Speedup serialization of numbers to PDF by a factor of 10

src/calibre/ebooks/conversion/plugins/pdf_output.py

@@ -232,7 +232,15 @@ class PDFOutput(OutputFormatPlugin):
         out_stream.seek(0)
         out_stream.truncate()
         self.log.debug('Rendering pages to PDF...')
+        import time
+        st = time.time()
+        if False:
+            import cProfile
+            cProfile.runctx('writer.dump(items, out_stream, PDFMetadata(self.metadata))',
+                        globals(), locals(), '/tmp/profile')
+        else:
             writer.dump(items, out_stream, PDFMetadata(self.metadata))
+        self.log('Rendered PDF in %g seconds:'%(time.time()-st))
 
         if close:
             out_stream.close()

src/calibre/ebooks/pdf/render/common.py

@@ -9,8 +9,10 @@ __docformat__ = 'restructuredtext en'
 
 import codecs, zlib
 from io import BytesIO
-from struct import pack
-from decimal import Decimal
+
+from calibre.constants import plugins, ispy3
+
+pdf_float = plugins['speedup'][0].pdf_float
 
 EOL = b'\n'
 
@@ -52,32 +54,25 @@ PAPER_SIZES = {k:globals()[k.upper()] for k in ('a0 a1 a2 a3 a4 a5 a6 b0 b1 b2'
 
 # Basic PDF datatypes {{{
 
-def format_float(f):
-    if abs(f) < 1e-7:
-        return '0'
-    places = 6
-    a, b = type(u'')(Decimal(f).quantize(Decimal(10)**-places)).partition('.')[0::2]
-    b = b.rstrip('0')
-    if not b:
-        return '0' if a == '-0' else a
-    return '%s.%s'%(a, b)
+ic = str if ispy3 else unicode
+icb = (lambda x: str(x).encode('ascii')) if ispy3 else bytes
 
 def fmtnum(o):
-    if isinstance(o, (int, long)):
-        return type(u'')(o)
-    return format_float(o)
+    if isinstance(o, float):
+        return pdf_float(o)
+    return ic(o)
 
 def serialize(o, stream):
-    if hasattr(o, 'pdf_serialize'):
-        o.pdf_serialize(stream)
-    elif isinstance(o, bool):
-        stream.write(b'true' if o else b'false')
+    if isinstance(o, float):
+        stream.write_raw(pdf_float(o).encode('ascii'))
     elif isinstance(o, (int, long)):
-        stream.write(type(u'')(o).encode('ascii'))
-    elif isinstance(o, float):
-        stream.write(format_float(o).encode('ascii'))
+        stream.write_raw(icb(o))
+    elif hasattr(o, 'pdf_serialize'):
+        o.pdf_serialize(stream)
     elif o is None:
-        stream.write(b'null')
+        stream.write_raw(b'null')
+    elif isinstance(o, bool):
+        stream.write_raw(b'true' if o else b'false')
     else:
         raise ValueError('Unknown object: %r'%o)
 
@@ -103,13 +98,6 @@ class String(unicode):
             raw = codecs.BOM_UTF16_BE + s.encode('utf-16-be')
         stream.write(b'('+raw+b')')
 
-class GlyphIndex(int):
-
-    def pdf_serialize(self, stream):
-        byts = bytearray(pack(b'>H', self))
-        stream.write('<%s>'%''.join(map(
-            lambda x: bytes(hex(x)[2:]).rjust(2, b'0'), byts)))
-
 class Dictionary(dict):
 
     def pdf_serialize(self, stream):
@@ -180,6 +168,9 @@ class Stream(BytesIO):
         super(Stream, self).write(raw if isinstance(raw, bytes) else
                                   raw.encode('ascii'))
 
+    def write_raw(self, raw):
+        BytesIO.write(self, raw)
+
 class Reference(object):
 
     def __init__(self, num, obj):

src/calibre/ebooks/pdf/render/serialize.py

@@ -9,14 +9,13 @@ __docformat__ = 'restructuredtext en'
 
 import hashlib
 from future_builtins import map
-from itertools import izip
 
 from PyQt4.Qt import QBuffer, QByteArray, QImage, Qt, QColor, qRgba
 
 from calibre.constants import (__appname__, __version__)
 from calibre.ebooks.pdf.render.common import (
     Reference, EOL, serialize, Stream, Dictionary, String, Name, Array,
-    GlyphIndex, fmtnum)
+    fmtnum)
 from calibre.ebooks.pdf.render.fonts import FontManager
 from calibre.ebooks.pdf.render.links import Links
 
@@ -166,54 +165,6 @@ class Path(object):
     def close(self):
         self.ops.append(('h',))
 
-class Text(object):
-
-    def __init__(self):
-        self.transform = self.default_transform = [1, 0, 0, 1, 0, 0]
-        self.font_name = 'Times-Roman'
-        self.font_path = None
-        self.horizontal_scale = self.default_horizontal_scale = 100
-        self.word_spacing = self.default_word_spacing = 0
-        self.char_space = self.default_char_space = 0
-        self.glyph_adjust = self.default_glyph_adjust = None
-        self.size = 12
-        self.text = ''
-
-    def set_transform(self, *args):
-        if len(args) == 1:
-            m = args[0]
-            vals = [m.m11(), m.m12(), m.m21(), m.m22(), m.dx(), m.dy()]
-        else:
-            vals = args
-        self.transform = vals
-
-    def pdf_serialize(self, stream, font_name):
-        if not self.text: return
-        stream.write_line('BT ')
-        serialize(Name(font_name), stream)
-        stream.write(' %s Tf '%fmtnum(self.size))
-        stream.write(' '.join(map(fmtnum, self.transform)) + ' Tm ')
-        if self.horizontal_scale != self.default_horizontal_scale:
-            stream.write('%s Tz '%fmtnum(self.horizontal_scale))
-        if self.word_spacing != self.default_word_spacing:
-            stream.write('%s Tw '%fmtnum(self.word_spacing))
-        if self.char_space != self.default_char_space:
-            stream.write('%s Tc '%fmtnum(self.char_space))
-        stream.write_line()
-        if self.glyph_adjust is self.default_glyph_adjust:
-            serialize(String(self.text), stream)
-            stream.write(' Tj ')
-        else:
-            chars = Array()
-            frac, widths = self.glyph_adjust
-            for c, width in izip(self.text, widths):
-                chars.append(String(c))
-                chars.append(int(width * frac))
-            serialize(chars, stream)
-            stream.write(' TJ ')
-        stream.write_line('ET')
-
-
 class Catalog(Dictionary):
 
     def __init__(self, pagetree):
@@ -244,7 +195,9 @@ class HashingStream(object):
         self.last_char = b''
 
     def write(self, raw):
-        raw = raw if isinstance(raw, bytes) else raw.encode('ascii')
+        self.write_raw(raw if isinstance(raw, bytes) else raw.encode('ascii'))
+
+    def write_raw(self, raw):
         self.f.write(raw)
         self.hashobj.update(raw)
         if raw:
@@ -420,9 +373,8 @@ class PDFStream(object):
         self.current_page.write(' %s Tf '%fmtnum(size))
         self.current_page.write('%s Tm '%' '.join(map(fmtnum, transform)))
         for x, y, glyph_id in glyphs:
-            self.current_page.write('%s %s Td '%(fmtnum(x), fmtnum(y)))
-            serialize(GlyphIndex(glyph_id), self.current_page)
-            self.current_page.write(' Tj ')
+            self.current_page.write_raw(('%s %s Td <%04X> Tj '%(
+                fmtnum(x), fmtnum(y), glyph_id)).encode('ascii'))
         self.current_page.write_line(b' ET')
 
     def get_image(self, cache_key):

src/calibre/ebooks/pdf/render/test.py

@@ -105,12 +105,15 @@ def pen(p, xmax, ymax):
     p.setPen(pen)
     p.drawRect(0, xmax/3, xmax/3, xmax/2)
 
+def text(p, xmax, ymax):
+    p.drawText(QPoint(0, ymax/3), 'Text')
+
 def main():
     app = QApplication([])
     app
     tdir = gettempdir()
     pdf = os.path.join(tdir, 'painter.pdf')
-    func = pen
+    func = full
     dpi = 100
     with open(pdf, 'wb') as f:
         dev = PdfDevice(f, xdpi=dpi, ydpi=dpi, compress=False)

src/calibre/utils/speedup.c

@@ -3,6 +3,9 @@
 
 #include <stdlib.h>
 
+#define min(x, y) ((x < y) ? x : y)
+#define max(x, y) ((x > y) ? x : y)
+
 static PyObject *
 speedup_parse_date(PyObject *self, PyObject *args) {
     const char *raw, *orig, *tz;
@@ -61,11 +64,48 @@ speedup_parse_date(PyObject *self, PyObject *args) {
             (tzh*60 + tzm)*sign*60);
 }
 
+
+static PyObject*
+speedup_pdf_float(PyObject *self, PyObject *args) {
+    double f = 0.0, a = 0.0;
+    char *buf = "0", *dot;
+    void *free_buf = NULL; 
+    int precision = 6, l = 0;
+    PyObject *ret;
+
+    if(!PyArg_ParseTuple(args, "d", &f)) return NULL;
+
+    a = fabs(f);
+
+    if (a > 1.0e-7) {
+        if(a > 1) precision = min(max(0, 6-(int)log10(a)), 6);
+        buf = PyOS_double_to_string(f, 'f', precision, 0, NULL);
+        if (buf != NULL) {
+            free_buf = (void*)buf;
+            if (precision > 0) {
+                l = strlen(buf) - 1;
+                while (l > 0 && buf[l] == '0') l--;
+                if (buf[l] == ',' || buf[l] == '.') buf[l] = 0;
+                else buf[l+1] = 0;
+                if ( (dot = strchr(buf, ',')) ) *dot = '.';
+            }
+        } else if (!PyErr_Occurred()) PyErr_SetString(PyExc_TypeError, "Float->str failed.");
+    }
+
+    ret = PyUnicode_FromString(buf);
+    if (free_buf != NULL) PyMem_Free(free_buf);
+    return ret;
+}
+
 static PyMethodDef speedup_methods[] = {
     {"parse_date", speedup_parse_date, METH_VARARGS,
         "parse_date()\n\nParse ISO dates faster."
     },
 
+    {"pdf_float", speedup_pdf_float, METH_VARARGS,
+        "pdf_float()\n\nConvert float to a string representation suitable for PDF"
+    },
+
     {NULL, NULL, 0, NULL}
 };