view: Avoid using cute unicode homomorphs for '/' and '\' and instead use ugly escaping, as the unicode method doesn't work on non-unicode supporting systems.

Annex/View.hs

@@ -226,32 +226,22 @@ toViewPath :: MetaValue -> FilePath
 toViewPath = concatMap escapeslash . fromMetaValue
   where
 	escapeslash c
-		| c == '/' = [pseudoSlash]
-		| c == '\\' = [pseudoBackslash]
-		| c == pseudoSlash = [pseudoSlash, pseudoSlash]
-		| c == pseudoBackslash = [pseudoBackslash, pseudoBackslash]
+		| c == '/' = "%_"
+		| c == '\\' = "%."
+		| c == '%' = "%%"
 		| otherwise = [c]
 
 fromViewPath :: FilePath -> MetaValue
 fromViewPath = toMetaValue . deescapeslash []
   where
 	deescapeslash s [] = reverse s
-	deescapeslash s (c:cs)
-		| c == pseudoSlash = case cs of
-			(c':cs')
-				| c' == pseudoSlash -> deescapeslash (pseudoSlash:s) cs'
-			_ -> deescapeslash ('/':s) cs
-		| c == pseudoBackslash = case cs of
-			(c':cs')
-				| c' == pseudoBackslash -> deescapeslash (pseudoBackslash:s) cs'
-			_ -> deescapeslash ('/':s) cs
-		| otherwise = deescapeslash (c:s) cs
+	deescapeslash s ('%':'_':cs) = deescapeslash ('/':s) cs
+	deescapeslash s ('%':'.':cs) = deescapeslash ('\\':s) cs
+	deescapeslash s ('%':'%':cs) = deescapeslash ('%':s) cs
+	deescapeslash s (c:cs) = deescapeslash (c:s) cs
 
-pseudoSlash :: Char
-pseudoSlash = '\8725' -- '∕' /= '/'
-
-pseudoBackslash :: Char
-pseudoBackslash = '\9586' -- '╲' /= '\'
+prop_viewPath_roundtrips :: MetaValue -> Bool
+prop_viewPath_roundtrips v = fromViewPath (toViewPath v) == v
 
 pathProduct :: [[FilePath]] -> [FilePath]
 pathProduct [] = []

Test.hs

@@ -173,6 +173,7 @@ properties = localOption (QuickCheckTests 1000) $ testGroup "QuickCheck"
 	, testProperty "prop_metadata_sane" Types.MetaData.prop_metadata_sane
 	, testProperty "prop_metadata_serialize" Types.MetaData.prop_metadata_serialize
 	, testProperty "prop_branchView_legal" Logs.View.prop_branchView_legal
+	, testProperty "prop_viewPath_roundtrips" Annex.View.prop_viewPath_roundtrips
 	, testProperty "prop_view_roundtrips" Annex.View.prop_view_roundtrips
 	, testProperty "prop_viewedFile_rountrips" Annex.View.ViewedFile.prop_viewedFile_roundtrips
 	, testProperty "prop_b64_roundtrips" Utility.Base64.prop_b64_roundtrips

debian/changelog

@@ -25,6 +25,7 @@ git-annex (6.20151219) UNRELEASED; urgency=medium
   * fix: Adjusts unlocked files as configured by annex.thin.
   * persistent-sqlite is now a hard build dependency, since v6 repository
     mode needs it.
+
   * status: On crippled filesystems, was displaying M for all annexed files
     that were present. Probably caused by a change to what git status
     displays in this situation. Fixed by treating files git thinks are
@@ -47,6 +48,9 @@ git-annex (6.20151219) UNRELEASED; urgency=medium
   * rekey: No longer copies over urls from the old to the new key.
     It makes sense for migrate to do that, but not for this low-level
     (and little used) plumbing command to.
+  * view: Avoid using cute unicode homomorphs for '/' and '\' and instead
+    use ugly escaping, as the unicode method doesn't work on non-unicode
+    supporting systems.
 
  -- Joey Hess <id@joeyh.name>  Sat, 19 Dec 2015 13:31:17 -0400
 

doc/bugs/slash_in_metadata_breaks_field__61____42___view.mdwn

@@ -39,4 +39,4 @@ I was trying out the metadata extraction via libextractor and for the mimetype t
 Apart from this git-annex is working very well for me. I mostly use it as an archive, distributing numerous copies on various hard drives and cloud providers and keeping track of what is where.Its an amazing tool for that.
 
 
-
+> [[done]] --[[Joey]] 

doc/bugs/slash_in_metadata_breaks_field__61____42___view/comment_1_249d786ce15ab0c91191987c0bab76ef._comment

@@ -0,0 +1,24 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 1"""
+ date="2016-01-08T15:52:22Z"
+ content="""
+I was able to reproduce this problem, but only with LANG=C. It works in a
+unicode locale.
+
+What's torpedoing this is a hack that it uses to handle "/" in a view.
+
+	pseudoSlash :: Char
+	pseudoSlash = '\8725' -- '∕' /= '/'
+
+It's necessary that in a view, each viewed metadata component yield exactly one
+level of directory hierarchy. Otherwise, it would be impossible to reverse
+"a/b/c/file" when viewing on 2 metadata components -- 
+is that "a/b" and "c" or "a" and "b/c"?
+
+Which is why I used this cutsey hack, but yeah, it requires working
+unicode support.
+
+Sigh, 2016 and still can't have nice things.. Suppose it'll have to use an
+ugly encoding for them instead.
+"""]]