Add flags to control zlib compression level (fixes #113)

Jay Berkenbilt
1 parent dac0598b
Showing 9 changed files with 178 additions and 22 deletions
ChangeLog
include/qpdf/QPDFWriter.hh
libqpdf/QPDFWriter.cc
manual/build.mk
manual/qpdf-manual.xml
qpdf/qpdf.cc
qpdf/qtest/qpdf.test
qpdf/qtest/qpdf/minimal-1.pdf
qpdf/qtest/qpdf/minimal-9.pdf
 2019-08-23  Jay Berkenbilt  <ejb@ql.org>
  
+	* Add --recompress-streams option to qpdf and
+	QPDFWriter::setRecompressFlate to cause QPDFWriter to recompress
+	streams that are already compressed with /FlateDecode.
+
 	* Add option Pl_Flate::setCompressionLevel to globally set the
 	zlib compression level used by all Pl_Flate pipelines.
  
+	* Add --compression-level flag to qpdf to set the zlib compression
+	level. When combined with --recompress-flate, this will cause most
+	of qpdf's streams to use the maximum compression level. This
+	results in only a very small amount of savings in size that comes
+	at a fairly significant performance cost, but it could be useful
+	for archival files or other cases where every byte counts and
+	creation time doesn't matter so much. Note that using
+	--object-streams=generate in combination with these options gives
+	you the biggest advantage. Fixes #113.
+
 2019-08-22  Jay Berkenbilt  <ejb@ql.org>
  
 	* In QPDFObjectHandle::ParserCallbacks, in addition to
@@ -189,10 +189,11 @@ class QPDFWriter
     // filters on the input. When combined with
     // setCompressStreams(true), which the default, the effect of this
     // is that streams filtered with these older and less efficient
-    // filters will be recompressed with the Flate filter. As a
-    // special case, if a stream is already compressed with
+    // filters will be recompressed with the Flate filter. By default,
+    // as a special case, if a stream is already compressed with
     // FlateDecode and setCompressStreams is enabled, the original
-    // compressed data will be preserved.
+    // compressed data will be preserved. This behavior can be
+    // overridden by calling setRecompressFlate(true).
     //
     // qpdf_dl_specialized: In addition to uncompressing the
     // generalized compression formats, supported non-lossy
@@ -209,6 +210,15 @@ class QPDFWriter
     QPDF_DLL
     void setDecodeLevel(qpdf_stream_decode_level_e);
  
+    // By default, when both the input and output contents of a stream
+    // are compressed with Flate, qpdf does not uncompress and
+    // recompress the stream. Passing true here causes it to do so.
+    // This can be useful if recompressing all streams with a higher
+    // compression level, which can be set by calling the static
+    // method Pl_Flate::setCompressionLevel.
+    QPDF_DLL
+    void setRecompressFlate(bool);
+
     // Set value of content stream normalization.  The default is
     // "false".  If true, we attempt to normalize newlines inside of
     // content streams.  Some constructs such as inline images may
@@ -597,6 +607,7 @@ class QPDFWriter
         bool compress_streams_set;
         qpdf_stream_decode_level_e stream_decode_level;
         bool stream_decode_level_set;
+        bool recompress_flate;
         bool qdf_mode;
         bool preserve_unreferenced_objects;
         bool newline_before_endstream;
@@ -37,6 +37,7 @@ QPDFWriter::Members::Members(QPDF&amp; pdf) :
     compress_streams_set(false),
     stream_decode_level(qpdf_dl_none),
     stream_decode_level_set(false),
+    recompress_flate(false),
     qdf_mode(false),
     preserve_unreferenced_objects(false),
     newline_before_endstream(false),
@@ -207,6 +208,12 @@ QPDFWriter::setDecodeLevel(qpdf_stream_decode_level_e val)
 }
  
 void
+QPDFWriter::setRecompressFlate(bool val)
+{
+    this->m->recompress_flate = val;
+}
+
+void
 QPDFWriter::setContentNormalization(bool val)
 {
     this->m->normalize_content_set = true;
@@ -1716,13 +1723,14 @@ QPDFWriter::unparseObject(QPDFObjectHandle object, int level,
 	if (this->m->compress_streams)
 	{
 	    // Don't filter if the stream is already compressed with
-	    // FlateDecode.  We don't want to make it worse by getting
-	    // rid of a predictor or otherwise messing with it.  We
-	    // should also avoid messing with anything that's
-	    // compressed with a lossy compression scheme, but we
-	    // don't support any of those right now.
+	    // FlateDecode. This way we don't make it worse if the
+	    // original file used a better Flate algorithm, and we
+	    // don't spend time and CPU cycles uncompressing and
+	    // recompressing stuff. This can be overridden with
+	    // setRecompressFlate(true).
 	    QPDFObjectHandle filter_obj = stream_dict.getKey("/Filter");
-            if ((! object.isDataModified()) &&
+            if ((! this->m->recompress_flate) &&
+                (! object.isDataModified()) &&
                 filter_obj.isName() &&
 		((filter_obj.getName() == "/FlateDecode") ||
 		 (filter_obj.getName() == "/Fl")))
@@ -26,7 +26,8 @@ endif
  
 $(OUTDOC).pdf: $(OUTDOC).fo qpdf/build/qpdf
 	$(FOP) $< -pdf $@.tmp
-	qpdf/build/qpdf --linearize $@.tmp $@
+	qpdf/build/qpdf --linearize --object-streams=generate \
+		--recompress-flate --compression-level=9 $@.tmp $@
  
 $(OUTDOC).html: $(INDOC).xml manual/html.xsl $(VALIDATE)
 	$(XSLTPROC) --output $@ manual/html.xsl $<
@@ -1433,27 +1433,32 @@ outfile.pdf&lt;/option&gt;
          <listitem>
           <para>
            <option>generalized</option>: decode streams filtered with
-           supported generalized filters: <option>/LZWDecode</option>,
-           <option>/FlateDecode</option>,
-           <option>/ASCII85Decode</option>, and
-           <option>/ASCIIHexDecode</option>. We define generalized
+           supported generalized filters:
+           <literal>/LZWDecode</literal>,
+           <literal>/FlateDecode</literal>,
+           <literal>/ASCII85Decode</literal>, and
+           <literal>/ASCIIHexDecode</literal>. We define generalized
            filters as those to be used for general-purpose compression
            or encoding, as opposed to filters specifically designed
-           for image data.
+           for image data. Note that, by default, streams already
+           compressed with <literal>/FlateDecode</literal> are not
+           uncompressed and recompressed unless you also specify
+           <option>--recompress-flate</option>.
           </para>
          </listitem>
          <listitem>
           <para>
            <option>specialized</option>: in addition to generalized,
            decode streams with supported non-lossy specialized
-           filters; currently this is just <option>/RunLengthDecode</option>
+           filters; currently this is just
+           <literal>/RunLengthDecode</literal>
           </para>
          </listitem>
          <listitem>
           <para>
            <option>all</option>: in addition to generalized and
            specialized, decode streams with supported lossy filters;
-           currently this is just <option>/DCTDecode</option> (JPEG)
+           currently this is just <literal>/DCTDecode</literal> (JPEG)
           </para>
          </listitem>
         </itemizedlist>
@@ -1476,7 +1481,10 @@ outfile.pdf&lt;/option&gt;
            <option>compress</option>: recompress stream data when
            possible (default); equivalent to
            <option>--compress-streams=y</option>
-           <option>--decode-level=generalized</option>
+           <option>--decode-level=generalized</option>. Does not
+           recompress streams already compressed with
+           <literal>/FlateDecode</literal> unless
+           <option>--recompress-flate</option> is also specified.
           </para>
          </listitem>
          <listitem>
@@ -1499,6 +1507,37 @@ outfile.pdf&lt;/option&gt;
       </listitem>
      </varlistentry>
      <varlistentry>
+      <term><option>--recompress-flate</option></term>
+      <listitem>
+       <para>
+        By default, streams already compressed with
+        <literal>/FlateDecode</literal> are left alone rather than
+        being uncompressed and recompressed. This option causes qpdf
+        to uncompress and recompress the streams. There is a
+        significant performance cost to using this option, but you
+        probably want to use it if you specify
+        <option>--compression-level</option>.
+       </para>
+      </listitem>
+     </varlistentry>
+     <varlistentry>
+      <term><option>--compression-level=<replaceable>level</replaceable></option></term>
+      <listitem>
+       <para>
+        When writing new streams that are compressed with
+        <literal>/FlateDecode</literal>, use the specified compression
+        level. The value of <option>level</option> should be a number
+        from 1 to 9 and is passed directly to zlib, which implements
+        deflate compression. Note that qpdf doesn't uncompress and
+        recompress streams by default. To have this option apply to
+        already compressed streams, you should also specify
+        <option>--recompress-flate</option>. If your goal is to shrink
+        the size of PDF files, you should also use
+        <option>--object-streams=generate</option>.
+       </para>
+      </listitem>
+     </varlistentry>
+     <varlistentry>
       <term><option>--normalize-content=[yn]</option></term>
       <listitem>
        <para>
@@ -4449,7 +4488,7 @@ print &quot;\n&quot;;
       </listitem>
       <listitem>
        <para>
-        Library Enhancements
+        Library and CLI Enhancements
        </para>
        <itemizedlist>
         <listitem>
@@ -4510,6 +4549,41 @@ print &quot;\n&quot;;
         </listitem>
         <listitem>
          <para>
+          Static method
+          <function>Pl_Flate::setCompressionLevel</function> can be
+          called to set the zlib compression level globally used by
+          all instances of Pl_Flate in deflate mode.
+         </para>
+        </listitem>
+        <listitem>
+         <para>
+          The method
+          <function>QPDFWriter::setRecompressFlate</function> can be
+          called to tell <classname>QPDFWriter</classname> to
+          uncompress and recompress streams already compressed with
+          <literal>/FlateDecode</literal>.
+         </para>
+        </listitem>
+        <listitem>
+         <para>
+          CLI enhancement: the <option>--recompress-flate</option>
+          instructs <command>qpdf</command> to recompress streams that
+          are already compressed with <literal>/FlateDecode</literal>.
+          Useful with <option>--compression-level</option>.
+         </para>
+        </listitem>
+        <listitem>
+         <para>
+          CLI enhancement: the
+          <option>--compression-level=<replaceable>level</replaceable></option>
+          sets the zlib compression level used for any streams
+          compressed by <literal>/FlateDecode</literal>. Most
+          effective when combined with
+          <option>--recompress-flate</option>.
+         </para>
+        </listitem>
+        <listitem>
+         <para>
           The underlying implementation of QPDF arrays has been
           enhanced to be much more memory efficient when dealing with
           arrays with lots of nulls. This enables qpdf to use
@@ -5699,9 +5773,9 @@ print &quot;\n&quot;;
       <listitem>
        <para>
         Disregard data check errors when uncompressing
-        <option>/FlateDecode</option> streams. This is consistent with
-        most other PDF readers and allows qpdf to recover data from
-        another class of malformed PDF files.
+        <literal>/FlateDecode</literal> streams. This is consistent
+        with most other PDF readers and allows qpdf to recover data
+        from another class of malformed PDF files.
        </para>
       </listitem>
       <listitem>
@@ -13,6 +13,7 @@
 #include <qpdf/Pl_Discard.hh>
 #include <qpdf/Pl_DCT.hh>
 #include <qpdf/Pl_Count.hh>
+#include <qpdf/Pl_Flate.hh>
 #include <qpdf/PointerHolder.hh>
  
 #include <qpdf/QPDF.hh>
@@ -124,6 +125,9 @@ struct Options
         stream_data_mode(qpdf_s_compress),
         compress_streams(true),
         compress_streams_set(false),
+        recompress_flate(false),
+        recompress_flate_set(false),
+        compression_level(-1),
         decode_level(qpdf_dl_generalized),
         decode_level_set(false),
         normalize_set(false),
@@ -217,6 +221,9 @@ struct Options
     qpdf_stream_data_e stream_data_mode;
     bool compress_streams;
     bool compress_streams_set;
+    bool recompress_flate;
+    bool recompress_flate_set;
+    int compression_level;
     qpdf_stream_decode_level_e decode_level;
     bool decode_level_set;
     bool normalize_set;
@@ -632,6 +639,8 @@ class ArgParser
     void argCollate();
     void argStreamData(char* parameter);
     void argCompressStreams(char* parameter);
+    void argRecompressFlate();
+    void argCompressionLevel(char* parameter);
     void argDecodeLevel(char* parameter);
     void argNormalizeContent(char* parameter);
     void argSuppressRecovery();
@@ -847,6 +856,9 @@ ArgParser::initOptionTable()
         &ArgParser::argStreamData, stream_data_choices);
     (*t)["compress-streams"] = oe_requiredChoices(
         &ArgParser::argCompressStreams, yn);
+    (*t)["recompress-flate"] = oe_bare(&ArgParser::argRecompressFlate);
+    (*t)["compression-level"] = oe_requiredParameter(
+        &ArgParser::argCompressionLevel, "level");
     char const* decode_level_choices[] =
         {"none", "generalized", "specialized", "all", 0};
     (*t)["decode-level"] = oe_requiredChoices(
@@ -1328,6 +1340,9 @@ ArgParser::argHelp()
         << "--stream-data=option      controls transformation of stream data (below)\n"
         << "--compress-streams=[yn]   controls whether to compress streams on output\n"
         << "--decode-level=option     controls how to filter streams from the input\n"
+        << "--recompress-flate        recompress streams already compressed with Flate\n"
+        << "--compression-level=n     set zlib compression level; most effective with\n"
+        << "                          --recompress-flate --object-streams=generate\n"
         << "--normalize-content=[yn]  enables or disables normalization of content streams\n"
         << "--object-streams=mode     controls handing of object streams\n"
         << "--preserve-unreferenced   preserve unreferenced objects\n"
@@ -1725,6 +1740,19 @@ ArgParser::argCompressStreams(char* parameter)
 }
  
 void
+ArgParser::argRecompressFlate()
+{
+    o.recompress_flate_set = true;
+    o.recompress_flate = true;
+}
+
+void
+ArgParser::argCompressionLevel(char* parameter)
+{
+    o.compression_level = QUtil::string_to_int(parameter);
+}
+
+void
 ArgParser::argDecodeLevel(char* parameter)
 {
     o.decode_level_set = true;
@@ -4889,6 +4917,10 @@ static void set_encryption_options(QPDF&amp; pdf, Options&amp; o, QPDFWriter&amp; w)
  
 static void set_writer_options(QPDF& pdf, Options& o, QPDFWriter& w)
 {
+    if (o.compression_level >= 0)
+    {
+        Pl_Flate::setCompressionLevel(o.compression_level);
+    }
     if (o.qdf_mode)
     {
         w.setQDFMode(true);
@@ -4913,6 +4945,10 @@ static void set_writer_options(QPDF&amp; pdf, Options&amp; o, QPDFWriter&amp; w)
     {
         w.setCompressStreams(o.compress_streams);
     }
+    if (o.recompress_flate_set)
+    {
+        w.setRecompressFlate(o.recompress_flate);
+    }
     if (o.decode_level_set)
     {
         w.setDecodeLevel(o.decode_level);
@@ -3876,8 +3876,20 @@ $td-&gt;runtest(&quot;convert inline-images to qdf&quot;,
 compare_pdfs("inline-images.pdf", "a.pdf");
  
 show_ntests();
+# ----------
+$td->notify("--- Compression Level ---");
+$n_tests += 4;
  
+check_pdf("recompress with level",
+	  "qpdf --static-id --recompress-flate --compression-level=9" .
+          " --object-streams=generate minimal.pdf",
+          "minimal-9.pdf", 0);
+check_pdf("recompress with level",
+	  "qpdf --static-id --recompress-flate --compression-level=1" .
+          " --object-streams=generate minimal.pdf",
+          "minimal-1.pdf", 0);
  
+show_ntests();
 # ----------
 $td->notify("--- Specialized filtering Tests ---");
 $n_tests += 3;