Add 'clean' option to pdfclean to clean (but not sanitize) content streams.

This goes well with the 'mutool clean -d' decompression option to debug
content streams, without doing the sanitize optimization pass.

3 files changed, 53 insertions, 31 deletions

diff --git a/source/pdf/pdf-clean.c b/source/pdf/pdf-clean.c
index 7fe0d6fe..0c9cb974 100644
--- a/source/pdf/pdf-clean.c
+++ b/source/pdf/pdf-clean.c
@@ -2,7 +2,9 @@
 #include "mupdf/pdf.h"
 
 static void
-pdf_clean_stream_object(fz_context *ctx, pdf_document *doc, pdf_obj *obj, pdf_obj *orig_res, fz_cookie *cookie, int own_res, int ascii, pdf_text_filter_fn *text_filter, pdf_after_text_object_fn *after_text, void *arg)
+pdf_clean_stream_object(fz_context *ctx, pdf_document *doc, pdf_obj *obj, pdf_obj *orig_res, fz_cookie *cookie, int own_res,
+		pdf_text_filter_fn *text_filter, pdf_after_text_object_fn *after_text, void *arg,
+		int sanitize, int ascii)
 {
 	pdf_processor *proc_buffer = NULL;
 	pdf_processor *proc_filter = NULL;
@@ -59,7 +61,7 @@ pdf_clean_stream_object(fz_context *ctx, pdf_document *doc, pdf_obj *obj, pdf_ob
 }
 
 static void
-pdf_clean_type3(fz_context *ctx, pdf_document *doc, pdf_obj *obj, pdf_obj *orig_res, fz_cookie *cookie, int ascii)
+pdf_clean_type3(fz_context *ctx, pdf_document *doc, pdf_obj *obj, pdf_obj *orig_res, fz_cookie *cookie, int sanitize, int ascii)
 {
 	pdf_processor *proc_buffer = NULL;
 	pdf_processor *proc_filter = NULL;
@@ -91,10 +93,16 @@ pdf_clean_type3(fz_context *ctx, pdf_document *doc, pdf_obj *obj, pdf_obj *orig_
 			fz_try(ctx)
 			{
 				proc_buffer = pdf_new_buffer_processor(ctx, buffer, ascii);
-				proc_filter = pdf_new_filter_processor(ctx, doc, proc_buffer, orig_res, res);
-
-				pdf_process_contents(ctx, proc_filter, doc, orig_res, val, cookie);
-				pdf_close_processor(ctx, proc_filter);
+				if (sanitize)
+				{
+					proc_filter = pdf_new_filter_processor(ctx, doc, proc_buffer, orig_res, res);
+					pdf_process_contents(ctx, proc_filter, doc, orig_res, val, cookie);
+					pdf_close_processor(ctx, proc_filter);
+				}
+				else
+				{
+					pdf_process_contents(ctx, proc_filter, doc, orig_res, val, cookie);
+				}
 				pdf_close_processor(ctx, proc_buffer);
 
 				pdf_update_stream(ctx, doc, val, buffer, 0);
@@ -127,12 +135,14 @@ pdf_clean_type3(fz_context *ctx, pdf_document *doc, pdf_obj *obj, pdf_obj *orig_
 	}
 }
 
-void pdf_clean_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page, fz_cookie *cookie, pdf_page_contents_process_fn *proc_fn, void *arg, int ascii)
+void pdf_clean_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page, fz_cookie *cookie, pdf_page_contents_process_fn *proc_fn, void *arg, int sanitize, int ascii)
 {
-	pdf_filter_page_contents(ctx, doc, page, cookie, proc_fn, NULL, NULL, arg, ascii);
+	pdf_filter_page_contents(ctx, doc, page, cookie, proc_fn, NULL, NULL, arg, sanitize, ascii);
 }
 
-void pdf_filter_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page, fz_cookie *cookie, pdf_page_contents_process_fn *proc_fn, pdf_text_filter_fn *text_filter, pdf_after_text_object_fn *after_text, void *proc_arg, int ascii)
+void pdf_filter_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page, fz_cookie *cookie,
+		pdf_page_contents_process_fn *proc_fn, pdf_text_filter_fn *text_filter, pdf_after_text_object_fn *after_text, void *proc_arg,
+		int sanitize, int ascii)
 {
 	pdf_processor *proc_buffer = NULL;
 	pdf_processor *proc_filter = NULL;
@@ -162,10 +172,17 @@ void pdf_filter_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page
 		resources = pdf_page_resources(ctx, page);
 
 		proc_buffer = pdf_new_buffer_processor(ctx, buffer, ascii);
-		proc_filter = pdf_new_filter_processor_with_text_filter(ctx, doc, proc_buffer, resources, res, text_filter, after_text, proc_arg);
+		if (sanitize)
+		{
+			proc_filter = pdf_new_filter_processor_with_text_filter(ctx, doc, proc_buffer, resources, res, text_filter, after_text, proc_arg);
 
-		pdf_process_contents(ctx, proc_filter, doc, resources, contents, cookie);
-		pdf_close_processor(ctx, proc_filter);
+			pdf_process_contents(ctx, proc_filter, doc, resources, contents, cookie);
+			pdf_close_processor(ctx, proc_filter);
+		}
+		else
+		{
+			pdf_process_contents(ctx, proc_buffer, doc, resources, contents, cookie);
+		}
 		pdf_close_processor(ctx, proc_buffer);
 
 		/* Deal with page content stream. */
@@ -207,7 +224,7 @@ void pdf_filter_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page
 				if (!o)
 					continue;
 				/* Transparency group XObject */
-				pdf_clean_stream_object(ctx, doc, o, resources, cookie, 1, ascii, text_filter, after_text, proc_arg);
+				pdf_clean_stream_object(ctx, doc, o, resources, cookie, 1, text_filter, after_text, proc_arg, sanitize, ascii);
 			}
 		}
 
@@ -223,7 +240,7 @@ void pdf_filter_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page
 				if (!pat)
 					continue;
 				if (pdf_to_int(ctx, pdf_dict_get(ctx, pat, PDF_NAME_PatternType)) == 1)
-					pdf_clean_stream_object(ctx, doc, pat, resources, cookie, 0, ascii, text_filter, after_text, proc_arg);
+					pdf_clean_stream_object(ctx, doc, pat, resources, cookie, 0, text_filter, after_text, proc_arg, sanitize, ascii);
 			}
 		}
 
@@ -239,7 +256,7 @@ void pdf_filter_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page
 				if (!xobj)
 					continue;
 				if (pdf_name_eq(ctx, PDF_NAME_Form, pdf_dict_get(ctx, xobj, PDF_NAME_Subtype)))
-					pdf_clean_stream_object(ctx, doc, xobj, resources, cookie, 1, ascii, text_filter, after_text, proc_arg);
+					pdf_clean_stream_object(ctx, doc, xobj, resources, cookie, 1, text_filter, after_text, proc_arg, sanitize, ascii);
 			}
 		}
 
@@ -255,7 +272,7 @@ void pdf_filter_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page
 				if (!o)
 					continue;
 				if (pdf_name_eq(ctx, PDF_NAME_Type3, pdf_dict_get(ctx, o, PDF_NAME_Subtype)))
-					pdf_clean_type3(ctx, doc, o, resources, cookie, ascii);
+					pdf_clean_type3(ctx, doc, o, resources, cookie, sanitize, ascii);
 			}
 		}
 
@@ -290,13 +307,13 @@ void pdf_filter_page_contents(fz_context *ctx, pdf_document *doc, pdf_page *page
 	}
 }
 
-void pdf_clean_annot_contents(fz_context *ctx, pdf_document *doc, pdf_annot *annot, fz_cookie *cookie, pdf_page_contents_process_fn *proc_fn, void *proc_arg, int ascii)
+void pdf_clean_annot_contents(fz_context *ctx, pdf_document *doc, pdf_annot *annot, fz_cookie *cookie, pdf_page_contents_process_fn *proc_fn, void *proc_arg, int sanitize, int ascii)
 {
-	pdf_filter_annot_contents(ctx, doc, annot, cookie, proc_fn, NULL, NULL, proc_arg, ascii);
+	pdf_filter_annot_contents(ctx, doc, annot, cookie, proc_fn, NULL, NULL, proc_arg, sanitize, ascii);
 }
 
 void pdf_filter_annot_contents(fz_context *ctx, pdf_document *doc, pdf_annot *annot, fz_cookie *cookie,
-	pdf_page_contents_process_fn *proc, pdf_text_filter_fn *text_filter, pdf_after_text_object_fn *after_text, void *arg, int ascii)
+	pdf_page_contents_process_fn *proc, pdf_text_filter_fn *text_filter, pdf_after_text_object_fn *after_text, void *arg, int sanitize, int ascii)
 {
 	pdf_obj *ap;
 	int i, n;
@@ -313,6 +330,6 @@ void pdf_filter_annot_contents(fz_context *ctx, pdf_document *doc, pdf_annot *an
 		if (v == NULL)
 			continue;
 
-		pdf_clean_stream_object(ctx, doc, v, NULL, cookie, 1, 1, text_filter, after_text, arg);
+		pdf_clean_stream_object(ctx, doc, v, NULL, cookie, 1, text_filter, after_text, arg, sanitize, ascii);
 	}
 }
diff --git a/source/pdf/pdf-write.c b/source/pdf/pdf-write.c
index 154655e0..9fcdbf0a 100644
--- a/source/pdf/pdf-write.c
+++ b/source/pdf/pdf-write.c
@@ -2703,7 +2703,7 @@ static void complete_signatures(fz_context *ctx, pdf_document *doc, pdf_write_st
 	}
 }
 
-static void sanitize(fz_context *ctx, pdf_document *doc, int ascii)
+static void clean_content_streams(fz_context *ctx, pdf_document *doc, int sanitize, int ascii)
 {
 	int n = pdf_count_pages(ctx, doc);
 	int i;
@@ -2712,11 +2712,11 @@ static void sanitize(fz_context *ctx, pdf_document *doc, int ascii)
 	{
 		pdf_annot *annot;
 		pdf_page *page = pdf_load_page(ctx, doc, i);
-		pdf_clean_page_contents(ctx, doc, page, NULL, NULL, NULL, ascii);
+		pdf_clean_page_contents(ctx, doc, page, NULL, NULL, NULL, sanitize, ascii);
 
 		for (annot = pdf_first_annot(ctx, page); annot != NULL; annot = pdf_next_annot(ctx, annot))
 		{
-			pdf_clean_annot_contents(ctx, doc, annot, NULL, NULL, NULL, ascii);
+			pdf_clean_annot_contents(ctx, doc, annot, NULL, NULL, NULL, sanitize, ascii);
 		}
 
 		fz_drop_page(ctx, &page->super);
@@ -2787,7 +2787,8 @@ const char *fz_pdf_write_options_usage =
 	"\tascii: ASCII hex encode binary streams\n"
 	"\tpretty: pretty-print objects with indentation\n"
 	"\tlinearize: optimize for web browsers\n"
-	"\tsanitize: clean up graphics commands in content streams\n"
+	"\tclean: pretty-print graphics commands in content streams\n"
+	"\tsanitize: sanitize graphics commands in content streams\n"
 	"\tgarbage: garbage collect unused objects\n"
 	"\tincremental: write changes as incremental update\n"
 	"\tcontinue-on-error: continue saving the document even if there is an error\n"
@@ -2816,8 +2817,10 @@ pdf_parse_write_options(fz_context *ctx, pdf_write_options *opts, const char *ar
 		opts->do_pretty = fz_option_eq(val, "yes");
 	if (fz_has_option(ctx, args, "linearize", &val))
 		opts->do_linear = fz_option_eq(val, "yes");
-	if (fz_has_option(ctx, args, "sanitize", &val))
+	if (fz_has_option(ctx, args, "clean", &val))
 		opts->do_clean = fz_option_eq(val, "yes");
+	if (fz_has_option(ctx, args, "sanitize", &val))
+		opts->do_sanitize = fz_option_eq(val, "yes");
 	if (fz_has_option(ctx, args, "incremental", &val))
 		opts->do_incremental = fz_option_eq(val, "yes");
 	if (fz_has_option(ctx, args, "continue-on-error", &val))
@@ -2851,9 +2854,9 @@ prepare_for_save(fz_context *ctx, pdf_document *doc, pdf_write_options *in_opts)
 {
 	doc->freeze_updates = 1;
 
-	/* Sanitize the operator streams */
-	if (in_opts->do_clean)
-		sanitize(ctx, doc, in_opts->do_ascii);
+	/* Rewrite (and possibly sanitize) the operator streams */
+	if (in_opts->do_clean || in_opts->do_sanitize)
+		clean_content_streams(ctx, doc, in_opts->do_sanitize, in_opts->do_ascii);
 
 	pdf_finish_edit(ctx, doc);
 	presize_unsaved_signature_byteranges(ctx, doc);
diff --git a/source/tools/pdfclean.c b/source/tools/pdfclean.c
index ff60bbb5..6d5a69be 100644
--- a/source/tools/pdfclean.c
+++ b/source/tools/pdfclean.c
@@ -31,7 +31,8 @@ static void usage(void)
 		"\t-z\tdeflate uncompressed streams\n"
 		"\t-f\tcompress font streams\n"
 		"\t-i\tcompress image streams\n"
-		"\t-s\tclean content streams\n"
+		"\t-c\tclean content streams\n"
+		"\t-s\tsanitize content streams\n"
 		"\tpages\tcomma separated list of page numbers and ranges\n"
 		);
 	exit(1);
@@ -50,7 +51,7 @@ int pdfclean_main(int argc, char **argv)
 	opts.continue_on_error = 1;
 	opts.errors = &errors;
 
-	while ((c = fz_getopt(argc, argv, "adfgilp:sz")) != -1)
+	while ((c = fz_getopt(argc, argv, "adfgilp:scz")) != -1)
 	{
 		switch (c)
 		{
@@ -63,7 +64,8 @@ int pdfclean_main(int argc, char **argv)
 		case 'a': opts.do_ascii += 1; break;
 		case 'g': opts.do_garbage += 1; break;
 		case 'l': opts.do_linear += 1; break;
-		case 's': opts.do_clean += 1; break;
+		case 'c': opts.do_clean += 1; break;
+		case 's': opts.do_sanitize += 1; break;
 		default: usage(); break;
 		}
 	}