Remove unreferenced resources only from relevant pages

6 files changed, 28 insertions, 15 deletions

diff --git a/ChangeLog b/ChangeLog
index 1e31efb4..40eb1014 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,10 @@
 2021-01-04  Jay Berkenbilt  <ejb@ql.org>
 
+	* When qpdf CLI extracts pages, it now only attempts to remove
+	unreferenced resourecs from the pages that it is keeping. This
+	change dramatically reduces the time it takes to extract a small
+	number of pages from a large, complex file.
+
 	* Move getNext()->write() calls in some pipelines to ensure that
 	state gates properly reset even if the next pipeline's write
 	throws an exception (fuzz issue 28262).
diff --git a/TODO b/TODO
index 7b620a96..9d687d56 100644
--- a/TODO
+++ b/TODO
@@ -1,10 +1,3 @@
-Candidates for upcoming release
-===============================
-
-* Remember to check work `qpdf` project for private issues
-  * file with very slow page extraction
-  * big page even with --remove-unreferenced-resources=yes, even with --empty
-
 Fuzz Errors
 ===========
 
diff --git a/manual/qpdf-manual.xml b/manual/qpdf-manual.xml
index f93f4a17..98fca51e 100644
--- a/manual/qpdf-manual.xml
+++ b/manual/qpdf-manual.xml
@@ -5001,6 +5001,15 @@ print "\n";
           <literal>/DecodeParms</literal>.
          </para>
         </listitem>
+        <listitem>
+         <para>
+          When extracting pages, the <command>qpdf</command> CLI only
+          removes unreferenced resources from the pages that are being
+          kept, resulting in a significant performance improvement
+          when extracting small numbers of pages from large, complex
+          documents.
+         </para>
+        </listitem>
        </itemizedlist>
       </listitem>
       <listitem>
diff --git a/qpdf/qpdf.cc b/qpdf/qpdf.cc
index 0d1ab988..2e35c96f 100644
--- a/qpdf/qpdf.cc
+++ b/qpdf/qpdf.cc
@@ -5120,6 +5120,7 @@ static void handle_page_specs(QPDF& pdf, Options& o)
                          page_spec.range));
     }
 
+    std::map<unsigned long long, bool> remove_unreferenced;
     if (o.remove_unreferenced_page_resources != re_no)
     {
         for (std::map<std::string, QPDF*>::iterator iter =
@@ -5134,10 +5135,11 @@ static void handle_page_specs(QPDF& pdf, Options& o)
                 cis->stayOpen(true);
             }
             QPDF& other(*((*iter).second));
-            if (should_remove_unreferenced_resources(other, o))
+            auto other_uuid = other.getUniqueId();
+            if (remove_unreferenced.count(other_uuid) == 0)
             {
-                QPDFPageDocumentHelper dh(other);
-                dh.removeUnreferencedResources();
+                remove_unreferenced[other_uuid] =
+                    should_remove_unreferenced_resources(other, o);
             }
             if (cis)
             {
@@ -5246,6 +5248,10 @@ static void handle_page_specs(QPDF& pdf, Options& o)
             else
             {
                 copied_pages[from_uuid].insert(to_copy_og);
+                if (remove_unreferenced[from_uuid])
+                {
+                    to_copy.removeUnreferencedResources();
+                }
             }
             dh.addPage(to_copy, false);
             if (page_data.qpdf == &pdf)
diff --git a/qpdf/qtest/qpdf.test b/qpdf/qtest/qpdf.test
index 6919bfcf..83cbacd3 100644
--- a/qpdf/qtest/qpdf.test
+++ b/qpdf/qtest/qpdf.test
@@ -2247,12 +2247,15 @@ $td->runtest("check output",
              {$td->FILE => "a.pdf"},
              {$td->FILE => "shared-images-errors-2-out.pdf"});
 
+# This test used to generate warnings about images on pages we didn't
+# care about, but qpdf was modified not to process those pages, so the
+# "irrelevant" errors went away.
 $td->runtest("shared resources irrelevant errors",
              {$td->COMMAND =>
                   "qpdf --qdf --static-id" .
                   " shared-images-errors.pdf --pages . 1 -- a.pdf"},
-             {$td->FILE => "shared-images-errors-1.out",
-                  $td->EXIT_STATUS => 3},
+             {$td->STRING => "",
+                  $td->EXIT_STATUS => 0},
              $td->NORMALIZE_NEWLINES);
 $td->runtest("check output",
              {$td->FILE => "a.pdf"},
diff --git a/qpdf/qtest/qpdf/shared-images-errors-1.out b/qpdf/qtest/qpdf/shared-images-errors-1.out
deleted file mode 100644
index 5b98f88f..00000000
--- a/qpdf/qtest/qpdf/shared-images-errors-1.out
+++ /dev/null
@@ -1,3 +0,0 @@
-WARNING: shared-images-errors.pdf (offset 4933): error decoding stream data for object 19 0: stream inflate: inflate: data: incorrect header check
-WARNING: shared-images-errors.pdf, object 4 0 at offset 676: Unable to parse content stream: content stream (content stream object 19 0): errors while decoding content stream; not attempting to remove unreferenced objects from this page
-qpdf: operation succeeded with warnings; resulting file may have some problems