Subversion Filesystem Schema v2.0 Proposal

Quickie Thoughts on Pros/Cons

Pros:

• Storage system independence; no marshaling
• Unique identifier for node revisions that isn't burdened by the need to carry extra data
• One-hop access to the node revision ID that birthed a line of history
• Access to all successors of a node revision ID
• Rename detection via dirent list/node revision ID comparison
• Cheaper DAG crawling using a single, simple table

Cons:

• Some information previously harvested from node revision IDs now requires actual reads of the node revisions to get at

Entities

Revision
REVNUM	Revision number (primary key)
ROOT-ID	Root associated with this revision

TransactionRoot
ROOT-ID	Primary key
ALIVE-P	True iff the transaction is not dead
BASE-ROOT-NODE-ID	Node ID of the root node in the tree on which this transaction is based

Root
ROOT-ID	Primary key
REVNUM	Revision number of committed root, or empty if root is not committed
ROOT-NODE-ID	Node ID of the root node in the tree represented by this root
PROP-ID	Property ID used to find properties associated with this root

NodeRevision
NODE-ID	Primary key
KIND	Node kind (file or dir, e.g.)
CREATED-ROOT-ID	ID of root in which this node revision was created
CREATED-PATH	Canonicalized path at which this node revision was created
BIRTH-NODE-ID	Node revision ID of the first node revision in this node revision's line of history (useful for answering is_related() questions)
PRED-NODE-ID	Node revision ID of this node revision's predecessor
PRED-COUNT	Number of predecessors this node revision has
COPY-ID	ID of the copy which caused this node revision to exist, if any
PROP-ID	Property ID used to find properties associated with this node revision
DATA-ID	Data ID used to find textual content associated with this node revision, if any
EDIT-DATA-ID	Data ID used to find not-yet-finalized textual content associated with this node revision, if any [feels too back-end specific]

Copy
COPY-ID	Primary key
SRC-PATH	Canonicalized path which was the source of this copy
SRC-ROOT-ID	Root ID in which the copy source was found
DST-NODE-ID	Node revision ID of the copy destination

NodeSuccessor
NODE-ID	Primary key (MULTI)
SUCC-NODE-ID	Node revision ID of a successor of the node revision

DirectoryEntry
NODE-ID	Primary key (MULTI)
CHILD-NAME	Name of a child of the directory node revision
CHILD-NODE-ID	Node revision ID of a child of the directory node revision (the child whose name is CHILD-NAME)

RootCopy
ROOT-ID	Primary key (MULTI)
COPY-ID	ID of a copy associated with this root