Remove unused fonts from generated PDF

src/calibre/ebooks/pdf/html_writer.py

@@ -31,7 +31,9 @@ from calibre.ebooks.pdf.render.serialize import PDFStream
 from calibre.gui2 import setup_unix_signals
 from calibre.gui2.webengine import secure_webengine
 from calibre.utils.logging import default_log
-from calibre.utils.podofo import get_podofo, set_metadata_implementation
+from calibre.utils.podofo import (
+    get_podofo, remove_unused_fonts, set_metadata_implementation
+)
 from calibre.utils.short_uuid import uuid4
 from polyglot.builtins import iteritems, map, range, unicode_type
 from polyglot.urllib import urlparse
@@ -563,11 +565,14 @@ def convert(opf_path, opts, metadata=None, output_path=None, log=default_log, co
         add_toc(PDFOutlineRoot(pdf_doc), toc)
     report_progress(0.75, _('Added links to PDF content'))
 
-    # TODO: Remove unused fonts
     # TODO: Remove duplicate fonts
     # TODO: Subset and embed fonts before rendering PDF
     # TODO: Support for mathematics
 
+    num_removed = remove_unused_fonts(pdf_doc)
+    if num_removed:
+        log('Removed', num_removed, 'unused fonts')
+
     if cover_data:
         add_cover(pdf_doc, cover_data, page_layout, opts)
 

src/calibre/utils/podofo/__init__.py

@@ -10,7 +10,7 @@ from calibre.constants import plugins, preferred_encoding
 from calibre.ebooks.metadata import authors_to_string
 from calibre.ptempfile import TemporaryDirectory
 from calibre.utils.ipc.simple_worker import WorkerError, fork_job
-from polyglot.builtins import range, unicode_type
+from polyglot.builtins import range, unicode_type, iteritems
 
 
 def get_podofo():
@@ -154,6 +154,23 @@ def list_fonts(pdf_doc):
     return ref_map
 
 
+def remove_unused_fonts(pdf_doc):
+    font_ref_map = list_fonts(pdf_doc)
+    unused = tuple(ref for ref, font in iteritems(font_ref_map) if not font['used'])
+    pdf_doc.remove_fonts(unused)
+    return len(tuple(f for f in unused if font_ref_map[f]['StreamRef']))
+
+
+def test_remove_unused_fonts(src):
+    podofo = get_podofo()
+    p = podofo.PDFDoc()
+    p.open(src)
+    remove_unused_fonts(p)
+    dest = src.rpartition('.')[0] + '-removed.pdf'
+    p.save(dest)
+    print('Modified pdf saved to:', dest)
+
+
 def test_list_fonts(src):
     podofo = get_podofo()
     p = podofo.PDFDoc()

src/calibre/utils/podofo/doc.cpp

@@ -715,12 +715,15 @@ static PyMethodDef PDFDoc_methods[] = {
     {"alter_links", (PyCFunction)PDFDoc_alter_links, METH_VARARGS,
      "alter_links() -> Change links in the document."
     },
-    {"list_fonts", (PyCFunction)list_fonts, METH_VARARGS,
+    {"list_fonts", (PyCFunction)list_fonts, METH_NOARGS,
      "list_fonts() -> Get list of fonts in document"
     },
     {"used_fonts_in_page_range", (PyCFunction)used_fonts_in_page_range, METH_VARARGS,
      "used_fonts_in_page_range() -> Get list of references to fonts used in the specified pages"
     },
+    {"remove_fonts", (PyCFunction)remove_fonts, METH_VARARGS,
+     "remove_fonts() -> Remove the specified font objects."
+    },
     {"delete_page", (PyCFunction)PDFDoc_delete_page, METH_VARARGS,
      "delete_page(page_num) -> Delete the specified page from the pdf (0 is the first page)."
     },

src/calibre/utils/podofo/fonts.cpp

@@ -17,9 +17,27 @@ ref_as_tuple(const PdfReference &ref) {
     return Py_BuildValue("kk", num, generation);
 }
 
+static inline const PdfObject*
+get_font_file(const PdfObject *descriptor) {
+    PdfObject *ff = descriptor->GetIndirectKey("FontFile");
+    if (!ff) ff = descriptor->GetIndirectKey("FontFile2");
+    if (!ff) ff = descriptor->GetIndirectKey("FontFile3");
+    return ff;
+}
+
+static void
+remove_font(PdfVecObjects &objects, PdfObject *font) {
+    PdfObject *descriptor = font->GetIndirectKey("FontDescriptor");
+    if (descriptor) {
+        const PdfObject *ff = get_font_file(descriptor);
+        if (ff) delete objects.RemoveObject(ff->Reference());
+        delete objects.RemoveObject(descriptor->Reference());
+    }
+    delete objects.RemoveObject(font->Reference());
+}
 
 static bool
-used_fonts_in_page(PdfPage *page, PyObject *ans) {
+used_fonts_in_page(PdfPage *page, int page_num, PyObject *ans) {
     PdfContentsTokenizer tokenizer(page);
     bool in_text_block = false;
     const char* token = NULL;
@@ -73,9 +91,7 @@ list_fonts(PDFDoc *self, PyObject *args) {
                     long long stream_len = 0;
                     pyunique_ptr descendant_font, stream_ref;
                     if (descriptor) {
-                        const PdfObject *ff = descriptor->GetIndirectKey("FontFile");
+                        const PdfObject *ff = get_font_file(descriptor);
-                        if (!ff) ff = descriptor->GetIndirectKey("FontFile2");
-                        if (!ff) ff = descriptor->GetIndirectKey("FontFile3");
                         if (ff) {
                             stream_ref.reset(ref_as_tuple(ff->Reference()));
                             if (!stream_ref) return NULL;
@@ -119,10 +135,27 @@ used_fonts_in_page_range(PDFDoc *self, PyObject *args) {
     for (int i = first - 1; i < last; i++) {
         try {
             PdfPage *page = self->doc->GetPage(i);
-            if (!used_fonts_in_page(page, ans.get())) return NULL;
+            if (!used_fonts_in_page(page, i, ans.get())) return NULL;
         } catch (const PdfError &err) { continue; }
     }
     return ans.release();
 }
 
+PyObject*
+remove_fonts(PDFDoc *self, PyObject *args) {
+    PyObject *fonts;
+    if (!PyArg_ParseTuple(args, "O!", &PyTuple_Type, &fonts)) return NULL;
+    PdfVecObjects &objects = self->doc->GetObjects();
+    for (Py_ssize_t i = 0; i < PyTuple_GET_SIZE(fonts); i++) {
+        unsigned long num, gen;
+        if (!PyArg_ParseTuple(PyTuple_GET_ITEM(fonts, i), "kk", &num, &gen)) return NULL;
+        PdfReference ref(num, gen);
+        PdfObject *font = objects.GetObject(ref);
+        if (font) {
+            remove_font(objects, font);
+        }
+    }
+    Py_RETURN_NONE;
+}
+
 }

src/calibre/utils/podofo/global.h

@@ -61,5 +61,6 @@ dictionary_has_key_name(const PdfDictionary &d, T key, const char *name) {
 extern "C" {
 PyObject* list_fonts(PDFDoc*, PyObject*);
 PyObject* used_fonts_in_page_range(PDFDoc *self, PyObject *args);
+PyObject* remove_fonts(PDFDoc *self, PyObject *args);
 }
 }