought of as having a strong |
930888e8 | 473 | reference to its associated value as long as the key is accessible. |
e5547d5f MV |
474 | Consequently, when the key is only accessible from within a guardian, |
475 | the reference from the key to the value is also considered to be | |
476 | coming from within a guardian. Thus, if there is no other reference | |
477 | to the value, it is eligible to be returned from a guardian. | |
07d83abe MV |
478 | @end deffn |
479 | ||
480 | ||
07d83abe MV |
481 | @c Local Variables: |
482 | @c TeX-master: "guile.texi" | |
483 | @c End: |