1 files changed, 525 insertions, 0 deletions

diff --git a/librazor/merger.c b/librazor/merger.c
new file mode 100644
index 0000000..636903a
--- /dev/null
+++ b/librazor/merger.c
@@ -0,0 +1,525 @@
+/*
+ * Copyright (C) 2008  Kristian Høgsberg <krh@redhat.com>
+ * Copyright (C) 2008  Red Hat, Inc
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+ */
+
+#include <string.h>
+#include "razor-internal.h"
+#include "razor.h"
+
+#define UPSTREAM_SOURCE 0x80
+
+struct source {
+	struct razor_set *set;
+	uint32_t *property_map;
+	uint32_t *file_map;
+};
+
+struct razor_merger {
+	struct razor_set *set;
+	struct hashtable table;
+	struct source source1;
+	struct source source2;
+};
+
+struct razor_merger *
+razor_merger_create(struct razor_set *set1, struct razor_set *set2)
+{
+	struct razor_merger *merger;
+	int count;
+	size_t size;
+
+	merger = zalloc(sizeof *merger);
+	merger->set = razor_set_create();
+	hashtable_init(&merger->table, &merger->set->string_pool);
+
+	merger->source1.set = set1;
+	count = set1->properties.size / sizeof (struct razor_property);
+	size = count * sizeof merger->source1.property_map[0];
+	merger->source1.property_map = zalloc(size);
+	count = set1->files.size / sizeof (struct razor_entry);
+	size = count * sizeof merger->source1.file_map[0];
+	merger->source1.file_map = zalloc(size);
+
+	merger->source2.set = set2;
+	count = set2->properties.size / sizeof (struct razor_property);
+	size = count * sizeof merger->source2.property_map[0];
+	merger->source2.property_map = zalloc(size);
+	count = set2->files.size / sizeof (struct razor_entry);
+	size = count * sizeof merger->source2.file_map[0];
+	merger->source2.file_map = zalloc(size);
+
+	return merger;
+}
+
+void
+razor_merger_add_package(struct razor_merger *merger,
+			 struct razor_package *package)
+{
+	char *pool;
+	struct list *r;
+	struct razor_package *p;
+	struct razor_set *set1;
+	struct source *source;
+	uint32_t flags;
+
+	set1 = merger->source1.set;
+	if (set1->packages.data <= (void *) package &&
+	    (void *) package < set1->packages.data + set1->packages.size) {
+		source = &merger->source1;
+		flags = 0;
+	} else {
+		source = &merger->source2;
+		flags = UPSTREAM_SOURCE;
+	}
+
+	pool = source->set->string_pool.data;
+	p = array_add(&merger->set->packages, sizeof *p);
+	p->name = hashtable_tokenize(&merger->table, &pool[package->name]);
+	p->flags = flags;
+	p->version = hashtable_tokenize(&merger->table,
+					&pool[package->version]);
+	p->arch = hashtable_tokenize(&merger->table,
+				     &pool[package->arch]);
+
+	p->properties = package->properties;
+	r = list_first(&package->properties, &source->set->property_pool);
+	while (r) {
+		source->property_map[r->data] = 1;
+		r = list_next(r);
+	}
+
+	p->files = package->files;
+	r = list_first(&package->files, &source->set->file_pool);
+	while (r) {
+		source->file_map[r->data] = 1;
+		r = list_next(r);
+	}
+}
+
+static uint32_t
+add_property(struct razor_merger *merger,
+	     const char *name, uint32_t flags, const char *version)
+{
+	struct razor_property *p;
+
+	p = array_add(&merger->set->properties, sizeof *p);
+	p->name = hashtable_tokenize(&merger->table, name);
+	p->flags = flags;
+	p->version = hashtable_tokenize(&merger->table, version);
+
+	return p - (struct razor_property *) merger->set->properties.data;
+}
+
+static void
+merge_properties(struct razor_merger *merger)
+{
+	struct razor_property *p1, *p2;
+	struct razor_set *set1, *set2;
+	uint32_t *map1, *map2;
+	int i, j, cmp, count1, count2;
+	char *pool1, *pool2;
+
+	set1 = merger->source1.set;
+	set2 = merger->source2.set;
+	map1 = merger->source1.property_map;
+	map2 = merger->source2.property_map;
+
+	i = 0;
+	j = 0;
+	pool1 = set1->string_pool.data;
+	pool2 = set2->string_pool.data;
+
+	count1 = set1->properties.size / sizeof *p1;
+	count2 = set2->properties.size / sizeof *p2;
+	while (i < count1 || j < count2) {
+		if (i < count1 && map1[i] == 0) {
+			i++;
+			continue;
+		}
+		if (j < count2 && map2[j] == 0) {
+			j++;
+			continue;
+		}
+		p1 = (struct razor_property *) set1->properties.data + i;
+		p2 = (struct razor_property *) set2->properties.data + j;
+		if (i < count1 && j < count2)
+			cmp = strcmp(&pool1[p1->name], &pool2[p2->name]);
+		else if (i < count1)
+			cmp = -1;
+		else
+			cmp = 1;
+		if (cmp == 0)
+			cmp = p1->flags - p2->flags;
+		if (cmp == 0)
+			cmp = razor_versioncmp(&pool1[p1->version],
+					       &pool2[p2->version]);
+		if (cmp < 0) {
+			map1[i++] = add_property(merger,
+						 &pool1[p1->name],
+						 p1->flags,
+						 &pool1[p1->version]);
+		} else if (cmp > 0) {
+			map2[j++] = add_property(merger,
+						 &pool2[p2->name],
+						 p2->flags,
+						 &pool2[p2->version]);
+		} else  {
+			map1[i++] = map2[j++] =
+				add_property(merger,
+					     &pool1[p1->name],
+					     p1->flags,
+					     &pool1[p1->version]);
+		}
+	}
+}
+
+static void
+emit_properties(struct list_head *properties, struct array *source_pool,
+		uint32_t *map, struct array *pool)
+{
+	uint32_t r;
+	struct list *p, *q;
+
+	r = pool->size / sizeof *q;
+	p = list_first(properties, source_pool);
+	while (p) {
+		q = array_add(pool, sizeof *q);
+		q->data = map[p->data];
+		q->flags = p->flags;
+		p = list_next(p);
+	}
+
+	list_set_ptr(properties, r);
+}
+
+static uint32_t
+add_file(struct razor_merger *merger, const char *name)
+{
+	struct razor_entry *e;
+
+	e = array_add(&merger->set->files, sizeof *e);
+	e->name = hashtable_tokenize(&merger->table, name);
+	e->flags = 0;
+	e->start = 0;
+
+	return e - (struct razor_entry *)merger->set->files.data;
+}
+
+/* FIXME. Blah */
+static int
+fix_file_map(uint32_t *map,
+	     struct razor_entry *files,
+	     struct razor_entry *top)
+{
+	uint32_t e;
+	int found_file = 0;
+
+	e = top->start;
+	do {
+		if (files[e].start)
+			fix_file_map(map, files, &files[e]);
+		if (map[e])
+			found_file = 1;
+	} while (!(files[e++].flags & RAZOR_ENTRY_LAST));
+
+	if (found_file)
+		map[top - files] = 1;
+	return found_file;
+}
+
+struct merge_directory {
+	uint32_t merged, dir1, dir2;
+};
+
+static void
+merge_one_directory(struct razor_merger *merger, struct merge_directory *md)
+{
+	struct razor_entry *root1, *root2, *mroot, *e1, *e2;
+	struct razor_set *set1, *set2;
+	struct array merge_stack;
+	struct merge_directory *child_md, *end_md;
+	uint32_t *map1, *map2, start, last;
+	int cmp;
+	char *pool1, *pool2;
+
+	set1 = merger->source1.set;
+	set2 = merger->source2.set;
+	map1 = merger->source1.file_map;
+	map2 = merger->source2.file_map;
+	pool1 = set1->string_pool.data;
+	pool2 = set2->string_pool.data;
+	root1 = (struct razor_entry *) set1->files.data;
+	root2 = (struct razor_entry *) set2->files.data;
+
+	array_init(&merge_stack);
+
+	start = merger->set->files.size / sizeof (struct razor_entry);
+	last = 0;
+	e1 = md->dir1 ? root1 + md->dir1 : NULL;
+	e2 = md->dir2 ? root2 + md->dir2 : NULL;
+	while (e1 || e2) {
+		if (!e2 && !map1[e1 - root1]) {
+			if ((e1++)->flags & RAZOR_ENTRY_LAST)
+				e1 = NULL;
+			continue;
+		}
+		if (!e1 && !map2[e2 - root2]) {
+			if ((e2++)->flags & RAZOR_ENTRY_LAST)
+				e2 = NULL;
+			continue;
+		}
+		if (e1 && !map1[e1 - root1] &&
+		    e2 && !map1[e2 - root2]) {
+			if ((e1++)->flags & RAZOR_ENTRY_LAST)
+				e1 = NULL;
+			if ((e2++)->flags & RAZOR_ENTRY_LAST)
+				e2 = NULL;
+			continue;
+		}
+
+		if (!e1)
+			cmp = 1;
+		else if (!e2)
+			cmp = -1;
+		else {
+			cmp = strcmp (&pool1[e1->name],
+				      &pool2[e2->name]);
+		}
+
+		if (cmp < 0) {
+			if (map1[e1 - root1]) {
+				map1[e1 - root1] = last =
+					add_file(merger, &pool1[e1->name]);
+				if (e1->start) {
+					child_md = array_add(&merge_stack, sizeof (struct merge_directory));
+					child_md->merged = last;
+					child_md->dir1 = e1->start;
+					child_md->dir2 = 0;
+				}
+			}
+			if ((e1++)->flags & RAZOR_ENTRY_LAST)
+				e1 = NULL;
+		} else if (cmp > 0) {
+			if (map2[e2 - root2]) {
+				map2[e2 - root2] = last =
+					add_file(merger, &pool2[e2->name]);
+				if (e2->start) {
+					child_md = array_add(&merge_stack, sizeof (struct merge_directory));
+					child_md->merged = last;
+					child_md->dir1 = 0;
+					child_md->dir2 = e2->start;
+				}
+			}
+			if ((e2++)->flags & RAZOR_ENTRY_LAST)
+				e2 = NULL;
+		} else {
+			map1[e1 - root1] = map2[e2- root2] = last =
+				add_file(merger, &pool1[e1->name]);
+			if (e1->start || e2->start) {
+				child_md = array_add(&merge_stack, sizeof (struct merge_directory));
+				child_md->merged = last;
+				child_md->dir1 = e1->start;
+				child_md->dir2 = e2->start;
+			}
+			if ((e1++)->flags & RAZOR_ENTRY_LAST)
+				e1 = NULL;
+			if ((e2++)->flags & RAZOR_ENTRY_LAST)
+				e2 = NULL;
+		}
+	}
+
+	mroot = (struct razor_entry *)merger->set->files.data;
+	if (last) {
+		mroot[last].flags = RAZOR_ENTRY_LAST;
+		mroot[md->merged].start = start;
+	} else
+		mroot[md->merged].start = 0;
+
+	end_md = merge_stack.data + merge_stack.size;
+	for (child_md = merge_stack.data; child_md < end_md; child_md++)
+		merge_one_directory(merger, child_md);
+	array_release(&merge_stack);
+}
+
+static void
+merge_files(struct razor_merger *merger)
+{
+	struct razor_entry *root;
+	struct merge_directory md;
+	uint32_t *map1, *map2;
+
+	map1 = merger->source1.file_map;
+	map2 = merger->source2.file_map;
+
+	md.merged = 0;
+
+	if (merger->source1.set->files.size) {
+		root = (struct razor_entry *) merger->source1.set->files.data;
+		if (root->start)
+			fix_file_map(map1, root, root);
+		md.dir1 = root->start;
+	} else
+		md.dir1 = 0;
+
+	if (merger->source2.set->files.size) {
+		root = (struct razor_entry *) merger->source2.set->files.data;
+		if (root->start)
+			fix_file_map(map2, root, root);
+		md.dir2 = root->start;
+	} else
+		md.dir2 = 0;
+
+	merge_one_directory(merger, &md);
+}
+
+static void
+emit_files(struct list_head *files, struct array *source_pool,
+	   uint32_t *map, struct array *pool)
+{
+	uint32_t r;
+	struct list *p, *q;
+
+	r = pool->size / sizeof *q;
+	p = list_first(files, source_pool);
+	while (p) {
+		q = array_add(pool, sizeof *q);
+		q->data = map[p->data];
+		q->flags = p->flags;
+		p = list_next(p);
+	}
+
+	list_set_ptr(files, r);
+}
+
+/* Rebuild property->packages maps.  We can't just remap these, as a
+ * property may have lost or gained a number of packages.  Allocate an
+ * array per property and loop through the packages and add them to
+ * the arrays for their properties. */
+static void
+rebuild_property_package_lists(struct razor_set *set)
+{
+	struct array *pkgs, *a;
+	struct razor_package *pkg, *pkg_end;
+	struct razor_property *prop, *prop_end;
+	struct list *r;
+	uint32_t *q;
+	int count;
+
+	count = set->properties.size / sizeof (struct razor_property);
+	pkgs = zalloc(count * sizeof *pkgs);
+	pkg_end = set->packages.data + set->packages.size;
+
+	for (pkg = set->packages.data; pkg < pkg_end; pkg++) {
+		r = list_first(&pkg->properties, &set->property_pool);
+		while (r) {
+			q = array_add(&pkgs[r->data], sizeof *q);
+			*q = pkg - (struct razor_package *) set->packages.data;
+			r = list_next(r);
+		}
+	}
+
+	prop_end = set->properties.data + set->properties.size;
+	a = pkgs;
+	for (prop = set->properties.data; prop < prop_end; prop++, a++) {
+		list_set_array(&prop->packages, &set->package_pool, a, 0);
+		array_release(a);
+	}
+	free(pkgs);
+}
+
+static void
+rebuild_file_package_lists(struct razor_set *set)
+{
+	struct array *pkgs, *a;
+	struct razor_package *pkg, *pkg_end;
+	struct razor_entry *entry, *entry_end;
+	struct list *r;
+	uint32_t *q;
+	int count;
+
+	count = set->files.size / sizeof (struct razor_entry);
+	pkgs = zalloc(count * sizeof *pkgs);
+	pkg_end = set->packages.data + set->packages.size;
+
+	for (pkg = set->packages.data; pkg < pkg_end; pkg++) {
+		r = list_first(&pkg->files, &set->file_pool);
+		while (r) {
+			q = array_add(&pkgs[r->data], sizeof *q);
+			*q = pkg - (struct razor_package *) set->packages.data;
+			r = list_next(r);
+		}
+	}
+
+	entry_end = set->files.data + set->files.size;
+	a = pkgs;
+	for (entry = set->files.data; entry < entry_end; entry++, a++) {
+		list_set_array(&entry->packages, &set->package_pool, a, 0);
+		array_release(a);
+	}
+	free(pkgs);
+}
+
+struct razor_set *
+razor_merger_finish(struct razor_merger *merger)
+{
+	struct razor_set *result;
+	struct razor_package *p, *pend;
+
+	/* As we built the package list, we filled out a bitvector of
+	 * the properties that are referenced by the packages in the
+	 * new set.  Now we do a parallel loop through the properties
+	 * and emit those marked in the bit vector to the new set.  In
+	 * the process, we update the bit vector to actually map from
+	 * indices in the old property list to indices in the new
+	 * property list for both sets. */
+
+	merge_properties(merger);
+	merge_files(merger);
+
+	/* Now we loop through the packages again and emit the
+	 * property lists, remapped to point to the new properties. */
+
+	pend = merger->set->packages.data + merger->set->packages.size;
+	for (p = merger->set->packages.data; p < pend; p++) {
+		struct source *src;
+
+		if (p->flags & UPSTREAM_SOURCE)
+			src = &merger->source2;
+		else
+			src = &merger->source1;
+
+		emit_properties(&p->properties,
+				&src->set->property_pool,
+				src->property_map,
+				&merger->set->property_pool);
+		emit_files(&p->files,
+			   &src->set->file_pool,
+			   src->file_map,
+			   &merger->set->file_pool);
+		p->flags &= ~UPSTREAM_SOURCE;
+	}
+
+	rebuild_property_package_lists(merger->set);
+	rebuild_file_package_lists(merger->set);
+
+	result = merger->set;
+	hashtable_release(&merger->table);
+	free(merger);
+
+	return result;
+}