Can batch be used to implement transactions?
#29
Comments
hannesm
added a commit
to hannesm/opam-repository
that referenced
this issue
Dec 12, 2022
CHANGES: * Use ptime directly for RO.last_modified, instead of the int * int64 pair (mirage/mirage-kv#34) * Add RW.allocate to allocate a key and fill it with zero bytes (mirage/mirage-kv#34) * RO.list: return Key.t instead of string (mirage/mirage-kv#37, fixes mirage/mirage-kv#33) * Introduce a custom error for RW.rename with a source which is a prefix of destination (mirage/mirage-kv#37, fixes mirage/mirage-kv#31) * Use Optint.Int63.t for RO.size, RO.get_partial, RO.set_partial (mirage/mirage-kv#37, fixes mirage/mirage-kv#32) * Remove RW.batch (mirage/mirage-kv#37, fixes mirage/mirage-kv#29 mirage/mirage-kv#36, discussed at the MirageOS meeting in November, and on the mirageos-devel mailing list in January 2022) * Key.pp: escape the entire string, not individual fragments (mirage/mirage-kv#35)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Can I use the
batchoperation in application code for doing a series of sets and deletes (potentially leading to an error), and the result will be that either all sets and deletes are done, and the result is positive, or the result is an error and nothing has been done.I suspect the answer is no, should we have two operations,
start_transactionandcommit_transaction, where the first would produce a token (to be used as a KV.t), and the second apply a transaction? We can as well have acancel_transaction... Somehow that'd be a desired semantics for lots of applications. Certainly that'd be more difficult to implement in some backends, so maybe that could also be a separate interface (KV_RW_atomic)? //cc @samohtThe text was updated successfully, but these errors were encountered: