Fix HN text parsing.

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

CHANGE.md

@@ -2,13 +2,17 @@
 ==================
 
 * Support HN text overlay.
+* Support HN page with text.
+* Handle inaccurate page count in CAJ and KDH.
 
-0.2.5 (2023-01-XX)
+0.2.5 (2023-01-05)
 ==================
 
 * Improve PDF parser.
 * Handle duplicated object in CAJ.
 * Handle duplicated image in HN.
+* Handle incomplete PDF object in CAJ and KDH.
+* Handle invalid PDF object token in CAJ and KDH.
 * Fix JBIG decoder.
 
 0.2.4 (2022-12-31)

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;
@@ -250,6 +258,10 @@ cnki_pdf(cnki_t **param)
 	if ((*param)->stat > 0)
 		printf("Discovered %d parent object(s)\n", parent[0]);
 
+	pdf_obj_sort(&pdf);
+
+	_pdf_obj_dedup(param, &pdf);
+
 	int8_t *parent_missing;
 	int *kid;
 
@@ -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)
@@ -363,9 +375,11 @@ cnki_pdf(cnki_t **param)
 			printf("Root object is %d.\n", root);
 	}
 
+	int root_gen;
+
 	pdf_object_t *tmp;
 
-	if (pdf_get_obj(&pdf, root, &tmp) != 0) {
+	if ((root_gen = pdf_get_obj(&pdf, root, &tmp)) != 0) {
 		if ((*param)->stat > 0)
 			printf("Root object is missing\n");
 
@@ -407,21 +421,6 @@ cnki_pdf(cnki_t **param)
 
 	int outline = _pdf_cnki_outline(param, &pdf);
 
-	if ((*param)->stat > 1)
-		printf("Searching for catalog object\n");
-
-	int catalog = pdf_get_catalog_id(&pdf);
-
-	if (catalog != 0) {
-		if ((*param)->stat > 0)
-			printf("Catalog object is %d.\n", catalog);
-	} else {
-		if ((*param)->stat > 0)
-			printf("Catalog object is missing\n");
-
-		if ((*param)->stat > 1)
-			printf("Generating catalog object\n");
-
 	snprintf(buf, 64,
 		"<<\n/Type /Catalog\n/Pages %d 0 R\n",
 		root);
@@ -436,6 +435,31 @@ cnki_pdf(cnki_t **param)
 
 	strcat(dictionary, ">>");
 
+	if ((*param)->stat > 1)
+		printf("Searching for catalog object\n");
+
+	int catalog = pdf_get_catalog_id(&pdf);
+
+	if (catalog != 0) {
+		if ((*param)->stat > 0)
+			printf("Catalog object is %d.\n", catalog);
+
+		if (root_gen != 0) {
+			if ((*param)->stat > 1)
+				printf("Replacing catalog object\n");
+
+			pdf_obj_replace(&pdf, catalog, NULL, dictionary, NULL, 0);
+
+			if ((*param)->stat > 0)
+				printf("Replaced catalog object\n");
+		}
+	} else {
+		if ((*param)->stat > 0)
+			printf("Catalog object is missing\n");
+
+		if ((*param)->stat > 1)
+			printf("Generating catalog object\n");
+
 		pdf_obj_append(&pdf, 0, NULL, dictionary, NULL, 0);
 
 		if ((*param)->stat > 0)
@@ -471,8 +495,6 @@ 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 +532,7 @@ 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) {
@@ -828,45 +850,7 @@ cnki_pdf_hn(cnki_t **param)
 			for (int i = 0, j = 0; i < ptr->text_size - 1;) {
 				switch (((unsigned char) ptr->text[i + 1] << 8) + (unsigned char) ptr->text[i]) {
 					case 0x8001:
-						if (ptr->address_next > ptr->address)
-							strcat(dictionary, "T*\n");
-					case 0x8070:
-						if (ptr->address_next > ptr->address) {
-							i += 4;
-
-							for (;;) {
-								if (i + 3 >= ptr->text_size ||
-									(unsigned char) ptr->text[i + 1] == 0x80)
-									break;
-
-								conv_src[0] = ptr->text[i + 3];
-								conv_src[1] = ptr->text[i + 2];
-
-								//snprintf(buf, 64, "%f %f Td\n");
-								//strcat(dictionary, buf);
-
-								conv_size = 6;
-
-								if (strconv(&conv_dst, "UTF-16BE",
-									conv_src, "GB18030", &conv_size) == 0) {
-									if (conv_size - 2 > 0) {
-										strcat(dictionary, "<feff");
-										for (int k = 0; k < conv_size - 2; k++) {
-											snprintf(conv_hex, 3,
-												"%02x", (unsigned char) conv_dst[k]);
-											strcat(dictionary, conv_hex);
-										}
-										strcat(dictionary, "> Tj\n");
-									}
-									free(conv_dst);
-								}
-
-								i += 4;
-							}
-
-							break;
-						}
-
+						if (ptr->address_next <= ptr->address) {
 							if (i + 7 >= ptr->text_size) {
 								i += 2;
 								break;
@@ -875,7 +859,7 @@ cnki_pdf_hn(cnki_t **param)
 							conv_src[0] = ptr->text[i + 7];
 							conv_src[1] = ptr->text[i + 6];
 
-						//snprintf(buf, 64, "%f %f Td\n");
+							//snprintf(buf, 64, "1 0 0 1 %d %d Tm\n")
 							//strcat(dictionary, buf);
 
 							conv_size = 6;
@@ -883,7 +867,7 @@ cnki_pdf_hn(cnki_t **param)
 							if (strconv(&conv_dst, "UTF-16BE",
 								conv_src, "GB18030", &conv_size) == 0) {
 								if (conv_size - 2 > 0) {
-								strcat(dictionary, "<feff");
+									strcat(dictionary, "<");
 									for (int k = 0; k < conv_size - 2; k++) {
 										snprintf(conv_hex, 3,
 											"%02x", (unsigned char) conv_dst[k]);
@@ -895,6 +879,46 @@ cnki_pdf_hn(cnki_t **param)
 							}
 
 							i += 8;
+							break;
+						}
+
+						strcat(dictionary, "T*\n");
+					case 0x8070:
+						i += 4;
+
+						if (ptr->address_next <= ptr->address)
+							break;
+
+						for (;;) {
+							if (i + 3 >= ptr->text_size ||
+								(unsigned char) ptr->text[i + 1] == 0x80)
+								break;
+
+							conv_src[0] = ptr->text[i + 3];
+							conv_src[1] = ptr->text[i + 2];
+
+							//snprintf(buf, 64, "1 0 0 1 %d %d Tm\n")
+							//strcat(dictionary, buf);
+
+							conv_size = 6;
+
+							if (strconv(&conv_dst, "UTF-16BE",
+								conv_src, "GB18030", &conv_size) == 0) {
+								if (conv_size - 2 > 0) {
+									strcat(dictionary, "<");
+									for (int k = 0; k < conv_size - 2; k++) {
+										snprintf(conv_hex, 3,
+											"%02x", (unsigned char) conv_dst[k]);
+										strcat(dictionary, conv_hex);
+									}
+									strcat(dictionary, "> Tj\n");
+								}
+								free(conv_dst);
+							}
+
+							i += 4;
+						}
+
 						break;
 					case 0x800a:
 						if (i + 27 >= ptr->text_size || j >= ptr->image_length) {

src/jbig2.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2022, yzrh <yzrh@noema.org>
+ * Copyright (c) 2022-2023, yzrh <yzrh@noema.org>
  *
  * SPDX-License-Identifier: Apache-2.0
  */
@@ -31,5 +31,6 @@ strdec_jbig2(char **bitmap,
 	}
 
 	jbig2_release_page(ctx, image);
+	jbig2_ctx_free(ctx);
 	return 0;
 }

src/pdf_parser.c

@@ -79,8 +79,25 @@ _locate(pdf_object_t **pdf, FILE **fp, int size_buf)
 			memset(buf + end - cur, 0, size_buf - end + cur);
 		}
 
-		if (head == 0 && (pos = _memmem_whitespace(buf, size_buf, " 0 obj", 6)) != NULL)
+		if (head == 0) {
+			/* Hack needed for invalid object */
+			pos = _memmem_whitespace(buf, size_buf, " 0 obj", 6);
+			tmp = memmem(buf, size_buf, " 0 obj", 6);
+
+			while (tmp != NULL && tmp[6] != 0x3c && tmp[6] != 0x5b)
+				tmp = memmem(tmp + 6, size_buf - (tmp - buf) - 6, " 0 obj", 6);
+
+			if (pos != NULL && tmp != NULL) {
+				if (pos - buf < tmp - buf)
 					head = cur + (pos - buf) + 7;
+				else
+					head = cur + (tmp - buf) + 6;
+			} else if (pos != NULL) {
+				head = cur + (pos - buf) + 7;
+			} else if (tmp != NULL) {
+				head = cur + (tmp - buf) + 6;
+			}
+		}
 
 		if (tail == 0 && (pos = _memmem_whitespace(buf, size_buf, "endobj", 6)) != NULL) {
 			/* We need to check if it is the object stored in stream */
@@ -156,8 +173,45 @@ pdf_load(pdf_object_t **pdf, FILE **fp, int size_buf)
 		if (buf == NULL)
 			return 1;
 
+		fseek(*fp, ptr->address, SEEK_SET);
+		fread(buf, ptr->size, 1, *fp);
+
+		/* Handle incomplete object */
+		head = buf;
+		while ((tmp = _memmem_whitespace(head,
+			ptr->size - (head - buf),
+			" 0 obj", 6)) != NULL)
+			head = tmp + 7;
+
+		/* Hack needed for invalid object */
+		while ((tmp = memmem(head,
+			ptr->size - (head - buf),
+			" 0 obj", 6)) != NULL)
+			head = tmp + 6;
+
+		if (head - buf > 0) {
+			ptr->address += head - buf;
+			ptr->size -= head - buf;
+
+			tmp = realloc(buf, ptr->size);
+
+			if (tmp == NULL)
+				return 1;
+
+			buf = tmp;
+
+			fseek(*fp, ptr->address, SEEK_SET);
+			fread(buf, ptr->size, 1, *fp);
+		}
+
+		/* Hack needed for invalid object */
+		fseek(*fp, ptr->address - 14, SEEK_SET);
+		fread(str, 8, 1, *fp);
+
+		if (str[7] < '0' || str[7] > '9') {
 			fseek(*fp, ptr->address - 15, SEEK_SET);
 			fread(str, 8, 1, *fp);
+		}
 
 		for (int i = 7; i >= 0; i--) {
 			if (str[i] < '0' || str[i] > '9') {
@@ -170,22 +224,35 @@ pdf_load(pdf_object_t **pdf, FILE **fp, int size_buf)
 			}
 		}
 
-		fseek(*fp, ptr->address, SEEK_SET);
-		fread(buf, ptr->size, 1, *fp);
-
 		if ((head = memmem(buf, ptr->size, "<<", 2)) != NULL &&
-			(tail = _memmem_whitespace(buf, ptr->size, ">>", 2)) != NULL) {
+			((tail = _memmem_whitespace(buf, ptr->size, ">>", 2)) != NULL ||
+			/* Hack needed for invalid object */
+			(tail = memmem(buf, ptr->size, ">>", 2)) != NULL)) {
+			if (memmem(buf, tail - buf, "stream\r\n", 8) != NULL) {
+				tail = memmem(buf, ptr->size, ">>", 2);
+
+				while (ptr->size - (tail - buf) > 2 &&
+					(tmp = memmem(tail + 2,
+					ptr->size - (tail - buf) - 2,
+					">>", 2)) != NULL &&
+					memmem(tail + 2,
+					(tmp - tail) - 2,
+					"stream\r\n", 8) == NULL)
+					tail = tmp;
+			} else {
 				/*
 				 * A dictionary object may have nested dictionary,
 				 * but it should not be in a stream
 				 */
-			while ((tmp = _memmem_whitespace(tail + 3,
+				while (ptr->size - (tail - buf) > 3 &&
+					(tmp = _memmem_whitespace(tail + 3,
 					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;
+			}
 
 			ptr->dictionary_size = tail - head + 2;
 			ptr->dictionary = malloc(ptr->dictionary_size + 1);

src/pdf_writer.c

@@ -1,19 +1,39 @@
 /*
- * Copyright (c) 2020-2022, yzrh <yzrh@noema.org>
+ * Copyright (c) 2020-2023, yzrh <yzrh@noema.org>
  *
  * SPDX-License-Identifier: Apache-2.0
  */
 
 #include <stdlib.h>
+#include <string.h>
 #include <time.h>
 
+#include "version.h"
 #include "md5.h"
 #include "pdf.h"
 
+static int
+_info_obj(pdf_object_t **pdf)
+{
+	char dictionary[128] = "<<\n"
+		"/Producer (Melon " VERSION "." RELEASE "." PATCH EXTRA ")\n"
+		"/CreationDate (D:";
+
+	char buf[64];
+
+	time_t timestamp = time(NULL);
+
+	strftime(buf, 64, "%Y%m%d%H%M%S", gmtime(&timestamp));
+	strcat(dictionary, buf);
+	strcat(dictionary, "+00'00')\n>>");
+
+	return pdf_obj_append(pdf, 0, NULL, dictionary, NULL, 0);
+}
+
 int
 pdf_dump_obj(pdf_object_t **pdf, FILE **fp)
 {
-	if (*pdf == NULL || *fp == NULL)
+	if (*pdf == NULL || *fp == NULL || _info_obj(pdf) != 0)
 		return 1;
 
 	long cur;
@@ -152,18 +172,11 @@ pdf_dump_trailer(pdf_object_t **pdf, FILE **fp, int xref)
 	while (ptr->next != NULL)
 		ptr = ptr->next;
 
-	/*
-	 * TODO: Document information dictionary
-	 * `"/Producer (Melon)"'
-	 * `"/CreationDate (D:YYYYMMDDHHmmSS+00'00')"'
-	 *
-	 * Trailer dictionary
-	 * `"/Info %d 0 R"'
-	 */
 	fprintf(*fp,
-		"/Size %d\n/Root %d 0 R\n",
+		"/Size %d\n/Root %d 0 R\n/Info %d 0 R\n",
 		ptr->id + 1,
-		pdf_get_catalog_id(pdf));
+		pdf_get_catalog_id(pdf),
+		ptr->id);
 
 	fputs("/ID [", *fp);
 

src/version.h

@@ -5,6 +5,6 @@
  */
 
 #define VERSION	"0"
-#define RELEASE	"2"
-#define PATCH	"5"
+#define RELEASE	"3"
+#define PATCH	"0"
 #define EXTRA	""