Add --jpeg-quality-level flag (fixes #488)

Thanks to github user @cdosborn for the basic enhancement.

Add --jpeg-quality-level flag (fixes #488)
Thanks to github user @cdosborn for the basic enhancement.
Jay Berkenbilt
1 parent e62f1e4e
Showing 17 changed files with 142 additions and 34 deletions
include/qpdf/Pl_DCT.hh
include/qpdf/QPDFJob.hh
include/qpdf/auto_job_c_main.hh
job.sums
job.yml
libqpdf/Pl_DCT.cc
libqpdf/QPDFJob.cc
libqpdf/QPDFJob_config.cc
libqpdf/qpdf/auto_job_help.hh
libqpdf/qpdf/auto_job_init.hh
libqpdf/qpdf/auto_job_json_init.hh
libqpdf/qpdf/auto_job_schema.hh
libtests/dct_compress.cc
manual/cli.rst
manual/qpdf.1
manual/release-notes.rst
qpdf/qtest/image-optimization.test
@@ -64,7 +64,7 @@ class QPDF_DLL_CLASS Pl_DCT: public Pipeline
     };
  
     QPDF_DLL
-    static std::shared_ptr<CompressConfig>
+    static std::unique_ptr<CompressConfig>
         make_compress_config(std::function<void(jpeg_compress_struct*)>);
  
     // Constructor for compressing image data
@@ -634,6 +634,7 @@ class QPDFJob
         bool recompress_flate{false};
         bool recompress_flate_set{false};
         int compression_level{-1};
+        int jpeg_quality{-1};
         qpdf_stream_decode_level_e decode_level{qpdf_dl_generalized};
         bool decode_level_set{false};
         bool normalize_set{false};
@@ -54,6 +54,7 @@ QPDF_DLL Config* verbose();
 QPDF_DLL Config* warningExit0();
 QPDF_DLL Config* withImages();
 QPDF_DLL Config* compressionLevel(std::string const& parameter);
+QPDF_DLL Config* jpegQuality(std::string const& parameter);
 QPDF_DLL Config* copyEncryption(std::string const& parameter);
 QPDF_DLL Config* encryptionFilePassword(std::string const& parameter);
 QPDF_DLL Config* forceVersion(std::string const& parameter);
@@ -4,17 +4,17 @@ generate_auto_job f64733b79dcee5a0e3e8ccc6976448e8ddf0e8b6529987a66a7d3ab2ebc10a
 include/qpdf/auto_job_c_att.hh 4c2b171ea00531db54720bf49a43f8b34481586ae7fb6cbf225099ee42bc5bb4
 include/qpdf/auto_job_c_copy_att.hh 50609012bff14fd82f0649185940d617d05d530cdc522185c7f3920a561ccb42
 include/qpdf/auto_job_c_enc.hh 28446f3c32153a52afa239ea40503e6cc8ac2c026813526a349e0cd4ae17ddd5
-include/qpdf/auto_job_c_main.hh 48e8ea475e8a8f4c96de86bdad10dff83a263deccc3798c8bed7f5e0e070a037
+include/qpdf/auto_job_c_main.hh b865eb827356554763bb8349eadfcbc5cb260f80e025a5e229467c525007356d
 include/qpdf/auto_job_c_pages.hh 09ca15649cc94fdaf6d9bdae28a20723f2a66616bf15aa86d83df31051d82506
 include/qpdf/auto_job_c_uo.hh 9c2f98a355858dd54d0bba444b73177a59c9e56833e02fa6406f429c07f39e62
-job.yml 9245e70c233dc2067827593403bd9e9feafc5aa0be6b12bb7b99a4b2cab84584
+job.yml e136726a6c7e43736b1f75f4de347fa50baf5f38ed1da58647ce2e980751fb29
 libqpdf/qpdf/auto_job_decl.hh 34ba07d3891c3e5cdd8712f991e508a0652c9db314c5d5bcdf4421b76e6f6e01
-libqpdf/qpdf/auto_job_help.hh 03bdaab05f84b16bfb15ad7993a4655b7dc14af070fa97fe3035943726d4b258
-libqpdf/qpdf/auto_job_init.hh 029d929f930f60b4055796c8c4ce2ed625f861316ac738ab638579eca46b2472
+libqpdf/qpdf/auto_job_help.hh bcfe600cc01447a7fae8661a8d101c7b15ce144bb06ba0beab656f3a655371d1
+libqpdf/qpdf/auto_job_init.hh 02c526c37ad4051cac956ac7c12ae1d020517264f3f3d3beabb066ae2529e4bf
 libqpdf/qpdf/auto_job_json_decl.hh 04965f6321e54b8b3b1dd2ca101d763a22ab44fa81c69e4b6fc0fd6bb7f50f92
-libqpdf/qpdf/auto_job_json_init.hh 42b402305b52fc217453206c0a372303d0b59d4d4227bb564b4fa639257d4411
-libqpdf/qpdf/auto_job_schema.hh 2d3c163c74498b638a13931eed71c2a4dc6b155a9d3e2c1b740070fac4293737
+libqpdf/qpdf/auto_job_json_init.hh b49378f00d521a9f3e0ce9086e30b082bc6ef8e43c845e2a3c99857b72448307
+libqpdf/qpdf/auto_job_schema.hh f6a3e8b663714bba50b594f5e31437bbcb96ca4609d2c150c3bbc172e3b000fa
 manual/_ext/qpdf.py 6add6321666031d55ed4aedf7c00e5662bba856dfcd66ccb526563bffefbb580
-manual/cli.rst 1094662a10db21528fd151739a9779a4504ebac75b483a11a53d42ab0430ee42
-manual/qpdf.1 c7d03b8b544b0c3b2a74149d746596d4564aefff50a53980e435aa5c841f7bed
+manual/cli.rst 6fae28c9589bfde5b55260c95a7c64ad48688875f14f195129606405b32a04c6
+manual/qpdf.1 95feb3b18b5ca3f868628591e2e91c3535125e9eedaedf4e3b160f32f1ff9f8f
 manual/qpdf.1.in 436ecc85d45c4c9e2dbd1725fb7f0177fb627179469f114561adf3cb6cbb677b
@@ -160,6 +160,7 @@ options:
       - split-pages
     required_parameter:
       compression-level: level
+      jpeg-quality: level
       copy-encryption: file
       encryption-file-password: password
       force-version: version
@@ -413,6 +414,7 @@ json:
   suppress-recovery:
   coalesce-contents:
   compression-level:
+  jpeg-quality:
   externalize-inline-images:
   ii-min-bytes:
   remove-unreferenced-resources:
@@ -424,8 +424,8 @@ Pl_DCT::decompress(void* cinfo_p, Buffer* b)
     next()->finish();
 }
  
-std::shared_ptr<Pl_DCT::CompressConfig>
+std::unique_ptr<Pl_DCT::CompressConfig>
 Pl_DCT::make_compress_config(std::function<void(jpeg_compress_struct*)> f)
 {
-    return std::make_shared<FunctionCallbackConfig>(f);
+    return std::make_unique<FunctionCallbackConfig>(f);
 }
@@ -45,6 +45,7 @@ namespace
             size_t oi_min_width,
             size_t oi_min_height,
             size_t oi_min_area,
+            int quality,
             QPDFObjectHandle& image);
         ~ImageOptimizer() override = default;
         void provideStreamData(QPDFObjGen const&, Pipeline* pipeline) override;
@@ -56,6 +57,7 @@ namespace
         size_t oi_min_width;
         size_t oi_min_height;
         size_t oi_min_area;
+        qpdf_stream_decode_level_e decode_level{qpdf_dl_specialized};
         QPDFObjectHandle image;
         std::shared_ptr<Pl_DCT::CompressConfig> config;
     };
@@ -109,6 +111,7 @@ ImageOptimizer::ImageOptimizer(
     size_t oi_min_width,
     size_t oi_min_height,
     size_t oi_min_area,
+    int quality,
     QPDFObjectHandle& image) :
     o(o),
     oi_min_width(oi_min_width),
@@ -116,7 +119,12 @@ ImageOptimizer::ImageOptimizer(
     oi_min_area(oi_min_area),
     image(image)
 {
-    config = Pl_DCT::make_compress_config([](jpeg_compress_struct* config) {});
+    if (quality >= 0) {
+        // Recompress existing jpeg.
+        decode_level = qpdf_dl_all;
+        config = Pl_DCT::make_compress_config(
+            [quality](jpeg_compress_struct* cinfo) { jpeg_set_quality(cinfo, quality, false); });
+    }
 }
  
 std::shared_ptr<Pipeline>
@@ -204,7 +212,8 @@ ImageOptimizer::makePipeline(std::string const&amp; description, Pipeline* next)
 bool
 ImageOptimizer::evaluate(std::string const& description)
 {
-    if (!image.pipeStreamData(nullptr, 0, qpdf_dl_specialized, true)) {
+    // Note: passing nullptr as pipeline (first argument) just tests whether we can filter.
+    if (!image.pipeStreamData(nullptr, 0, decode_level, true)) {
         QTC::TC("qpdf", "QPDFJob image optimize no pipeline");
         o.doIfVerbose([&](Pipeline& v, std::string const& prefix) {
             v << prefix << ": " << description
@@ -219,7 +228,7 @@ ImageOptimizer::evaluate(std::string const&amp; description)
         // message issued by makePipeline
         return false;
     }
-    if (!image.pipeStreamData(p.get(), 0, qpdf_dl_specialized)) {
+    if (!image.pipeStreamData(p.get(), 0, decode_level)) {
         return false;
     }
     long long orig_length = image.getDict().getKey("/Length").getIntValue();
@@ -249,7 +258,7 @@ ImageOptimizer::provideStreamData(QPDFObjGen const&amp;, Pipeline* pipeline)
         pipeline->finish();
         return;
     }
-    image.pipeStreamData(p.get(), 0, qpdf_dl_specialized, false, false);
+    image.pipeStreamData(p.get(), 0, decode_level, false, false);
 }
  
 QPDFJob::PageSpec::PageSpec(
@@ -2196,7 +2205,12 @@ QPDFJob::handleTransformations(QPDF&amp; pdf)
                 [this, pageno, &pdf](
                     QPDFObjectHandle& obj, QPDFObjectHandle& xobj_dict, std::string const& key) {
                     auto io = std::make_unique<ImageOptimizer>(
-                        *this, m->oi_min_width, m->oi_min_height, m->oi_min_area, obj);
+                        *this,
+                        m->oi_min_width,
+                        m->oi_min_height,
+                        m->oi_min_area,
+                        m->jpeg_quality,
+                        obj);
                     if (io->evaluate("image " + key + " on page " + std::to_string(pageno))) {
                         QPDFObjectHandle new_image = pdf.newStream();
                         new_image.replaceDict(obj.getDict().shallowCopy());
@@ -140,6 +140,13 @@ QPDFJob::Config::compressionLevel(std::string const&amp; parameter)
 }
  
 QPDFJob::Config*
+QPDFJob::Config::jpegQuality(std::string const& parameter)
+{
+    o.m->jpeg_quality = QUtil::string_to_int(parameter.c_str());
+    return this;
+}
+
+QPDFJob::Config*
 QPDFJob::Config::copyEncryption(std::string const& parameter)
 {
     o.m->encryption_file = parameter;
@@ -239,6 +239,14 @@ gzip), which is the default compression for most PDF files.
 You need --recompress-flate with this option if you want to
 change already compressed streams.
 )");
+ap.addOptionHelp("--jpeg-quality", "transformation", "set jpeg quality level for jpeg", R"(--jpeg-quality=level
+
+When rewriting images with --optimize-images, set a quality
+level from 0 (lowest) to 100 (highest) for writing new images.
+Higher quality results in larger images, and lower quality
+results in smaller images. This option is only effective when
+combined with --optimize-images.
+)");
 ap.addOptionHelp("--normalize-content", "transformation", "fix newlines in content streams", R"(--normalize-content=[y|n]
  
 Normalize newlines to UNIX-style newlines in PDF content
@@ -284,15 +292,15 @@ version. The version number format is
 to "major.minor" and the extension level, if specified, to
 "extension-level".
 )");
+}
+static void add_help_4(QPDFArgParser& ap)
+{
 ap.addOptionHelp("--force-version", "transformation", "set output PDF version", R"(--force-version=version
  
 Force the output PDF file's PDF version header to be the specified
 value, even if the file uses features that may not be available
 in that version.
 )");
-}
-static void add_help_4(QPDFArgParser& ap)
-{
 ap.addHelpTopic("page-ranges", "page range syntax", R"(A full description of the page range syntax, with examples, can be
 found in the manual. In summary, a range is a comma-separated list
 of groups. A group is a number or a range of numbers separated by a
@@ -419,11 +427,11 @@ Don&#39;t optimize images whose area in pixels is below the specified value.
 )");
 ap.addOptionHelp("--keep-inline-images", "modification", "exclude inline images from optimization", R"(Prevent inline images from being considered by --optimize-images.
 )");
-ap.addOptionHelp("--remove-info", "modification", "remove file information", R"(Exclude file information (except modification date) from the output file.
-)");
 }
 static void add_help_5(QPDFArgParser& ap)
 {
+ap.addOptionHelp("--remove-info", "modification", "remove file information", R"(Exclude file information (except modification date) from the output file.
+)");
 ap.addOptionHelp("--remove-metadata", "modification", "remove metadata", R"(Exclude metadata from the output file.
 )");
 ap.addOptionHelp("--remove-page-labels", "modification", "remove explicit page numbers", R"(Exclude page labels (explicit page numbers) from the output file.
@@ -642,12 +650,12 @@ version to be at least 1.6. This option is only available with
 128-bit encryption. The default is "n" for compatibility
 reasons. Use 256-bit encryption instead.
 )");
-ap.addOptionHelp("--allow-insecure", "encryption", "allow empty owner passwords", R"(Allow creation of PDF files with empty owner passwords and
-non-empty user passwords when using 256-bit encryption.
-)");
 }
 static void add_help_6(QPDFArgParser& ap)
 {
+ap.addOptionHelp("--allow-insecure", "encryption", "allow empty owner passwords", R"(Allow creation of PDF files with empty owner passwords and
+non-empty user passwords when using 256-bit encryption.
+)");
 ap.addOptionHelp("--force-V4", "encryption", "force V=4 in encryption dictionary", R"(This option is for testing and is never needed in practice since
 qpdf does this automatically when needed.
 )");
@@ -825,14 +833,14 @@ ap.addOptionHelp(&quot;--mimetype&quot;, &quot;add-attachment&quot;, &quot;attachment mime type, e.g. app
 Specify the mime type for the attachment, such as text/plain,
 application/pdf, image/png, etc.
 )");
+}
+static void add_help_7(QPDFArgParser& ap)
+{
 ap.addOptionHelp("--description", "add-attachment", "set attachment's description", R"(--description="text"
  
 Supply descriptive text for the attachment, displayed by some
 PDF viewers.
 )");
-}
-static void add_help_7(QPDFArgParser& ap)
-{
 ap.addOptionHelp("--replace", "add-attachment", "replace attachment with same key", R"(Indicate that any existing attachment with the same key should
 be replaced by the new attachment. Otherwise, qpdf gives an
 error if an attachment with that key is already present.
@@ -919,12 +927,12 @@ object and for each content stream associated with the page.
 ap.addOptionHelp("--with-images", "inspection", "include image details with --show-pages", R"(When used with --show-pages, also shows the object and
 generation numbers for the image objects on each page.
 )");
-ap.addOptionHelp("--list-attachments", "inspection", "list embedded files", R"(Show the key and stream number for each embedded file. Combine
-with --verbose for more detailed information.
-)");
 }
 static void add_help_8(QPDFArgParser& ap)
 {
+ap.addOptionHelp("--list-attachments", "inspection", "list embedded files", R"(Show the key and stream number for each embedded file. Combine
+with --verbose for more detailed information.
+)");
 ap.addOptionHelp("--show-attachment", "inspection", "export an embedded file", R"(--show-attachment=key
  
 Write the contents of the specified attachment to standard
@@ -94,6 +94,7 @@ this-&gt;ap.addBare(&quot;verbose&quot;, [this](){c_main-&gt;verbose();});
 this->ap.addBare("warning-exit-0", [this](){c_main->warningExit0();});
 this->ap.addBare("with-images", [this](){c_main->withImages();});
 this->ap.addRequiredParameter("compression-level", [this](std::string const& x){c_main->compressionLevel(x);}, "level");
+this->ap.addRequiredParameter("jpeg-quality", [this](std::string const& x){c_main->jpegQuality(x);}, "level");
 this->ap.addRequiredParameter("copy-encryption", [this](std::string const& x){c_main->copyEncryption(x);}, "file");
 this->ap.addRequiredParameter("encryption-file-password", [this](std::string const& x){c_main->encryptionFilePassword(x);}, "password");
 this->ap.addRequiredParameter("force-version", [this](std::string const& x){c_main->forceVersion(x);}, "version");
@@ -314,6 +314,9 @@ popHandler(); // key: coalesceContents
 pushKey("compressionLevel");
 addParameter([this](std::string const& p) { c_main->compressionLevel(p); });
 popHandler(); // key: compressionLevel
+pushKey("jpegQuality");
+addParameter([this](std::string const& p) { c_main->jpegQuality(p); });
+popHandler(); // key: jpegQuality
 pushKey("externalizeInlineImages");
 addBare([this]() { c_main->externalizeInlineImages(); });
 popHandler(); // key: externalizeInlineImages
@@ -104,6 +104,7 @@ static constexpr char const* JOB_SCHEMA_DATA = R&quot;({
   "suppressRecovery": "suppress error recovery",
   "coalesceContents": "combine content streams",
   "compressionLevel": "set compression level for flate",
+  "jpegQuality": "set jpeg quality level for jpeg",
   "externalizeInlineImages": "convert inline to regular images",
   "iiMinBytes": "set minimum size for externalizeInlineImages",
   "removeUnreferencedResources": "remove unreferenced page resources",
+#include <qpdf/assert_test.h>
+
 #include <qpdf/Pl_DCT.hh>
 #include <qpdf/Pl_StdioFile.hh>
 #include <qpdf/QUtil.hh>
-#include <qpdf/assert_test.h>
  
 #include <cstdio>
 #include <cstdlib>
@@ -1028,6 +1028,26 @@ Related Options
    defers to the compression library's default behavior. See also
    :ref:`small-files`.
  
+.. qpdf:option:: --jpeg-quality=level
+
+   .. help: set jpeg quality level for jpeg
+
+      When rewriting images with --optimize-images, set a quality
+      level from 0 (lowest) to 100 (highest) for writing new images.
+      Higher quality results in larger images, and lower quality
+      results in smaller images. This option is only effective when
+      combined with --optimize-images.
+
+   When rewriting images with :qpdf:ref:`--optimize-images`, set a
+   quality level from 0 (lowest) to 100 (highest) for writing new
+   images. Higher quality results in larger images, and lower quality
+   results in smaller images. Be sure to check your output to see if
+   the quality is acceptable. This option is only effective when
+   combined with :qpdf:ref:`--optimize-images`. This option also
+   causes files that are already compressed with JPEG compression to
+   be uncompressed and recompressed, potentially introducing
+   additional loss of image quality. See also :ref:`small-files`.
+
 .. qpdf:option:: --normalize-content=[y|n]
  
    .. help: fix newlines in content streams
@@ -4005,6 +4025,13 @@ generate:
   lossy, so images may have artifacts. These are not usually
   noticeable to the casual observer.
  
+- ``--jpeg-quality=n``: set the JPEG quality used by :qpdf:ref:`--optimize-images`
+  when writing JPEG files. Use a lower number for ``n`` for smaller,
+  lower-quality images. The default of most JPEG libraries is 75.
+  Smaller numbers result in lower-quality but smaller images. The
+  :qpdf:ref:`--jpeg-quality` option was added in qpdf 12.1. This only
+  works in combination with :qpdf:ref:`--optimize-images`.
+
 - ``--object-streams=generate``: generate object streams, which means
   that more of the PDF file's structural content will be compressed
   (see :qpdf:ref:`--object-streams`)
@@ -317,6 +317,15 @@ gzip), which is the default compression for most PDF files.
 You need --recompress-flate with this option if you want to
 change already compressed streams.
 .TP
+.B --jpeg-quality \-\- set jpeg quality level for jpeg
+--jpeg-quality=level
+
+When rewriting images with --optimize-images, set a quality
+level from 0 (lowest) to 100 (highest) for writing new images.
+Higher quality results in larger images, and lower quality
+results in smaller images. This option is only effective when
+combined with --optimize-images.
+.TP
 .B --normalize-content \-\- fix newlines in content streams
 --normalize-content=[y|n]
  
@@ -41,14 +41,18 @@ more detail.
   - Library Enhancements
  
     - Add function ``Pl_DCT::make_compress_config`` to return a
-      ``Pl_DCT::CompressConfig`` shared pointer from a
-      ``std::function`` for a more modern configuration option.
+      ``Pl_DCT::CompressConfig`` unique pointer to a
+      ``CompressConfig`` from a ``std::function`` for a more modern
+      configuration option.
  
   - CLI Enhancements
  
     - New :qpdf:ref:`--remove-structure` option to exclude the document
       structure tree from the output PDF.
  
+    - New :qpdf:ref:`--jpeg-quality` option to set jpeg quality used
+      with :qpdf:ref:`--optimize-images`.
+
   - Other enhancements
  
     - There have been further enhancements to how files with damaged xref
@@ -37,7 +37,7 @@ my @image_opt = (
      '--oi-min-width=0 --oi-min-height=0 --oi-min-area=0']
     );
  
-my $n_tests = 2 * scalar(@image_opt);
+my $n_tests = 2 * scalar(@image_opt) + 5;
  
 foreach my $d (@image_opt)
 {
@@ -58,5 +58,34 @@ foreach my $d (@image_opt)
                  $td->NORMALIZE_NEWLINES);
 }
  
+$td->runtest("quality = 100",
+             {$td->COMMAND =>
+                  "qpdf --static-id --optimize-images --jpeg-quality=100" .
+                  " large-inline-image.pdf a.pdf"},
+             {$td->STRING => "", $td->EXIT_STATUS => 0});
+$td->runtest("quality = 50",
+             {$td->COMMAND =>
+                  "qpdf --static-id --optimize-images --jpeg-quality=50" .
+                  " large-inline-image.pdf b.pdf"},
+             {$td->STRING => "", $td->EXIT_STATUS => 0});
+$td->runtest("quality = 50 from DCT",
+             {$td->COMMAND =>
+                  "qpdf --static-id --optimize-images --jpeg-quality=50" .
+                  " a.pdf c.pdf"},
+             {$td->STRING => "", $td->EXIT_STATUS => 0});
+my $size100 = (stat("a.pdf"))[7];
+my $size50 = (stat("b.pdf"))[7];
+my $size50b = (stat("c.pdf"))[7];
+my $result = $size50 < $size100 ? "ok\n" : "failed\n";
+$td->runtest("quality 50 < quality 100",
+             {$td->STRING => $result},
+             {$td->STRING => "ok\n"},
+             $td->NORMALIZE_NEWLINES);
+$result = $size50b < $size100 ? "ok\n" : "failed\n";
+$td->runtest("quality 50 from DCT < quality 100",
+             {$td->STRING => $result},
+             {$td->STRING => "ok\n"},
+             $td->NORMALIZE_NEWLINES);
+
 cleanup();
 $td->report($n_tests);