Merge git://ozlabs.org/~paulus/gitk
[git.git] / list-objects-filter.c
blobee449de3f77e2b8663d1ae2d43da049f15269249
1 #include "cache.h"
2 #include "dir.h"
3 #include "tag.h"
4 #include "commit.h"
5 #include "tree.h"
6 #include "blob.h"
7 #include "diff.h"
8 #include "tree-walk.h"
9 #include "revision.h"
10 #include "list-objects.h"
11 #include "list-objects-filter.h"
12 #include "list-objects-filter-options.h"
13 #include "oidmap.h"
14 #include "oidset.h"
15 #include "object-store.h"
17 /* Remember to update object flag allocation in object.h */
19 * FILTER_SHOWN_BUT_REVISIT -- we set this bit on tree objects
20 * that have been shown, but should be revisited if they appear
21 * in the traversal (until we mark it SEEN). This is a way to
22 * let us silently de-dup calls to show() in the caller. This
23 * is subtly different from the "revision.h:SHOWN" and the
24 * "sha1-name.c:ONELINE_SEEN" bits. And also different from
25 * the non-de-dup usage in pack-bitmap.c
27 #define FILTER_SHOWN_BUT_REVISIT (1<<21)
30 * A filter for list-objects to omit ALL blobs from the traversal.
31 * And to OPTIONALLY collect a list of the omitted OIDs.
33 struct filter_blobs_none_data {
34 struct oidset *omits;
37 static enum list_objects_filter_result filter_blobs_none(
38 struct repository *r,
39 enum list_objects_filter_situation filter_situation,
40 struct object *obj,
41 const char *pathname,
42 const char *filename,
43 void *filter_data_)
45 struct filter_blobs_none_data *filter_data = filter_data_;
47 switch (filter_situation) {
48 default:
49 BUG("unknown filter_situation: %d", filter_situation);
51 case LOFS_BEGIN_TREE:
52 assert(obj->type == OBJ_TREE);
53 /* always include all tree objects */
54 return LOFR_MARK_SEEN | LOFR_DO_SHOW;
56 case LOFS_END_TREE:
57 assert(obj->type == OBJ_TREE);
58 return LOFR_ZERO;
60 case LOFS_BLOB:
61 assert(obj->type == OBJ_BLOB);
62 assert((obj->flags & SEEN) == 0);
64 if (filter_data->omits)
65 oidset_insert(filter_data->omits, &obj->oid);
66 return LOFR_MARK_SEEN; /* but not LOFR_DO_SHOW (hard omit) */
70 static void *filter_blobs_none__init(
71 struct oidset *omitted,
72 struct list_objects_filter_options *filter_options,
73 filter_object_fn *filter_fn,
74 filter_free_fn *filter_free_fn)
76 struct filter_blobs_none_data *d = xcalloc(1, sizeof(*d));
77 d->omits = omitted;
79 *filter_fn = filter_blobs_none;
80 *filter_free_fn = free;
81 return d;
85 * A filter for list-objects to omit ALL trees and blobs from the traversal.
86 * Can OPTIONALLY collect a list of the omitted OIDs.
88 struct filter_trees_depth_data {
89 struct oidset *omits;
92 * Maps trees to the minimum depth at which they were seen. It is not
93 * necessary to re-traverse a tree at deeper or equal depths than it has
94 * already been traversed.
96 * We can't use LOFR_MARK_SEEN for tree objects since this will prevent
97 * it from being traversed at shallower depths.
99 struct oidmap seen_at_depth;
101 unsigned long exclude_depth;
102 unsigned long current_depth;
105 struct seen_map_entry {
106 struct oidmap_entry base;
107 size_t depth;
110 /* Returns 1 if the oid was in the omits set before it was invoked. */
111 static int filter_trees_update_omits(
112 struct object *obj,
113 struct filter_trees_depth_data *filter_data,
114 int include_it)
116 if (!filter_data->omits)
117 return 0;
119 if (include_it)
120 return oidset_remove(filter_data->omits, &obj->oid);
121 else
122 return oidset_insert(filter_data->omits, &obj->oid);
125 static enum list_objects_filter_result filter_trees_depth(
126 struct repository *r,
127 enum list_objects_filter_situation filter_situation,
128 struct object *obj,
129 const char *pathname,
130 const char *filename,
131 void *filter_data_)
133 struct filter_trees_depth_data *filter_data = filter_data_;
134 struct seen_map_entry *seen_info;
135 int include_it = filter_data->current_depth <
136 filter_data->exclude_depth;
137 int filter_res;
138 int already_seen;
141 * Note that we do not use _MARK_SEEN in order to allow re-traversal in
142 * case we encounter a tree or blob again at a shallower depth.
145 switch (filter_situation) {
146 default:
147 BUG("unknown filter_situation: %d", filter_situation);
149 case LOFS_END_TREE:
150 assert(obj->type == OBJ_TREE);
151 filter_data->current_depth--;
152 return LOFR_ZERO;
154 case LOFS_BLOB:
155 filter_trees_update_omits(obj, filter_data, include_it);
156 return include_it ? LOFR_MARK_SEEN | LOFR_DO_SHOW : LOFR_ZERO;
158 case LOFS_BEGIN_TREE:
159 seen_info = oidmap_get(
160 &filter_data->seen_at_depth, &obj->oid);
161 if (!seen_info) {
162 seen_info = xcalloc(1, sizeof(*seen_info));
163 oidcpy(&seen_info->base.oid, &obj->oid);
164 seen_info->depth = filter_data->current_depth;
165 oidmap_put(&filter_data->seen_at_depth, seen_info);
166 already_seen = 0;
167 } else {
168 already_seen =
169 filter_data->current_depth >= seen_info->depth;
172 if (already_seen) {
173 filter_res = LOFR_SKIP_TREE;
174 } else {
175 int been_omitted = filter_trees_update_omits(
176 obj, filter_data, include_it);
177 seen_info->depth = filter_data->current_depth;
179 if (include_it)
180 filter_res = LOFR_DO_SHOW;
181 else if (filter_data->omits && !been_omitted)
183 * Must update omit information of children
184 * recursively; they have not been omitted yet.
186 filter_res = LOFR_ZERO;
187 else
188 filter_res = LOFR_SKIP_TREE;
191 filter_data->current_depth++;
192 return filter_res;
196 static void filter_trees_free(void *filter_data) {
197 struct filter_trees_depth_data *d = filter_data;
198 if (!d)
199 return;
200 oidmap_free(&d->seen_at_depth, 1);
201 free(d);
204 static void *filter_trees_depth__init(
205 struct oidset *omitted,
206 struct list_objects_filter_options *filter_options,
207 filter_object_fn *filter_fn,
208 filter_free_fn *filter_free_fn)
210 struct filter_trees_depth_data *d = xcalloc(1, sizeof(*d));
211 d->omits = omitted;
212 oidmap_init(&d->seen_at_depth, 0);
213 d->exclude_depth = filter_options->tree_exclude_depth;
214 d->current_depth = 0;
216 *filter_fn = filter_trees_depth;
217 *filter_free_fn = filter_trees_free;
218 return d;
222 * A filter for list-objects to omit large blobs.
223 * And to OPTIONALLY collect a list of the omitted OIDs.
225 struct filter_blobs_limit_data {
226 struct oidset *omits;
227 unsigned long max_bytes;
230 static enum list_objects_filter_result filter_blobs_limit(
231 struct repository *r,
232 enum list_objects_filter_situation filter_situation,
233 struct object *obj,
234 const char *pathname,
235 const char *filename,
236 void *filter_data_)
238 struct filter_blobs_limit_data *filter_data = filter_data_;
239 unsigned long object_length;
240 enum object_type t;
242 switch (filter_situation) {
243 default:
244 BUG("unknown filter_situation: %d", filter_situation);
246 case LOFS_BEGIN_TREE:
247 assert(obj->type == OBJ_TREE);
248 /* always include all tree objects */
249 return LOFR_MARK_SEEN | LOFR_DO_SHOW;
251 case LOFS_END_TREE:
252 assert(obj->type == OBJ_TREE);
253 return LOFR_ZERO;
255 case LOFS_BLOB:
256 assert(obj->type == OBJ_BLOB);
257 assert((obj->flags & SEEN) == 0);
259 t = oid_object_info(r, &obj->oid, &object_length);
260 if (t != OBJ_BLOB) { /* probably OBJ_NONE */
262 * We DO NOT have the blob locally, so we cannot
263 * apply the size filter criteria. Be conservative
264 * and force show it (and let the caller deal with
265 * the ambiguity).
267 goto include_it;
270 if (object_length < filter_data->max_bytes)
271 goto include_it;
273 if (filter_data->omits)
274 oidset_insert(filter_data->omits, &obj->oid);
275 return LOFR_MARK_SEEN; /* but not LOFR_DO_SHOW (hard omit) */
278 include_it:
279 if (filter_data->omits)
280 oidset_remove(filter_data->omits, &obj->oid);
281 return LOFR_MARK_SEEN | LOFR_DO_SHOW;
284 static void *filter_blobs_limit__init(
285 struct oidset *omitted,
286 struct list_objects_filter_options *filter_options,
287 filter_object_fn *filter_fn,
288 filter_free_fn *filter_free_fn)
290 struct filter_blobs_limit_data *d = xcalloc(1, sizeof(*d));
291 d->omits = omitted;
292 d->max_bytes = filter_options->blob_limit_value;
294 *filter_fn = filter_blobs_limit;
295 *filter_free_fn = free;
296 return d;
300 * A filter driven by a sparse-checkout specification to only
301 * include blobs that a sparse checkout would populate.
303 * The sparse-checkout spec can be loaded from a blob with the
304 * given OID or from a local pathname. We allow an OID because
305 * the repo may be bare or we may be doing the filtering on the
306 * server.
308 struct frame {
310 * defval is the usual default include/exclude value that
311 * should be inherited as we recurse into directories based
312 * upon pattern matching of the directory itself or of a
313 * containing directory.
315 int defval;
318 * 1 if the directory (recursively) contains any provisionally
319 * omitted objects.
321 * 0 if everything (recursively) contained in this directory
322 * has been explicitly included (SHOWN) in the result and
323 * the directory may be short-cut later in the traversal.
325 unsigned child_prov_omit : 1;
328 struct filter_sparse_data {
329 struct oidset *omits;
330 struct exclude_list el;
332 size_t nr, alloc;
333 struct frame *array_frame;
336 static enum list_objects_filter_result filter_sparse(
337 struct repository *r,
338 enum list_objects_filter_situation filter_situation,
339 struct object *obj,
340 const char *pathname,
341 const char *filename,
342 void *filter_data_)
344 struct filter_sparse_data *filter_data = filter_data_;
345 int val, dtype;
346 struct frame *frame;
348 switch (filter_situation) {
349 default:
350 BUG("unknown filter_situation: %d", filter_situation);
352 case LOFS_BEGIN_TREE:
353 assert(obj->type == OBJ_TREE);
354 dtype = DT_DIR;
355 val = is_excluded_from_list(pathname, strlen(pathname),
356 filename, &dtype, &filter_data->el,
357 r->index);
358 if (val < 0)
359 val = filter_data->array_frame[filter_data->nr].defval;
361 ALLOC_GROW(filter_data->array_frame, filter_data->nr + 1,
362 filter_data->alloc);
363 filter_data->nr++;
364 filter_data->array_frame[filter_data->nr].defval = val;
365 filter_data->array_frame[filter_data->nr].child_prov_omit = 0;
368 * A directory with this tree OID may appear in multiple
369 * places in the tree. (Think of a directory move or copy,
370 * with no other changes, so the OID is the same, but the
371 * full pathnames of objects within this directory are new
372 * and may match is_excluded() patterns differently.)
373 * So we cannot mark this directory as SEEN (yet), since
374 * that will prevent process_tree() from revisiting this
375 * tree object with other pathname prefixes.
377 * Only _DO_SHOW the tree object the first time we visit
378 * this tree object.
380 * We always show all tree objects. A future optimization
381 * may want to attempt to narrow this.
383 if (obj->flags & FILTER_SHOWN_BUT_REVISIT)
384 return LOFR_ZERO;
385 obj->flags |= FILTER_SHOWN_BUT_REVISIT;
386 return LOFR_DO_SHOW;
388 case LOFS_END_TREE:
389 assert(obj->type == OBJ_TREE);
390 assert(filter_data->nr > 0);
392 frame = &filter_data->array_frame[filter_data->nr];
393 filter_data->nr--;
396 * Tell our parent directory if any of our children were
397 * provisionally omitted.
399 filter_data->array_frame[filter_data->nr].child_prov_omit |=
400 frame->child_prov_omit;
403 * If there are NO provisionally omitted child objects (ALL child
404 * objects in this folder were INCLUDED), then we can mark the
405 * folder as SEEN (so we will not have to revisit it again).
407 if (!frame->child_prov_omit)
408 return LOFR_MARK_SEEN;
409 return LOFR_ZERO;
411 case LOFS_BLOB:
412 assert(obj->type == OBJ_BLOB);
413 assert((obj->flags & SEEN) == 0);
415 frame = &filter_data->array_frame[filter_data->nr];
417 dtype = DT_REG;
418 val = is_excluded_from_list(pathname, strlen(pathname),
419 filename, &dtype, &filter_data->el,
420 r->index);
421 if (val < 0)
422 val = frame->defval;
423 if (val > 0) {
424 if (filter_data->omits)
425 oidset_remove(filter_data->omits, &obj->oid);
426 return LOFR_MARK_SEEN | LOFR_DO_SHOW;
430 * Provisionally omit it. We've already established that
431 * this pathname is not in the sparse-checkout specification
432 * with the CURRENT pathname, so we *WANT* to omit this blob.
434 * However, a pathname elsewhere in the tree may also
435 * reference this same blob, so we cannot reject it yet.
436 * Leave the LOFR_ bits unset so that if the blob appears
437 * again in the traversal, we will be asked again.
439 if (filter_data->omits)
440 oidset_insert(filter_data->omits, &obj->oid);
443 * Remember that at least 1 blob in this tree was
444 * provisionally omitted. This prevents us from short
445 * cutting the tree in future iterations.
447 frame->child_prov_omit = 1;
448 return LOFR_ZERO;
453 static void filter_sparse_free(void *filter_data)
455 struct filter_sparse_data *d = filter_data;
456 /* TODO free contents of 'd' */
457 free(d);
460 static void *filter_sparse_oid__init(
461 struct oidset *omitted,
462 struct list_objects_filter_options *filter_options,
463 filter_object_fn *filter_fn,
464 filter_free_fn *filter_free_fn)
466 struct filter_sparse_data *d = xcalloc(1, sizeof(*d));
467 d->omits = omitted;
468 if (add_excludes_from_blob_to_list(filter_options->sparse_oid_value,
469 NULL, 0, &d->el) < 0)
470 die("could not load filter specification");
472 ALLOC_GROW(d->array_frame, d->nr + 1, d->alloc);
473 d->array_frame[d->nr].defval = 0; /* default to include */
474 d->array_frame[d->nr].child_prov_omit = 0;
476 *filter_fn = filter_sparse;
477 *filter_free_fn = filter_sparse_free;
478 return d;
481 static void *filter_sparse_path__init(
482 struct oidset *omitted,
483 struct list_objects_filter_options *filter_options,
484 filter_object_fn *filter_fn,
485 filter_free_fn *filter_free_fn)
487 struct filter_sparse_data *d = xcalloc(1, sizeof(*d));
488 d->omits = omitted;
489 if (add_excludes_from_file_to_list(filter_options->sparse_path_value,
490 NULL, 0, &d->el, NULL) < 0)
491 die("could not load filter specification");
493 ALLOC_GROW(d->array_frame, d->nr + 1, d->alloc);
494 d->array_frame[d->nr].defval = 0; /* default to include */
495 d->array_frame[d->nr].child_prov_omit = 0;
497 *filter_fn = filter_sparse;
498 *filter_free_fn = filter_sparse_free;
499 return d;
502 typedef void *(*filter_init_fn)(
503 struct oidset *omitted,
504 struct list_objects_filter_options *filter_options,
505 filter_object_fn *filter_fn,
506 filter_free_fn *filter_free_fn);
509 * Must match "enum list_objects_filter_choice".
511 static filter_init_fn s_filters[] = {
512 NULL,
513 filter_blobs_none__init,
514 filter_blobs_limit__init,
515 filter_trees_depth__init,
516 filter_sparse_oid__init,
517 filter_sparse_path__init,
520 void *list_objects_filter__init(
521 struct oidset *omitted,
522 struct list_objects_filter_options *filter_options,
523 filter_object_fn *filter_fn,
524 filter_free_fn *filter_free_fn)
526 filter_init_fn init_fn;
528 assert((sizeof(s_filters) / sizeof(s_filters[0])) == LOFC__COUNT);
530 if (filter_options->choice >= LOFC__COUNT)
531 BUG("invalid list-objects filter choice: %d",
532 filter_options->choice);
534 init_fn = s_filters[filter_options->choice];
535 if (init_fn)
536 return init_fn(omitted, filter_options,
537 filter_fn, filter_free_fn);
538 *filter_fn = NULL;
539 *filter_free_fn = NULL;
540 return NULL;