Like most relational databases, RethinkDB has several different commands to delete records from a database.<\/p>\n
To test the behavior, I made several copies of an existing table I had, which took ~1 minute per:<\/p>\n
\nr.db('performance')\n .tableCreate('query_timings2')\n\nr.db('performance')\n .table('query_timings2')\n .insert(r.db('performance')\n .table('query_timings'))\n<\/pre>\nCurrently, the SQL to ReQL comparision guide lists this as the equivalent of “truncate”:<\/p>\n
\nr.db('performance')\n .table('query_timings')\n .delete()\n<\/pre>\nIn my database, this took about the same amount of time as creating the original table. I believe this runs in a transaction, as when I abort the query the number of rows remains the same.<\/p>\n
In many relational databases, “truncate” has a specific meaning; in Oracle it deletes the underlying filesystem from disk immediately and outside a transaction. In Postgres it runs within a transaction but skips any effort required to verify rows, which appears to match what an unqualified delete does in RethinkDB.<\/p>\n
If you would like to replicate the Oracle behavior, dropping and re-creating a table seems to be much faster (I only know this experimentally, so I don’t know if this runs outside a transaction)<\/p>\n
\nr.db('performance')\n .tableDrop('query_timings') -- 45 ms\n\nr.db('performance')\n .tableCreate('query_timings')\n<\/pre>\nRethinkDB has two other neat feature that I’ve not seen in other databases; the option to do a “hard” vs “soft” delete, and whether to return the deleted rows:<\/p>\n
\nr.db('performance')\n .table('query_timings2')\n .delete({durability=\"soft\", returnChanges=false})\n<\/pre>\nIn my test, the “soft” delete took about half the time of the “hard” delete (basically it is not waiting for the OS to confirm changes were written to disk).<\/p>\n","protected":false},"excerpt":{"rendered":"
Like most relational databases, RethinkDB has several different commands to delete records from a database. To test the behavior, I made several copies of an existing table I had, which took ~1 minute per: r.db(‘performance’) .tableCreate(‘query_timings2’) r.db(‘performance’) .table(‘query_timings2’) .insert(r.db(‘performance’) .table(‘query_timings’)) Currently, the SQL to ReQL comparision guide lists this as the equivalent of “truncate”: r.db(‘performance’) … <\/p>\n