Docx reader: Recognize media inside textboxes (#11515)

Closes #11053.

Author: RayBB

src/Text/Pandoc/Readers/Docx/Parse.hs

@@ -160,21 +160,32 @@ unwrapElement ns element
   | isElem ns "w" "smartTag" element
   = concatMap (unwrapElement ns) (elChildren element)
   | isElem ns "w" "p" element
-  , textboxes@(_:_) <- findChildrenByName ns "w" "r" element >>=
-                       findChildrenByName ns "mc" "AlternateContent" >>=
-                       findChildrenByName ns "mc" "Fallback" >>=
-                       findChildrenByName ns "w" "pict" >>=
-                       (\e -> findChildrenByName ns "v" "shape" e <>
-                              findChildrenByName ns "v" "rect" e) >>=
-                       findChildrenByName ns "v" "textbox" >>=
-                       findChildrenByName ns "w" "txbxContent"
-  = concatMap (unwrapElement ns) (concatMap elChildren textboxes) -- handle #9214
+  , textboxes@(_:_) <- findChildrenByName ns "w" "r" element >>= findTextboxes
+  = concatMap (unwrapElement ns) (concatMap elChildren textboxes) -- handle #9214, #11053
   | isElem ns "w" "r" element
   , Just fallback <- findChildByName ns "mc" "AlternateContent" element >>=
                      findChildByName ns "mc" "Fallback"
   = [element{ elContent = concatMap (unwrapContent ns) (elContent fallback) }]
   | otherwise
   = [element{ elContent = concatMap (unwrapContent ns) (elContent element) }]
+  where
+    -- Search textbox content in the run's effective children.
+    -- If AlternateContent is present, use only the fallback branch,
+    -- matching the w:r unwrapping logic and avoiding duplicate textbox
+    -- extraction when both direct and fallback encodings are present.
+    findRunFallback run =
+      findChildByName ns "mc" "AlternateContent" run >>=
+      findChildByName ns "mc" "Fallback"
+    findTextboxes run =
+      findTextboxContent =<<
+      case findRunFallback run of
+        Just fallback -> findChildrenByName ns "w" "pict" fallback
+        Nothing -> findChildrenByName ns "w" "pict" run
+    findTextboxContent pict =
+      (findChildrenByName ns "v" "shape" pict <>
+       findChildrenByName ns "v" "rect" pict) >>=
+      findChildrenByName ns "v" "textbox" >>=
+      findChildrenByName ns "w" "txbxContent"
 
 unwrapContent :: NameSpaces -> Content -> [Content]
 unwrapContent ns (Elem element) = map Elem $ unwrapElement ns element

test/Tests/Readers/Docx.hs

@@ -215,6 +215,14 @@ tests = [ testGroup "document"
             "text in shape format"
             "docx/text_in_shape_format.docx"
             "docx/text_in_shape_format.native"
+          , testCompare
+            "image inside textbox content"
+            "docx/textbox_image.docx"
+            "docx/textbox_image.native"
+          , testCompare
+            "image inside textbox content with duplicate encoding"
+            "docx/textbox_image_duplicate_encoding.docx"
+            "docx/textbox_image_duplicate_encoding.native"
           ]
         , testGroup "blocks"
           [ testCompare
@@ -473,6 +481,12 @@ tests = [ testGroup "document"
           [ testMediaBag
             "image extraction"
             "docx/image.docx"
+          , testMediaBag
+            "image inside textbox content populates media bag"
+            "docx/textbox_image.docx"
+          , testMediaBag
+            "image inside textbox content with duplicate encoding populates media bag"
+            "docx/textbox_image_duplicate_encoding.docx"
           ]
         , testGroup "custom styles"
           [ testCompare

test/docx/textbox_image.docx

@@ -0,0 +1,30 @@
+[ Para
+    [ Str "The"
+    , Space
+    , Str "image"
+    , Space
+    , Str "is"
+    , Space
+    , Str "below."
+    ]
+, Para
+    [ Image
+        ( ""
+        , []
+        , [ ( "width" , "4.543038057742782in" )
+          , ( "height" , "2.9166666666666665in" )
+          ]
+        )
+        []
+        ( "media/image1.png" , "" )
+    ]
+, Para
+    [ Str "The"
+    , Space
+    , Str "image"
+    , Space
+    , Str "is"
+    , Space
+    , Str "above."
+    ]
+]

test/docx/textbox_image.native

@@ -0,0 +1,30 @@
+[ Para
+    [ Str "The"
+    , Space
+    , Str "image"
+    , Space
+    , Str "is"
+    , Space
+    , Str "below."
+    ]
+, Para
+    [ Image
+        ( ""
+        , []
+        , [ ( "width" , "4.543038057742782in" )
+          , ( "height" , "2.9166666666666665in" )
+          ]
+        )
+        []
+        ( "media/image1.png" , "" )
+    ]
+, Para
+    [ Str "The"
+    , Space
+    , Str "image"
+    , Space
+    , Str "is"
+    , Space
+    , Str "above."
+    ]
+]