[PATCH v6 01/12] objects.c: introduce `exclude_level` enum
- From
Teng Long <dyroneteng@gmail.com>
- Date
- Oct 19, 2021, 11:38 UTC
- Message-ID
- <3a885678c9cee6ed4a70a7c88c15a7fa1b332150.1634634814.git.tenglong@alibaba-inc.com>
- In-Reply-To
- <cover.1634634814.git.tenglong@alibaba-inc.com>
From: Teng Long <dyroneteng@gmail.com>
Currently packfile-uri supports the exclusion of blob objects, but in some scenarios, users may wish to exclude more types of objects, such as commits, trees and tags, the difference with blob is they are more complicated. In addition to the meaning of a single object itself, it may also represent a collection of the objects that it includes (trees and blob in a root-tree) or reaches (ancestors of a commit). The exclusion range is designed by an enum named `exclude_level` in "pack-objects.c" that enumerate three values, "ET_SELF", "ET_INCLUDE" and "ET_REACHABLE".
Here are some explanations for their differences:
- Scene 1: "ET_SELF" for excluding object itself.
Reason to support the "ET_SELF" is because the definition of exclusion-level should be better to be consistent, no matter what the object type is. Excluding a single object itself has meaning for blobs, because blob is the smallest granularity among object types, and indeed, sometimes there are some frequently used big-size blobs in repository.
If you want to exclude more ownership or reachable objects, it can be considered to use "ET_INCLUDE" or "ET_REACHABLE".
- Scene 2: "ET_INCLUDE" for excluding object itself and objects it contains.
When a commit is specified to be excluded as packfile-uri, more offen, it's hoped to exclude the trees and blobs contained in its top-level tree, as well as the commit itself, but not the ancestors of the commit. This applies to scenarios where we want to exclude a specified non-blob object that includes some big-size objects.
Commit, tag and tree are suitable for this scenario. When a tag is specified, it will exclude the dereference commit, and all trees and blobs contained in its top-tree, as well as the tag itself if it's not a lightweight one.
- Scene 3: "ET_REACHABLE" for excluding object itself, all the objects it contains, and its ancestors.
For further exclusion range, the ancestors are needed to excluded together, for example, the clone scenes.
This commit only defines the `exclude_level` enum type. The implementations of non-blob object types will be added in subsequent commits.
Signed-off-by: Teng Long <dyroneteng@gmail.com> --- builtin/pack-objects.c | 44 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 44 insertions(+)
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c index 6d13cd3e1a..73b92a0c90 100644 --- a/builtin/pack-objects.c +++ b/builtin/pack-objects.c @@ -127,9 +127,52 @@ enum missing_action { }; static enum missing_action arg_missing_action; static show_object_fn fn_show_object; +/* + * exclude_level defines how to make the exclusion when the object matches + * a packfile-uri (uploadpack.excludeobject or uploadpack.blobpackfileuri) + * configuration. + * + * - ET_SELF; + * This type means only the object itself will be excluded, and all other + * objects it includes or reachable will not. For example, if object type is: + * - BLOB: The blob object will be excluded + * - TREE: The tree object will be excluded, the sub-trees and blobs it + * includes will not be excluded. + * - COMMIT: The commit object will be excluded, all the trees and blobs + * that be included in its top-level tree will not be excluded. + * - TAG: TAG object will be excluded, the referrenced commit object will + * not be excluded. + * - ET_INCLUDE; + * This type means that not only the object itself will be excluded, but + * also the objects it includes. For example, if object type is: + * - BLOB: Same with 'ET_SELF' + * - TREE: The tree object, and also the sub-trees and blobs that + * the object includes will be excluded. + * - COMMIT: The commit object, and also all the trees and blobs + * contained in its top-level tree will be excluded. + * - TAG: The TAG object will be excluded, and also the referrenced + * commit will be excluded (the referrenced commit exclusion will + * treat as a 'ET_INCLUDE' way). + * - ET_REACHABLE; + * This type means that not only the object and all the objects it includes + * will be excluded, but also the reachable objects. For exmple, if object + * type is: + * - BLOB: Same with 'ET_INCLUDE' + * - TREE: Same with 'ET_INCLUDE' + * - COMMIT: The Objects in the case of 'ET_INCLUDE' will be excluded, + * and also the ancestors of the commit will be excluded. + * - TAG: The Objects in the case of 'ET_INCLUDE' will be excluded, and + * also the ancestors of the referrenced commit will be excluded. + */ +enum exclude_level { + ET_SELF, + ET_INCLUDE, + ET_REACHABLE, +}; struct configured_exclusion { struct oidmap_entry e; + int level; char *pack_hash_hex; char *uri; }; @@ -3003,6 +3046,7 @@ static int git_pack_config(const char *k, const char *v, void *cb) if (oidmap_get(&configured_exclusions, &ex->e.oid)) die(_("object already configured in another " "uploadpack.blobpackfileuri (got '%s')"), v); + ex->level = ET_SELF; ex->pack_hash_hex = xcalloc(1, pack_end - oid_end); memcpy(ex->pack_hash_hex, oid_end + 1, pack_end - oid_end - 1); ex->uri = xstrdup(pack_end + 1);
-- 2.31.1.453.g945ddc3a74.dirty