From 4134dad500f6cb360a9ceb957e0c5ec2ec59b2a8 Mon Sep 17 00:00:00 2001
From: John MacFarlane <jgm@berkeley.edu>
Date: Sun, 10 Jul 2011 16:54:46 -0700
Subject: Make HTML reader more forgiving of bad HTML.

* Skip spaces after <b>, <emph>, etc.
* Convert Plain elements into Para when they're in a list
  item with Para, Pre, BlockQuote, CodeBlock.

An example of HTML that pandoc handles better now:

~~~~
<h4> Testing html to markdown </h4>
<ul>
<li>
<b> An item in a list </b>
<p> An introductory sentence.
<pre>
Some preformatted text
at this stage comes next.

But alas! much havoc
is wrought by Pandoc.
</pre>
</ul>
~~~~

Thanks to Dirk Laurie for reporting the issues.
---
 src/Text/Pandoc/Readers/HTML.hs | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)

(limited to 'src')

diff --git a/src/Text/Pandoc/Readers/HTML.hs b/src/Text/Pandoc/Readers/HTML.hs
index 18e3113d3..ba25d8ad8 100644
--- a/src/Text/Pandoc/Readers/HTML.hs
+++ b/src/Text/Pandoc/Readers/HTML.hs
@@ -107,7 +107,7 @@ pBulletList = try $ do
   -- treat it as a list item, though it's not valid xhtml...
   skipMany nonItem
   items <- manyTill (pInTags "li" block >>~ skipMany nonItem) (pCloses "ul")
-  return [BulletList items]
+  return [BulletList $ map fixPlains items]
 
 pOrderedList :: TagParser [Block]
 pOrderedList = try $ do
@@ -138,7 +138,7 @@ pOrderedList = try $ do
   -- treat it as a list item, though it's not valid xhtml...
   skipMany nonItem
   items <- manyTill (pInTags "li" block >>~ skipMany nonItem) (pCloses "ol")
-  return [OrderedList (start, style, DefaultDelim) items]
+  return [OrderedList (start, style, DefaultDelim) $ map fixPlains items]
 
 pDefinitionList :: TagParser [Block]
 pDefinitionList = try $ do
@@ -154,7 +154,19 @@ pDefListItem = try $ do
   defs  <- many1 (try $ skipMany nonItem >> pInTags "dd" block)
   skipMany nonItem
   let term = intercalate [LineBreak] terms
-  return (term, defs)
+  return (term, map fixPlains defs)
+
+fixPlains :: [Block] -> [Block]
+fixPlains bs = if any isParaish bs
+                  then map plainToPara bs
+                  else bs
+  where isParaish (Para _) = True
+        isParaish (CodeBlock _ _) = True
+        isParaish (Header _ _) = True
+        isParaish (BlockQuote _) = True
+        isParaish _        = False
+        plainToPara (Plain xs) = Para xs
+        plainToPara x = x
 
 pRawTag :: TagParser String
 pRawTag = do
@@ -358,7 +370,7 @@ pInlinesInTags :: String -> ([Inline] -> Inline)
                -> TagParser [Inline]
 pInlinesInTags tagtype f = do
   contents <- pInTags tagtype inline
-  return [f contents]
+  return [f $ normalizeSpaces contents]
 
 pInTags :: String -> TagParser [a]
         -> TagParser [a]
-- 
cgit v1.2.3