Merge branch 'jc/core-checkstat-2.0'
[gitweb.git] / builtin / cat-file.c
index 045cee7bce0cfc130d6964bab941a44558ec35a6..d5a93e0e911ccbe4560a03ed78436bfb174e0f68 100644 (file)
@@ -13,9 +13,6 @@
 #include "userdiff.h"
 #include "streaming.h"
 
-#define BATCH 1
-#define BATCH_CHECK 2
-
 static int cat_one_file(int opt, const char *exp_type, const char *obj_name)
 {
        unsigned char sha1[20];
@@ -48,6 +45,14 @@ static int cat_one_file(int opt, const char *exp_type, const char *obj_name)
        case 'e':
                return !has_sha1_file(sha1);
 
+       case 'c':
+               if (!obj_context.path[0])
+                       die("git cat-file --textconv %s: <object> must be <sha1:path>",
+                           obj_name);
+
+               if (textconv_object(obj_context.path, obj_context.mode, sha1, 1, &buf, &size))
+                       break;
+
        case 'p':
                type = sha1_object_info(sha1, NULL);
                if (type < 0)
@@ -70,16 +75,6 @@ static int cat_one_file(int opt, const char *exp_type, const char *obj_name)
                /* otherwise just spit out the data */
                break;
 
-       case 'c':
-               if (!obj_context.path[0])
-                       die("git cat-file --textconv %s: <object> must be <sha1:path>",
-                           obj_name);
-
-               if (!textconv_object(obj_context.path, obj_context.mode, sha1, 1, &buf, &size))
-                       die("git cat-file --textconv: unable to run textconv on %s",
-                           obj_name);
-               break;
-
        case 0:
                if (type_from_string(exp_type) == OBJ_BLOB) {
                        unsigned char blob_sha1[20];
@@ -117,54 +112,211 @@ static int cat_one_file(int opt, const char *exp_type, const char *obj_name)
        return 0;
 }
 
-static int batch_one_object(const char *obj_name, int print_contents)
-{
+struct expand_data {
        unsigned char sha1[20];
-       enum object_type type = 0;
+       enum object_type type;
        unsigned long size;
-       void *contents = NULL;
+       unsigned long disk_size;
+       const char *rest;
+       unsigned char delta_base_sha1[20];
+
+       /*
+        * If mark_query is true, we do not expand anything, but rather
+        * just mark the object_info with items we wish to query.
+        */
+       int mark_query;
+
+       /*
+        * Whether to split the input on whitespace before feeding it to
+        * get_sha1; this is decided during the mark_query phase based on
+        * whether we have a %(rest) token in our format.
+        */
+       int split_on_whitespace;
+
+       /*
+        * After a mark_query run, this object_info is set up to be
+        * passed to sha1_object_info_extended. It will point to the data
+        * elements above, so you can retrieve the response from there.
+        */
+       struct object_info info;
+};
+
+static int is_atom(const char *atom, const char *s, int slen)
+{
+       int alen = strlen(atom);
+       return alen == slen && !memcmp(atom, s, alen);
+}
+
+static void expand_atom(struct strbuf *sb, const char *atom, int len,
+                       void *vdata)
+{
+       struct expand_data *data = vdata;
+
+       if (is_atom("objectname", atom, len)) {
+               if (!data->mark_query)
+                       strbuf_addstr(sb, sha1_to_hex(data->sha1));
+       } else if (is_atom("objecttype", atom, len)) {
+               if (data->mark_query)
+                       data->info.typep = &data->type;
+               else
+                       strbuf_addstr(sb, typename(data->type));
+       } else if (is_atom("objectsize", atom, len)) {
+               if (data->mark_query)
+                       data->info.sizep = &data->size;
+               else
+                       strbuf_addf(sb, "%lu", data->size);
+       } else if (is_atom("objectsize:disk", atom, len)) {
+               if (data->mark_query)
+                       data->info.disk_sizep = &data->disk_size;
+               else
+                       strbuf_addf(sb, "%lu", data->disk_size);
+       } else if (is_atom("rest", atom, len)) {
+               if (data->mark_query)
+                       data->split_on_whitespace = 1;
+               else if (data->rest)
+                       strbuf_addstr(sb, data->rest);
+       } else if (is_atom("deltabase", atom, len)) {
+               if (data->mark_query)
+                       data->info.delta_base_sha1 = data->delta_base_sha1;
+               else
+                       strbuf_addstr(sb, sha1_to_hex(data->delta_base_sha1));
+       } else
+               die("unknown format element: %.*s", len, atom);
+}
+
+static size_t expand_format(struct strbuf *sb, const char *start, void *data)
+{
+       const char *end;
+
+       if (*start != '(')
+               return 0;
+       end = strchr(start + 1, ')');
+       if (!end)
+               die("format element '%s' does not end in ')'", start);
+
+       expand_atom(sb, start + 1, end - start - 1, data);
+
+       return end - start + 1;
+}
+
+static void print_object_or_die(int fd, struct expand_data *data)
+{
+       const unsigned char *sha1 = data->sha1;
+
+       assert(data->info.typep);
+
+       if (data->type == OBJ_BLOB) {
+               if (stream_blob_to_fd(fd, sha1, NULL, 0) < 0)
+                       die("unable to stream %s to stdout", sha1_to_hex(sha1));
+       }
+       else {
+               enum object_type type;
+               unsigned long size;
+               void *contents;
+
+               contents = read_sha1_file(sha1, &type, &size);
+               if (!contents)
+                       die("object %s disappeared", sha1_to_hex(sha1));
+               if (type != data->type)
+                       die("object %s changed type!?", sha1_to_hex(sha1));
+               if (data->info.sizep && size != data->size)
+                       die("object %s changed size!?", sha1_to_hex(sha1));
+
+               write_or_die(fd, contents, size);
+               free(contents);
+       }
+}
+
+struct batch_options {
+       int enabled;
+       int print_contents;
+       const char *format;
+};
+
+static int batch_one_object(const char *obj_name, struct batch_options *opt,
+                           struct expand_data *data)
+{
+       struct strbuf buf = STRBUF_INIT;
 
        if (!obj_name)
           return 1;
 
-       if (get_sha1(obj_name, sha1)) {
+       if (get_sha1(obj_name, data->sha1)) {
                printf("%s missing\n", obj_name);
                fflush(stdout);
                return 0;
        }
 
-       if (print_contents == BATCH)
-               contents = read_sha1_file(sha1, &type, &size);
-       else
-               type = sha1_object_info(sha1, &size);
-
-       if (type <= 0) {
+       if (sha1_object_info_extended(data->sha1, &data->info, LOOKUP_REPLACE_OBJECT) < 0) {
                printf("%s missing\n", obj_name);
                fflush(stdout);
-               if (print_contents == BATCH)
-                       free(contents);
                return 0;
        }
 
-       printf("%s %s %lu\n", sha1_to_hex(sha1), typename(type), size);
-       fflush(stdout);
+       strbuf_expand(&buf, opt->format, expand_format, data);
+       strbuf_addch(&buf, '\n');
+       write_or_die(1, buf.buf, buf.len);
+       strbuf_release(&buf);
 
-       if (print_contents == BATCH) {
-               write_or_die(1, contents, size);
-               printf("\n");
-               fflush(stdout);
-               free(contents);
+       if (opt->print_contents) {
+               print_object_or_die(1, data);
+               write_or_die(1, "\n", 1);
        }
-
        return 0;
 }
 
-static int batch_objects(int print_contents)
+static int batch_objects(struct batch_options *opt)
 {
        struct strbuf buf = STRBUF_INIT;
+       struct expand_data data;
+
+       if (!opt->format)
+               opt->format = "%(objectname) %(objecttype) %(objectsize)";
+
+       /*
+        * Expand once with our special mark_query flag, which will prime the
+        * object_info to be handed to sha1_object_info_extended for each
+        * object.
+        */
+       memset(&data, 0, sizeof(data));
+       data.mark_query = 1;
+       strbuf_expand(&buf, opt->format, expand_format, &data);
+       data.mark_query = 0;
+
+       /*
+        * If we are printing out the object, then always fill in the type,
+        * since we will want to decide whether or not to stream.
+        */
+       if (opt->print_contents)
+               data.info.typep = &data.type;
+
+       /*
+        * We are going to call get_sha1 on a potentially very large number of
+        * objects. In most large cases, these will be actual object sha1s. The
+        * cost to double-check that each one is not also a ref (just so we can
+        * warn) ends up dwarfing the actual cost of the object lookups
+        * themselves. We can work around it by just turning off the warning.
+        */
+       warn_on_object_refname_ambiguity = 0;
 
        while (strbuf_getline(&buf, stdin, '\n') != EOF) {
-               int error = batch_one_object(buf.buf, print_contents);
+               int error;
+
+               if (data.split_on_whitespace) {
+                       /*
+                        * Split at first whitespace, tying off the beginning
+                        * of the string and saving the remainder (or NULL) in
+                        * data.rest.
+                        */
+                       char *p = strpbrk(buf.buf, " \t");
+                       if (p) {
+                               while (*p && strchr(" \t", *p))
+                                       *p++ = '\0';
+                       }
+                       data.rest = p;
+               }
+
+               error = batch_one_object(buf.buf, opt, &data);
                if (error)
                        return error;
        }
@@ -186,10 +338,29 @@ static int git_cat_file_config(const char *var, const char *value, void *cb)
        return git_default_config(var, value, cb);
 }
 
+static int batch_option_callback(const struct option *opt,
+                                const char *arg,
+                                int unset)
+{
+       struct batch_options *bo = opt->value;
+
+       if (unset) {
+               memset(bo, 0, sizeof(*bo));
+               return 0;
+       }
+
+       bo->enabled = 1;
+       bo->print_contents = !strcmp(opt->long_name, "batch");
+       bo->format = arg;
+
+       return 0;
+}
+
 int cmd_cat_file(int argc, const char **argv, const char *prefix)
 {
-       int opt = 0, batch = 0;
+       int opt = 0;
        const char *exp_type = NULL, *obj_name = NULL;
+       struct batch_options batch = {0};
 
        const struct option options[] = {
                OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
@@ -200,12 +371,12 @@ int cmd_cat_file(int argc, const char **argv, const char *prefix)
                OPT_SET_INT('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
                OPT_SET_INT(0, "textconv", &opt,
                            N_("for blob objects, run textconv on object's content"), 'c'),
-               OPT_SET_INT(0, "batch", &batch,
-                           N_("show info and content of objects fed from the standard input"),
-                           BATCH),
-               OPT_SET_INT(0, "batch-check", &batch,
-                           N_("show info about objects fed from the standard input"),
-                           BATCH_CHECK),
+               { OPTION_CALLBACK, 0, "batch", &batch, "format",
+                       N_("show info and content of objects fed from the standard input"),
+                       PARSE_OPT_OPTARG, batch_option_callback },
+               { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
+                       N_("show info about objects fed from the standard input"),
+                       PARSE_OPT_OPTARG, batch_option_callback },
                OPT_END()
        };
 
@@ -222,19 +393,19 @@ int cmd_cat_file(int argc, const char **argv, const char *prefix)
                else
                        usage_with_options(cat_file_usage, options);
        }
-       if (!opt && !batch) {
+       if (!opt && !batch.enabled) {
                if (argc == 2) {
                        exp_type = argv[0];
                        obj_name = argv[1];
                } else
                        usage_with_options(cat_file_usage, options);
        }
-       if (batch && (opt || argc)) {
+       if (batch.enabled && (opt || argc)) {
                usage_with_options(cat_file_usage, options);
        }
 
-       if (batch)
-               return batch_objects(batch);
+       if (batch.enabled)
+               return batch_objects(&batch);
 
        return cat_one_file(opt, exp_type, obj_name);
 }