[PATCH] Rewrite ls-tree to behave more like "/bin/ls -a"
[git.git] / ls-tree.c
index 26cea3f..d2a57d0 100644 (file)
--- a/ls-tree.c
+++ b/ls-tree.c
  * Copyright (C) Linus Torvalds, 2005
  */
 #include "cache.h"
+#include "blob.h"
+#include "tree.h"
 
 static int line_termination = '\n';
-static int recursive = 0;
+#define LS_RECURSIVE 1
+#define LS_TREE_ONLY 2
+static int ls_options = 0;
 
-struct path_prefix {
-       struct path_prefix *prev;
-       const char *name;
-};
+static struct tree_entry_list root_entry;
 
-#define DEBUG(fmt, ...)        
+static void prepare_root(unsigned char *sha1)
+{
+       unsigned char rsha[20];
+       unsigned long size;
+       void *buf;
+       struct tree *root_tree;
+
+       buf = read_object_with_reference(sha1, "tree", &size, rsha);
+       free(buf);
+       if (!buf)
+               die("Could not read %s", sha1_to_hex(sha1));
+
+       root_tree = lookup_tree(rsha);
+       if (!root_tree)
+               die("Could not read %s", sha1_to_hex(sha1));
+
+       /* Prepare a fake entry */
+       root_entry.directory = 1;
+       root_entry.executable = root_entry.symlink = 0;
+       root_entry.mode = S_IFDIR;
+       root_entry.name = "";
+       root_entry.item.tree = root_tree;
+       root_entry.parent = NULL;
+}
 
-static int string_path_prefix(char *buff, size_t blen, struct path_prefix *prefix)
+static int prepare_children(struct tree_entry_list *elem)
 {
-       int len = 0;
-       if (prefix) {
-               if (prefix->prev) {
-                       len = string_path_prefix(buff,blen,prefix->prev);
-                       buff += len;
-                       blen -= len;
-                       if (blen > 0) {
-                               *buff = '/';
-                               len++;
-                               buff++;
-                               blen--;
-                       }
-               }
-               strncpy(buff,prefix->name,blen);
-               return len + strlen(prefix->name);
+       if (!elem->directory)
+               return -1;
+       if (!elem->item.tree->object.parsed) {
+               struct tree_entry_list *e;
+               if (parse_tree(elem->item.tree))
+                       return -1;
+               /* Set up the parent link */
+               for (e = elem->item.tree->entries; e; e = e->next)
+                       e->parent = elem;
        }
-
        return 0;
 }
 
-static void print_path_prefix(struct path_prefix *prefix)
+static struct tree_entry_list *find_entry_0(struct tree_entry_list *elem,
+                                           const char *path,
+                                           const char *path_end)
 {
-       if (prefix) {
-               if (prefix->prev) {
-                       print_path_prefix(prefix->prev);
-                       putchar('/');
+       const char *ep;
+       int len;
+
+       while (path < path_end) {
+               if (prepare_children(elem))
+                       return NULL;
+
+               /* In elem->tree->entries, find the one that has name
+                * that matches what is between path and ep.
+                */
+               elem = elem->item.tree->entries;
+
+               ep = strchr(path, '/');
+               if (!ep || path_end <= ep)
+                       ep = path_end;
+               len = ep - path;
+
+               while (elem) {
+                       if ((strlen(elem->name) == len) &&
+                           !strncmp(elem->name, path, len))
+                               break;
+                       elem = elem->next;
                }
-               fputs(prefix->name, stdout);
+               if (path_end <= ep || !elem)
+                       return elem;
+               while (*ep == '/' && ep < path_end)
+                       ep++;
+               path = ep;
        }
+       return NULL;
 }
 
-/*
- * return:
- *     -1 if prefix is *not* a subset of path
- *      0 if prefix == path
- *      1 if prefix is a subset of path
- */
-static int pathcmp(const char *path, struct path_prefix *prefix)
+static struct tree_entry_list *find_entry(const char *path,
+                                         const char *path_end)
 {
-       char buff[PATH_MAX];
-       int len,slen;
-
-       if (prefix == NULL)
-               return 1;
-
-       len = string_path_prefix(buff, sizeof buff, prefix);
-       slen = strlen(path);
-
-       if (slen < len)
-               return -1;
+       /* Find tree element, descending from root, that
+        * corresponds to the named path, lazily expanding
+        * the tree if possible.
+        */
+       if (path == path_end) {
+               /* Special.  This is the root level */
+               return &root_entry;
+       }
+       return find_entry_0(&root_entry, path, path_end);
+}
 
-       if (strncmp(path,buff,len) == 0) {
-               if (slen == len)
-                       return 0;
-               else
-                       return 1;
+static void show_entry_name(struct tree_entry_list *e)
+{
+       /* This is yucky.  The root level is there for
+        * our convenience but we really want to do a
+        * forest.
+        */
+       if (e->parent && e->parent != &root_entry) {
+               show_entry_name(e->parent);
+               putchar('/');
        }
+       printf("%s", e->name);
+}
 
-       return -1;
-}      
+static const char *entry_type(struct tree_entry_list *e)
+{
+       return (e->directory ? "tree" : "blob");
+}
 
-/*
- * match may be NULL, or a *sorted* list of paths
- */
-static void list_recursive(void *buffer,
-                          const char *type,
-                          unsigned long size,
-                          struct path_prefix *prefix,
-                          char **match, int matches)
+static const char *entry_hex(struct tree_entry_list *e)
 {
-       struct path_prefix this_prefix;
-       this_prefix.prev = prefix;
-
-       if (strcmp(type, "tree"))
-               die("expected a 'tree' node");
-
-       if (matches)
-               recursive = 1;
-
-       while (size) {
-               int namelen = strlen(buffer)+1;
-               void *eltbuf = NULL;
-               char elttype[20];
-               unsigned long eltsize;
-               unsigned char *sha1 = buffer + namelen;
-               char *path = strchr(buffer, ' ') + 1;
-               unsigned int mode;
-               const char *matched = NULL;
-               int mtype = -1;
-               int mindex;
-
-               if (size < namelen + 20 || sscanf(buffer, "%o", &mode) != 1)
-                       die("corrupt 'tree' file");
-               buffer = sha1 + 20;
-               size -= namelen + 20;
-
-               this_prefix.name = path;
-               for ( mindex = 0; mindex < matches; mindex++) {
-                       mtype = pathcmp(match[mindex],&this_prefix);
-                       if (mtype >= 0) {
-                               matched = match[mindex];
-                               break;
-                       }
-               }
+       return sha1_to_hex(e->directory
+                          ? e->item.tree->object.sha1
+                          : e->item.blob->object.sha1);
+}
 
-               /*
-                * If we're not matching, or if this is an exact match,
-                * print out the info
-                */
-               if (!matches || (matched != NULL && mtype == 0)) {
-                       printf("%06o %s %s\t", mode,
-                              S_ISDIR(mode) ? "tree" : "blob",
-                              sha1_to_hex(sha1));
-                       print_path_prefix(&this_prefix);
-                       putchar(line_termination);
-               }
+/* forward declaration for mutually recursive routines */
+static int show_entry(struct tree_entry_list *, int);
 
-               if (! recursive || ! S_ISDIR(mode))
-                       continue;
+static int show_children(struct tree_entry_list *e, int level)
+{
+       if (prepare_children(e))
+               die("internal error: ls-tree show_children called with non tree");
+       e = e->item.tree->entries;
+       while (e) {
+               show_entry(e, level);
+               e = e->next;
+       }
+       return 0;
+}
 
-               if (matches && ! matched)
-                       continue;
+static int show_entry(struct tree_entry_list *e, int level)
+{
+       int err = 0; 
 
-               if (! (eltbuf = read_sha1_file(sha1, elttype, &eltsize)) ) {
-                       error("cannot read %s", sha1_to_hex(sha1));
-                       continue;
-               }
+       if (e != &root_entry) {
+               printf("%06o %s %s      ", e->mode, entry_type(e),
+                      entry_hex(e));
+               show_entry_name(e);
+               putchar(line_termination);
+       }
 
-               /* If this is an exact directory match, we may have
-                * directory files following this path. Match on them.
-                * Otherwise, we're at a pach subcomponent, and we need
-                * to try to match again.
+       if (e->directory) {
+               /* If this is a directory, we have the following cases:
+                * (1) This is the top-level request (explicit path from the
+                *     command line, or "root" if there is no command line).
+                *  a. Without any flag.  We show direct children.  We do not 
+                *     recurse into them.
+                *  b. With -r.  We do recurse into children.
+                *  c. With -d.  We do not recurse into children.
+                * (2) We came here because our caller is either (1-a) or
+                *     (1-b).
+                *  a. Without any flag.  We do not show our children (which
+                *     are grandchildren for the original request).
+                *  b. With -r.  We continue to recurse into our children.
+                *  c. With -d.  We should not have come here to begin with.
                 */
-               if (mtype == 0)
-                       mindex++;
-
-               list_recursive(eltbuf, elttype, eltsize, &this_prefix, &match[mindex], matches-mindex);
-               free(eltbuf);
+               if (level == 0 && !(ls_options & LS_TREE_ONLY))
+                       /* case (1)-a and (1)-b */
+                       err = err | show_children(e, level+1);
+               else if (level && ls_options & LS_RECURSIVE)
+                       /* case (2)-b */
+                       err = err | show_children(e, level+1);
        }
+       return err;
 }
 
-static int qcmp(const void *a, const void *b)
+static int list_one(const char *path, const char *path_end)
 {
-       return strcmp(*(char **)a, *(char **)b);
+       int err = 0;
+       struct tree_entry_list *e = find_entry(path, path_end);
+       if (!e) {
+               /* traditionally ls-tree does not complain about
+                * missing path.  We may change this later to match
+                * what "/bin/ls -a" does, which is to complain.
+                */
+               return err;
+       }
+       err = err | show_entry(e, 0);
+       return err;
 }
 
-static int list(unsigned char *sha1,char **path)
+static int list(char **path)
 {
-       void *buffer;
-       unsigned long size;
-       int npaths;
-
-       for (npaths = 0; path[npaths] != NULL; npaths++)
-               ;
-
-       qsort(path,npaths,sizeof(char *),qcmp);
-
-       buffer = read_object_with_reference(sha1, "tree", &size, NULL);
-       if (!buffer)
-               die("unable to read sha1 file");
-       list_recursive(buffer, "tree", size, NULL, path, npaths);
-       free(buffer);
-       return 0;
+       int i;
+       int err = 0;
+       for (i = 0; path[i]; i++) {
+               int len = strlen(path[i]);
+               while (0 <= len && path[i][len] == '/')
+                       len--;
+               err = err | list_one(path[i], path[i] + len);
+       }
+       return err;
 }
 
-static const char *ls_tree_usage = "git-ls-tree [-r] [-z] <key> [paths...]";
+static const char *ls_tree_usage =
+       "git-ls-tree [-d] [-r] [-z] <tree-ish> [path...]";
 
 int main(int argc, char **argv)
 {
+       static char *path0[] = { "", NULL };
+       char **path;
        unsigned char sha1[20];
 
        while (1 < argc && argv[1][0] == '-') {
@@ -194,7 +223,10 @@ int main(int argc, char **argv)
                        line_termination = 0;
                        break;
                case 'r':
-                       recursive = 1;
+                       ls_options |= LS_RECURSIVE;
+                       break;
+               case 'd':
+                       ls_options |= LS_TREE_ONLY;
                        break;
                default:
                        usage(ls_tree_usage);
@@ -206,7 +238,10 @@ int main(int argc, char **argv)
                usage(ls_tree_usage);
        if (get_sha1(argv[1], sha1) < 0)
                usage(ls_tree_usage);
-       if (list(sha1, &argv[2]) < 0)
+
+       path = (argc == 2) ? path0 : (argv + 2);
+       prepare_root(sha1);
+       if (list(path) < 0)
                die("list failed");
        return 0;
 }