Support linked images

4 files changed, 104 insertions, 16 deletions

diff --git a/mammoth/docx/__init__.py b/mammoth/docx/__init__.py
index ec8d363..1ca8cb0 100644
--- a/mammoth/docx/__init__.py
+++ b/mammoth/docx/__init__.py
@@ -1,5 +1,6 @@
 import zipfile
 import contextlib
+import os
 
 from .. import results, lists
 from .xmlparser import parse_xml
@@ -9,6 +10,7 @@ from .relationships_xml import read_relationships_xml_element, Relationships
 from .numbering_xml import read_numbering_xml_element, Numbering
 from .styles_xml import read_styles_xml_element
 from .notes_xml import create_footnotes_reader, create_endnotes_reader
+from .files import Files
 from . import body_xml
 
 
@@ -26,7 +28,7 @@ _namespaces = [
 
 def read(fileobj):
     zip_file = zipfile.ZipFile(fileobj)
-    body_readers = _body_readers(zip_file)
+    body_readers = _body_readers(getattr(fileobj, "name"), zip_file)
     
     return _read_notes(zip_file, body_readers).bind(lambda notes:
         _read_document(zip_file, body_readers, notes))
@@ -55,7 +57,7 @@ def _read_document(zip_file, body_readers, notes):
         )
 
 
-def _body_readers(zip_file):
+def _body_readers(document_path, zip_file):
     with _open_entry(zip_file, "[Content_Types].xml") as content_types_fileobj:
         content_types = read_content_types_xml_element(_parse_docx_xml(content_types_fileobj))
 
@@ -77,6 +79,7 @@ def _body_readers(zip_file):
             relationships=relationships,
             styles=styles,
             docx_file=zip_file,
+            files=Files(os.path.dirname(document_path)),
         )
     
     return for_name
diff --git a/mammoth/docx/body_xml.py b/mammoth/docx/body_xml.py
index cb42f05..54f9dd3 100644
--- a/mammoth/docx/body_xml.py
+++ b/mammoth/docx/body_xml.py
@@ -10,7 +10,8 @@ def reader(numbering=None,
         content_types=None,
         relationships=None,
         styles=None,
-        docx_file=None):
+        docx_file=None,
+        files=None):
     
     read, read_all = _create_reader(
         numbering=numbering,
@@ -18,6 +19,7 @@ def reader(numbering=None,
         relationships=relationships,
         styles=styles,
         docx_file=docx_file,
+        files=files,
     )
     return _BodyReader(read, read_all)
 
@@ -37,7 +39,7 @@ class _BodyReader(object):
         return results.Result(result.value, result.messages)
 
 
-def _create_reader(numbering, content_types, relationships, styles, docx_file):
+def _create_reader(numbering, content_types, relationships, styles, docx_file, files):
     _handlers = {}
     _ignored_elements = set([
         "w:bookmarkStart",
@@ -221,17 +223,9 @@ def _create_reader(numbering, content_types, relationships, styles, docx_file):
         return _combine_results(map(lambda blip: _read_blip(blip, alt_text), blips))
     
     def _read_blip(element, alt_text):
-        relationship_id = element.attributes["r:embed"]
-        image_path = "word/" + relationships[relationship_id].target.lstrip("/")
+        image_path, open_image = _read_blip_image(element)
         content_type = content_types.find_content_type(image_path)
         
-        def open_image():
-            image_file = docx_file.open(image_path)
-            if hasattr(image_file, "__exit__"):
-                return image_file
-            else:
-                return contextlib.closing(image_file)
-        
         image = documents.image(alt_text=alt_text, content_type=content_type, open=open_image)
         
         if content_type in ["image/png", "image/gif", "image/jpeg", "image/svg+xml", "image/tiff"]:
@@ -241,6 +235,35 @@ def _create_reader(numbering, content_types, relationships, styles, docx_file):
             
         return _element_result_with_messages(image, messages)
     
+    def _read_blip_image(element):
+        embed_relationship_id = element.attributes.get("r:embed")
+        link_relationship_id = element.attributes.get("r:link")
+        if embed_relationship_id is not None:
+            return _read_embedded_blip_image(embed_relationship_id)
+        elif link_relationship_id is not None:
+            return _read_linked_blip_image(link_relationship_id)
+    
+    def _read_embedded_blip_image(relationship_id):
+        image_path = "word/" + relationships[relationship_id].target.lstrip("/")
+        
+        def open_image():
+            image_file = docx_file.open(image_path)
+            if hasattr(image_file, "__exit__"):
+                return image_file
+            else:
+                return contextlib.closing(image_file)
+        
+        return image_path, open_image
+    
+    
+    def _read_linked_blip_image(relationship_id):
+        image_path = relationships[relationship_id].target
+        
+        def open_image():
+            return files.open(image_path)
+        
+        return image_path, open_image
+    
     def note_reference_reader(note_type):
         @handler("w:{0}Reference".format(note_type))
         def note_reference(element):
diff --git a/mammoth/docx/content_types_xml.py b/mammoth/docx/content_types_xml.py
index a2eb14c..f0406b7 100644
--- a/mammoth/docx/content_types_xml.py
+++ b/mammoth/docx/content_types_xml.py
@@ -7,7 +7,10 @@ def read_content_types_xml_element(element):
         _read_override,
         element.find_children("content-types:Override")
     ))
-    return ContentTypes(extension_defaults, overrides)
+    return _ContentTypes([
+        _XmlContentTypes(extension_defaults, overrides),
+        _FallbackContentTypes(),
+    ])
 
 
 def _read_default(element):
@@ -22,7 +25,37 @@ def _read_override(element):
     return part_name.lstrip("/"), content_type
 
 
-class ContentTypes(object):
+class _ContentTypes(object):
+    def __init__(self, finders):
+        self._finders = finders
+    
+    def find_content_type(self, path):
+        for finder in self._finders:
+            content_type = finder.find_content_type(path)
+            if content_type is not None:
+                return content_type
+
+
+class _FallbackContentTypes(object):
+    _image_content_types = {
+        "png": "png",
+        "gif": "gif",
+        "jpeg": "jpeg",
+        "jpg": "jpeg",
+        "tif": "tiff",
+        "tiff": "tiff",
+        "bmp": "bmp",
+    }
+    
+    def find_content_type(self, path):
+        extension = _get_extension(path).lower()
+        if extension in self._image_content_types:
+            return "image/" + self._image_content_types[extension]
+        else:
+            return None
+
+
+class _XmlContentTypes(object):
     def __init__(self, extension_defaults, overrides):
         self._extension_defaults = extension_defaults
         self._overrides = overrides
@@ -31,5 +64,9 @@ class ContentTypes(object):
         if path in self._overrides:
             return self._overrides[path]
         else:
-            extension = path.rpartition(".")[2]
+            extension = _get_extension(path)
             return self._extension_defaults.get(extension)
+
+
+def _get_extension(path):
+    return path.rpartition(".")[2]
diff --git a/mammoth/docx/files.py b/mammoth/docx/files.py
new file mode 100644
index 0000000..d919de9
--- /dev/null
+++ b/mammoth/docx/files.py
@@ -0,0 +1,25 @@
+import os
+import contextlib
+try:
+    from urllib2 import urlopen
+except ImportError:
+    from urllib.request import urlopen
+try:
+    from urllib.parse import urlparse
+except ImportError:
+    from urlparse import urlparse
+
+
+class Files(object):
+    def __init__(self, base):
+        self._base = base
+    
+    def open(self, uri):
+        if _is_absolute(uri):
+            return contextlib.closing(urlopen(uri))
+        else:
+            return open(os.path.join(self._base, uri), "rb")
+
+
+def _is_absolute(url):
+    return urlparse(url).scheme != ""