Prioritise incomplete object during deduplication.

Signed-off-by: yzrh <yzrh@noema.org>

3 files changed, 45 insertions, 13 deletions

diff --git a/CHANGE.md b/CHANGE.md
index 063d93a..3e05e10 100644
--- a/CHANGE.md
+++ b/CHANGE.md
@@ -2,6 +2,8 @@
 ==================
 
 * Support HN text overlay.
+* Handle invalid PDF object token in CAJ and KDH.
+* Handle inaccuracy page count in CAJ and KDH.
 
 0.2.5 (2023-01-XX)
 ==================
@@ -9,6 +11,7 @@
 * Improve PDF parser.
 * Handle duplicated object in CAJ.
 * Handle duplicated image in HN.
+* Handle incomplete PDF object in CAJ and KDH.
 * Fix JBIG decoder.
 
 0.2.4 (2022-12-31)
diff --git a/src/cnki_pdf.c b/src/cnki_pdf.c
index af38aa6..76931ea 100644
--- a/src/cnki_pdf.c
+++ b/src/cnki_pdf.c
@@ -163,10 +163,16 @@ _pdf_obj_dedup(cnki_t **param, pdf_object_t **pdf)
 		printf("Deleting duplicated object\n");
 
 	ptr = *pdf;
-	while (ptr != NULL && ptr->next != NULL) {
-		if (ptr->id == ptr->next->id) {
-			pdf_get_obj(&ptr, ptr->id, &tmp);
-			pdf_obj_del(&ptr, ptr->id);
+	while (ptr->next != NULL && ptr->next->next != NULL) {
+		if (ptr->next->id == ptr->next->next->id) {
+			/* Keep the bigger one, the smaller one is usually incomplete */
+			if (ptr->next->size < ptr->next->next->size) {
+				pdf_get_obj(&ptr, ptr->next->id, &tmp);
+				pdf_obj_del(&ptr, ptr->next->id);
+			} else {
+				pdf_get_obj(&ptr->next, ptr->next->id, &tmp);
+				pdf_obj_del(&ptr->next, ptr->next->id);
+			}
 
 			tmp->next = NULL;
 			pdf_obj_destroy(&tmp);
@@ -174,7 +180,9 @@ _pdf_obj_dedup(cnki_t **param, pdf_object_t **pdf)
 			ret++;
 
 			if ((*param)->stat > 1)
-				printf("Deleted duplicated object %d.\n", ptr->id);
+				printf("Deleted duplicated object %d.\n", ptr->next->id);
+
+			continue;
 		}
 
 		ptr = ptr->next;
@@ -236,6 +244,10 @@ cnki_pdf(cnki_t **param)
 		printf("Loaded %d object(s)\n",
 			pdf_get_count(&pdf));
 
+	pdf_obj_sort(&pdf);
+
+	_pdf_obj_dedup(param, &pdf);
+
 	int dictionary_size;
 	char *dictionary;
 
@@ -301,7 +313,7 @@ cnki_pdf(cnki_t **param)
 
 			snprintf(buf, 64,
 				"]\n/Count %d\n>>",
-				pdf_get_kid_count(&pdf, parent[i]));
+				pdf_get_kid_count(&pdf, parent[i]) > 0 ? pdf_get_kid_count(&pdf, parent[i]) : kid[0]);
 			strcat(dictionary, buf);
 
 			pdf_obj_prepend(&pdf, parent[i], NULL, dictionary, NULL, 0);
@@ -354,7 +366,7 @@ cnki_pdf(cnki_t **param)
 		} else {
 			for (int i = 0; i < parent[0]; i++)
 				if (parent_missing[i] == 1)
-					root = i;
+					root = parent[i + 1];
 		}
 
 		if (root == 0)
@@ -471,8 +483,6 @@ cnki_pdf(cnki_t **param)
 
 	_pdf_obj_sort(param, &pdf);
 
-	_pdf_obj_dedup(param, &pdf);
-
 	_pdf_dump(param, &pdf);
 
 	pdf_obj_destroy(&pdf);
@@ -510,7 +520,7 @@ cnki_pdf_hn(cnki_t **param)
 	if (root_kid == NULL)
 		return 1;
 
-	memset(root_kid, 0, (*param)->file_stat->page);
+	memset(root_kid, 0, (*param)->file_stat->page * sizeof(int));
 
 	object_hn_t *ptr = (*param)->object_hn;
 	while (ptr != NULL) {
diff --git a/src/pdf_parser.c b/src/pdf_parser.c
index 781bafa..ed7bfba 100644
--- a/src/pdf_parser.c
+++ b/src/pdf_parser.c
@@ -183,7 +183,7 @@ pdf_load(pdf_object_t **pdf, FILE **fp, int size_buf)
 				ptr->size - (tail - buf) - 3,
 				">>", 2)) != NULL &&
 				memmem(tail + 3,
-				ptr->size - (tail - buf) - 3,
+				(tmp - tail) - 3,
 				"stream\r\n", 8) == NULL)
 				tail = tmp;
 
@@ -226,8 +226,27 @@ pdf_load(pdf_object_t **pdf, FILE **fp, int size_buf)
 
 			free(buf);
 		} else {
-			ptr->object_size = ptr->size;
-			ptr->object = buf;
+			/* Handle incomplete object */
+			head = buf;
+			while ((tmp = _memmem_whitespace(head,
+				ptr->size - (head - buf),
+				" 0 obj", 6)) != NULL)
+				head = tmp + 7;
+
+			if (head - buf > 0) {
+				ptr->object_size = ptr->size - (head - buf);
+				ptr->object = malloc(ptr->object_size);
+
+				if (ptr->object == NULL)
+					return 1;
+
+				memcpy(ptr->object, head, ptr->object_size);
+
+				free(buf);
+			} else {
+				ptr->object_size = ptr->size;
+				ptr->object = buf;
+			}
 		}
 
 		ptr = ptr->next;