# 
# 
# patch "ChangeLog"
#  from [1a0ea70e9c673c115a12bc000804d467b71dd570]
#    to [44bceb604bcb5b776129356bacdb9f793cd53403]
# 
# patch "cset.hh"
#  from [7bd817258ac15fa2b6e4c3473e9b97858721969d]
#    to [2d06997a4c2475aa45550dbe967c09397575d908]
# 
# patch "paths.cc"
#  from [31e473f534a09b58cab1b815a4aaf533115c9265]
#    to [b77cfcbc73e191883145492b2c7517f09365ce7f]
# 
# patch "paths.hh"
#  from [efc61a747969595d633387c4e3bb0705b7cd4a10]
#    to [97283f2f57949e69eee43772c097e1e1707481bd]
# 
# patch "roster.cc"
#  from [9d8a98859ba806ec4946bac03e274b349f149696]
#    to [f8dadb5487d8385b6572fdc658dd30f1c4481e3f]
# 
# patch "vocab.cc"
#  from [7cf1d84a1a30eebaeb5edfbb8c3a46d1598d56a4]
#    to [f74bcb1c80f9c49a91923861feaef0b8693da0c6]
# 
============================================================
--- ChangeLog	1a0ea70e9c673c115a12bc000804d467b71dd570
+++ ChangeLog	44bceb604bcb5b776129356bacdb9f793cd53403
@@ -1,3 +1,12 @@
+2006-05-17  Matt Johnston  <address@hidden>
+
+	* paths.cc (internal_string_to_split_path): move from cset.hh to
+	here, make it share the path splitting with fully_normalized_path().
+	(has_bad_chars): use lookup table to check bad chars.
+	(bad_component): don't compare string vs char* every time.
+	* roster.cc, cset.hh: internal_string_to_split_path changes
+	* vocab.cc (verify(hexenc<id>): .empty() rather than == ""
+
 2006-05-16  Matt Johnston  <address@hidden>
 
 	* roster.cc (parse_from): don't use lexical_cast, avoid copying a
============================================================
--- cset.hh	7bd817258ac15fa2b6e4c3473e9b97858721969d
+++ cset.hh	2d06997a4c2475aa45550dbe967c09397575d908
@@ -128,15 +128,4 @@
 template <> void
 dump(cset const & cs, std::string & out);
 
-
-// Some helpers.
-
-inline split_path
-internal_string_to_split_path(std::string const & str)
-{
-  split_path sp;
-  file_path_internal(str).split(sp);
-  return sp;
-}
-
 #endif // __CSET_HH__
============================================================
--- paths.cc	31e473f534a09b58cab1b815a4aaf533115c9265
+++ paths.cc	b77cfcbc73e191883145492b2c7517f09365ce7f
@@ -113,21 +113,50 @@
 static inline bool
 bad_component(std::string const & component)
 {
-  if (component == "")
+  static const std::string dot(".");
+  static const std::string dotdot("..");
+  if (component.empty())
     return true;
-  if (component == ".")
+  if (component == dot)
     return true;
-  if (component == "..")
+  if (component == dotdot)
     return true;
   return false;
 }
 
 static inline bool
-fully_normalized_path(std::string const & path)
+has_bad_chars(std::string const & path)
 {
-  // FIXME: probably should make this a 256-byte static lookup table
-  const static std::string bad_chars = std::string("\\") + constants::illegal_path_bytes + std::string(1, '\0');
-  
+  static bool bad_chars_init(false);
+  static u8 bad_table[128] = {0};
+  if (UNLIKELY(!bad_chars_init))
+    {
+      std::string bad_chars = std::string("\\") + constants::illegal_path_bytes + std::string(1, '\0');
+      for (std::string::const_iterator b = bad_chars.begin(); b != bad_chars.end(); b++)
+        {
+          u8 x = (u8)*b;
+          I((x) < sizeof(bad_table));
+          bad_table[x] = 1;
+        }
+      bad_chars_init = true;
+    }
+
+  for (std::string::const_iterator c = path.begin(); c != path.end(); c++)
+    {
+      u8 x = (u8)*c;
+      if (x < sizeof(bad_table) && bad_table[x])
+          return true;
+    }
+  return false;
+}
+
+// fully_normalized_path performs very similar function to file_path.split().
+// if want_split is set, split_path will be filled with the '/' separated
+// components of the path.
+static inline bool
+fully_normalized_path_split(std::string const & path, bool want_split,
+                            split_path & sp)
+{
   // empty path is fine
   if (path.empty())
     return true;
@@ -136,7 +165,7 @@
   if (path.size() > 1 && path[1] == ':')
     return false;
   // first scan for completely illegal bytes
-  if (path.find_first_of(bad_chars) != std::string::npos)
+  if (has_bad_chars(path))
     return false;
   // now check each component
   std::string::size_type start, stop;
@@ -146,17 +175,30 @@
       stop = path.find('/', start);
       if (stop == std::string::npos)
         {
-          if (bad_component(path.substr(start)))
+          std::string const & s(path.substr(start));
+          if (bad_component(s))
             return false;
+          if (want_split)
+            sp.push_back(s);
           break;
         }
-      if (bad_component(path.substr(start, stop - start)))
+      std::string const & s(path.substr(start, stop - start));
+      if (bad_component(s))
         return false;
+      if (want_split)
+        sp.push_back(s);
       start = stop + 1;
     }
   return true;
 }
 
+static inline bool
+fully_normalized_path(std::string const & path)
+{
+  split_path sp;
+  return fully_normalized_path_split(path, false, sp);
+}
+
 // This function considers _MTN, _MTn, _MtN, _mtn etc. to all be bookkeeping
 // paths, because on case insensitive filesystems, files put in any of them
 // may end up in _MTN instead.  This allows arbitrary code execution.  A
@@ -189,6 +231,19 @@
           && !in_bookkeeping_dir(path));
 }
 
+// equivalent to file_path_internal(path).split(sp), but
+// avoids splitting the string twice
+void
+internal_string_to_split_path(std::string const & path, split_path & sp)
+{
+  I(utf8_validate(path));
+  I(!in_bookkeeping_dir(path));
+  sp.clear();
+  sp.reserve(8);
+  sp.push_back(the_null_component);
+  I(fully_normalized_path_split(path, true, sp));
+}
+
 file_path::file_path(file_path::source_type type, std::string const & path)
 {
   MM(path);
============================================================
--- paths.hh	efc61a747969595d633387c4e3bb0705b7cd4a10
+++ paths.hh	97283f2f57949e69eee43772c097e1e1707481bd
@@ -254,4 +254,7 @@
 
 typedef std::set<split_path> path_set;
 
+// equivalent to file_path_internal(path).split(sp) but more efficient.
+void internal_string_to_split_path(std::string const & path, split_path & sp);
+
 #endif
============================================================
--- roster.cc	9d8a98859ba806ec4946bac03e274b349f149696
+++ roster.cc	f8dadb5487d8385b6572fdc658dd30f1c4481e3f
@@ -2516,7 +2516,9 @@
       else
         {
           I(!pth.empty());
-          attach_node(n->self, internal_string_to_split_path(pth));
+          split_path sp;
+          internal_string_to_split_path(pth, sp);
+          attach_node(n->self, sp);
         }
 
       // Non-dormant attrs
============================================================
--- vocab.cc	7cf1d84a1a30eebaeb5edfbb8c3a46d1598d56a4
+++ vocab.cc	f74bcb1c80f9c49a91923861feaef0b8693da0c6
@@ -61,7 +61,7 @@
   if (val.ok)
     return;
 
-  if (val() == "")
+  if (val().empty())
     return;
 
   N(val().size() == constants::idlen,