CREATE TABLE /*$wgDBprefix*/blob_tracking (
-- page.page_id
-- This may be zero for orphan or deleted text
+ -- Note that this is for compression grouping only -- it doesn't need to be
+ -- accurate at the time recompressTracked is run. Operations such as a
+ -- delete/undelete cycle may make it inaccurate.
bt_page integer not null,
-- revision.rev_id
-- This may be zero for orphan or deleted text
+ -- Like bt_page, it does not need to be accurate when recompressTracked is run.
bt_rev_id integer not null,
-- text.old_id
-- The CGZ content hash, or null
bt_cgz_hash varbinary(255),
- PRIMARY KEY (bt_rev_id, bt_text_id),
+ -- The URL this blob is to be moved to
+ bt_new_url varbinary(255),
- -- Sort by page for easy CGZ recompression
- KEY (bt_page, bt_rev_id),
+ -- True if the text table has been updated to point to bt_new_url
+ bt_moved bool not null default 0,
+
+ -- Primary key
+ -- Note that text_id is not unique due to null edits (protection, move)
+ -- moveTextRow(), commit(), trackOrphanText()
+ PRIMARY KEY (bt_text_id, bt_rev_id),
- -- For fast orphan searches
- KEY (bt_text_id),
+ -- Sort by page for easy CGZ recompression
+ -- doAllPages(), doAllOrphans(), doPage(), finishIncompleteMoves()
+ KEY (bt_moved, bt_page, bt_text_id),
-- Key for determining the revisions using a given blob
+ -- Not used by any scripts yet
KEY (bt_cluster, bt_blob_id, bt_cgz_hash)
+
+) /*$wgDBTableOptions*/;
+
+-- Tracking table for blob rows that aren't tracked by the text table
+CREATE TABLE /*$wgDBprefix*/blob_orphans (
+ bo_cluster varbinary(255),
+ bo_blob_id integer not null,
+
+ PRIMARY KEY (bo_cluster, bo_blob_id)
) /*$wgDBTableOptions*/;