Re: path canonicalization problem (issue #559)

On Mon, Jan 21, 2002 at 11:37:50AM -0600, Karl Fogel wrote:
> Great, thanks Garrett! I've left issue #559 assigned to Branko;
> recommend discussing with him whatever solution you come up with, it
> seems that he's thought about this a lot and will have useful things
> to say, no matter who writes the code.

here's a preliminary patch to make svn_path_canonicalize do some
actual canonicalization. right now it just removes extraneous /'s and
.'s from paths. eventually i'd like it to also take care of extra
/../'s, but that's going to take a bit more work.

this passes 'make check', and seems to work for me...

-garrett

Make svn_path_canonicalize do some more canonicalization.

* subversion/libsvn_subr/path.c
  (svn_path_canonicalize): remove superfluous /'s and .'s from the
  path. also, be careful to not do such things inside the host portion
  of a url, since many urls are passed into this function. add a
  comment noting that having this function turn relative paths into
  absolute paths (which would simplify removal of extraneous ..'s) is a
  bad idea, and causes many many things to break.
* subversion/include/svn_path.h
  (svn_path_canonicalize): update docstring.

Index: subversion/include/svn_path.h
===================================================================
--- subversion/include/svn_path.h
+++ subversion/include/svn_path.h Mon Jan 21 14:05:00 2002
@@ -108,9 +108,9 @@
 int svn_path_is_empty (const svn_stringbuf_t *path);
 
 
-/* Remove trailing separators that don't affect the meaning of the path.
- (At some future point, this may make other semantically inoperative
- transformations.) */
+/* Remove trailing separators, extra separators, and extraneous references
+ to '.', which don't affect the meaning of the path. (At some future
+ point, this should also remove unnecessary references to '..'.) */
 void svn_path_canonicalize (svn_stringbuf_t *path);
 
 
Index: subversion/libsvn_subr/path.c
===================================================================
--- subversion/libsvn_subr/path.c
+++ subversion/libsvn_subr/path.c Mon Jan 21 14:27:21 2002
@@ -85,11 +85,40 @@
 void
 svn_path_canonicalize (svn_stringbuf_t *path)
 {
- /* At some point this could eliminate redundant components.
- For now, it just makes sure there is no trailing slash. */
+ char *start_here = NULL;
+ char *tmp, *tmp1;
 
- /* kff todo: maybe should be implemented with a new routine in
- libsvn_string. */
+ /* if this is a url, we want to start condensing at the third slash,
+ otherwise we start at the beginning of the path */
+ if (svn_path_is_url ((const svn_string_t *)path))
+ {
+ int i, num_slashes = 0;
+
+ for (i = 0; i < path->len; ++i)
+ if (path->data[i] == '/')
+ if (++num_slashes == 3)
+ start_here = &path->data[i];
+
+ if (start_here == NULL)
+ return; /* this should never happen */
+ }
+ else
+ start_here = path->data;
+
+ /* condense extra /'s and /./'s */
+ while ((tmp = strstr (start_here, "//")) != NULL)
+ strcpy (tmp, tmp + 1);
+
+ while ((tmp = strstr (start_here, "/./")) != NULL)
+ strcpy (tmp, tmp + 2);
+
+ /* ### should also remove ..'s where possible, but that code will be a bit
+ icky, and i'm feeling lazy at the moment. we can't take the easy way out
+ and just turn our paths into absolute paths, which would make this
+ considerably easier, since doing that breaks all sorts of things for
+ reasons i don't really understand. - rooneg */
+
+ path->len = strlen(path->data);
 
   /* Remove trailing separators from the end of the path. */
   while ((path->len > 0)

-- 
garrett rooney                     Unix was not designed to stop you from 
rooneg@electricjellyfish.net       doing stupid things, because that would  
http://electricjellyfish.net/      stop you from doing clever things.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org