Add support for … and … XML entities.

[dead/halcyon.git] / src / Twitter / Xml.hs

diff --git a/src/Twitter/Xml.hs b/src/Twitter/Xml.hs
index 09013bf0cc9b623b7ca8ef3a12aeac31ea467cca..b34c6fbae94c4ca8629eec48ad849becb98c23e8 100644 (file)
--- a/src/Twitter/Xml.hs
+++ b/src/Twitter/Xml.hs
@@ -1,48 +1,72 @@
+-- |Application-specific XML functions.
 module Twitter.Xml
 where
 
 import Data.Maybe
+import Test.HUnit
 import Text.Regex (mkRegex, subRegex)
 import Text.XML.HaXml
 
+-- |Returns the 'CharData' contained within the given 'Content', or
+-- 'Nothing' if no acceptable CharData was found. It will parse either
+-- a 'CString' ('String') or 'CRef' (XML entity reference).
 get_char_data :: Content -> (Maybe CharData)
 get_char_data (CString _ cd) = Just cd
 get_char_data (CRef ref) = Just (verbatim ref) -- Entities.
 get_char_data _ = Nothing
 
 
+-- |A 'CFilter' returning all <status> tags within <statuses>.
 all_statuses :: CFilter
 all_statuses = (tag "statuses" /> tag "status")
 
--- Called unique_id here because status_id is used elsewhere.
+-- |Finds the text of the <id> element contained within some other
+-- content. Called unique_id here because status_id is used elsewhere.
 unique_id :: CFilter
 unique_id = keep /> (tag "id") /> txt
 
+-- |Finds the text of the <created_at> element contained within some
+-- other element.
 status_created_at :: CFilter
 status_created_at = keep /> (tag "created_at") /> txt
 
+-- |Finds the text of the <text> element contained within some
+-- other element.
 status_text :: CFilter
 status_text = keep /> (tag "text") /> txt
 
+-- |Finds the XML of the <user> element contained within some other
+-- element.
 status_user :: CFilter
 status_user = keep /> (tag "user")
-              
+
+-- |Finds the text of the <screen_name> element contained within some
+-- other element.
 user_screen_name :: CFilter
 user_screen_name = keep /> (tag "screen_name") /> txt
 
 
+-- |A list of tuples whose first entry is a regular expression
+-- matching XML entities, and whose second entry is the ASCII
+-- character represented by that entity.
 xml_entities :: [(String, String)]
 xml_entities = [("[lr]dquo", "\""),
+                ("quot",     "\""),
                 ("[mn]dash", "-"),
                 ("nbsp",     " "),
                 ("#8217",    "'"),
                 ("amp",      "&"),
                 ("lt",       "<"),
-                ("gt",       ">")]
+                ("gt",       ">"),
+                ("#8230",    "..."),
+                ("hellip",   "...")]
 
+-- |Replace all of the XML entities in target.
 replace_entities :: String -> String
 replace_entities target = unescape_recursive xml_entities target
 
+-- |The recursive function which does the real work for
+-- 'replace_entities'.
 unescape_recursive :: [(String, String)] -> String -> String
 unescape_recursive [] target = target
 unescape_recursive replacements target =
@@ -51,3 +75,16 @@ unescape_recursive replacements target =
       replacement = (replacements !! 0)
       from = "&" ++ (fst replacement) ++ ";"
       to = (snd replacement)
+
+
+
+xml_tests :: [Test]
+xml_tests = [ test_replace_entities ]
+
+
+test_replace_entities :: Test
+test_replace_entities =
+    TestCase $ assertEqual "All entities are replaced correctly." expected_text actual_text
+    where
+      actual_text = (replace_entities "&quot;The moon is gay&#8230;&hellip;&quot; said &lt;insert the current president of the United States of America&gt;. &ldquo;It&#8217;s OK&mdash;&ndash;he&#8217;s not a real doctor.&rdquo;")
+      expected_text = "\"The moon is gay......\" said <insert the current president of the United States of America>. \"It's OK--he's not a real doctor.\""