git-svn-id: https://pdfminerr.googlecode.com/svn/trunk/pdfminer@14 1aa58f4a-7d42-0410-adbc-911cccaed67c

extent.py

@@ -0,0 +1,89 @@
+#!/usr/bin/env python
+import sys
+INF = sys.maxint
+
+
+##  Rect
+##
+class Rect:
+  
+  def __init__(self, x0=-INF, y0=-INF, w=None, h=None):
+    self.x0 = x0
+    self.y0 = y0
+    if w == None:
+      self.x1 = INF
+    else:      
+      self.x1 = x0+w
+    if h == None:
+      self.y1 = INF
+    else:
+      self.y1 = y0+h
+    return
+
+  def overlap(self, rect):
+    return not (rect.x1 <= self.x0 or self.x1 <= rect.x0 or
+                rect.y1 <= self.y0 or self.y1 <= rect.y0)
+
+
+##  ExtSet
+##
+class ExtSet:
+  
+  def __init__(self, gridsize):
+    self.gridsize = gridsize
+    self.grid = []
+    return
+  
+  def cells(self, x0, x1):
+    i = int(x0 / self.gridsize)
+    x = i * self.gridsize
+    while x < x1:
+      yield i
+      x += self.gridsize
+      i += 1
+    return
+  
+  def add(self, x0, x1, obj):
+    for i in self.cells(x0, x1):
+      self.grid[i].append(obj)
+    return
+  
+  def get(self, x0, x1):
+    objs = set()
+    for i in self.cells(x0, x1):
+      objs.update(self.grid[i])
+    return objs
+
+def test_extset():
+  e=ExtSet(10)
+  assert list(e.cells(-1, 1)) == [-1,0]
+  assert list(e.cells(0, 1)) == [0]
+  assert list(e.cells(0, 10)) == [0]
+  assert list(e.cells(0, 11)) == [0,1]
+  assert list(e.cells(1, 11)) == [0,1]
+  assert list(e.cells(10, 11)) == [1]
+  assert list(e.cells(0, 20)) == [0,1]
+  assert list(e.cells(10, 20)) == [1]
+  assert list(e.cells(1,21)) == [0,1,2]
+  assert list(e.cells(11,21)) == [1,2]
+  return
+
+
+##  ExtGrid
+##
+class ExtGrid:
+  
+  def __init__(self, gridsize):
+    self.hext = ExtSet(gridsize)
+    self.vext = ExtSet(gridsize)
+    return
+  
+  def add(self, rect):
+    self.hext.add(rect.x0, rect.x1, rect)
+    self.vext.add(rect.y0, rect.y1, rect)
+    return
+  
+  def get(self, rect):
+    rects = self.hext.get(rect.x0, rect.x1)
+    rects.update_intersect(self.vext.get(rect.y0, rect.y1))
+    return rects

pdf2txt.py

@@ -23,13 +23,20 @@ class TextConverter(PDFDevice):
     self.outfp.write('\n')
     return
   
-  def begin_block(self, name, (x0,y0,x1,y1)):
-    self.outfp.write('<block name="%s" x0="%d" y0="%d" x1="%d" y1="%d">\n' %
+  def begin_page(self, name, (x0,y0,x1,y1)):
+    self.outfp.write('<page name="%s" x0="%d" y0="%d" x1="%d" y1="%d">\n' %
                      (name,x0,y0,x1,y1))
     return
-  
-  def end_block(self):
-    self.outfp.write('</block>\n')
+  def end_page(self, _):
+    self.outfp.write('</page>\n')
+    return
+
+  def begin_figure(self, name, (x0,y0,x1,y1)):
+    self.outfp.write('<figure name="%s" x0="%d" y0="%d" x1="%d" y1="%d">\n' %
+                     (name,x0,y0,x1,y1))
+    return
+  def end_figure(self, _):
+    self.outfp.write('</figure>\n')
     return
 
   def handle_undefined_char(self, cidcoding, cid):
@@ -73,6 +80,7 @@ class TextConverter(PDFDevice):
 # pdf2txt
 def pdf2txt(outfp, rsrc, fname, pages, codec, debug=0):
   device = TextConverter(outfp, rsrc, codec)
+  outfp.write('<document>')
   doc = PDFDocument(debug=debug)
   fp = file(fname)
   parser = PDFParser(doc, fp, debug=debug)
@@ -81,6 +89,7 @@ def pdf2txt(outfp, rsrc, fname, pages, codec, debug=0):
     if pages and (i not in pages): continue
     interpreter.process_page(page)
   fp.close()
+  outfp.write('</document>')
   device.close()
   return
 

pdfinterp.py

@@ -189,7 +189,8 @@ class PDFType3Font(PDFSimpleFont):
     if 'FontDescriptor' in spec:
       descriptor = dict_value(spec['FontDescriptor'])
     else:
-      descriptor = {'FontName':None, 'Ascent':0, 'Descent':0,
+      descriptor = {'FontName':spec.get('Name'),
+                    'Ascent':0, 'Descent':0,
                     'FontBBox':spec['FontBBox']}
     PDFSimpleFont.__init__(self, descriptor, widths, spec)
     return
@@ -442,9 +443,13 @@ class PDFDevice:
     self.ctm = ctm
     return
 
-  def begin_block(self, name, bbox):
+  def begin_page(self, name, bbox):
     return
-  def end_block(self):
+  def end_page(self, name):
+    return
+  def begin_figure(self, name, bbox):
+    return
+  def end_figure(self, name):
     return
   
   def render_string(self, textstate, textmatrix, size, seq):
@@ -820,26 +825,23 @@ class PDFPageInterpreter:
       ctm = mult_matrix(xobj.dic.get('Matrix', MATRIX_IDENTITY), self.ctm)
       (x0,y0) = apply_matrix(ctm, (x0,y0))
       (x1,y1) = apply_matrix(ctm, (x1,y1))
-      interpreter.render_contents(xobjid,
-                                  (x0,y0,x1,y1),
-                                  xobj.dic.get('Resources'),
-                                  [xobj],
-                                  ctm=ctm)
+      bbox = (x0,y0,x1,y1)
+      self.device.begin_figure(xobjid, bbox)
+      interpreter.render_contents(xobj.dic.get('Resources'),
+                                  [xobj], ctm=ctm)
+      self.device.end_figure(xobjid)
     return
 
   def process_page(self, page):
     if 1 <= self.debug:
       print >>stderr, 'Processing page: %r' % page
-    self.render_contents('page-%d' % page.pageid,
-                         page.mediabox,
-                         page.resources,
-                         page.contents)
+    self.device.begin_page(page.pageid, page.mediabox)
+    self.render_contents(page.resources, page.contents)
+    self.device.end_page(page.pageid)
     return
 
-  def render_contents(self, contid, mediabox, resources, contents,
-                      ctm=MATRIX_IDENTITY):
+  def render_contents(self, resources, contents, ctm=MATRIX_IDENTITY):
     self.initpage(ctm)
-    self.device.begin_block(contid, mediabox)
     # Handle resource declarations.
     def get_colorspace(spec):
       if isinstance(spec, list):
@@ -874,7 +876,6 @@ class PDFPageInterpreter:
     data = ''.join( stream_value(stream).get_data() 
                     for stream in list_value(contents) )
     self.execute(data)
-    self.device.end_block()
     return
   
   def execute(self, data):

pdfparser.py

@@ -117,7 +117,7 @@ def str_value(x):
 
 def list_value(x):
   x = resolve1(x)
-  if not isinstance(x, list):
+  if not (isinstance(x, list) or isinstance(x, tuple)):
     raise PDFTypeError('list required: %r' % x)
   return x