amsa / doop Goto Github PK

Just talked with Te,

For benchmark, we think for benchmarking:

1. Various workload(different portion of insert and select)
2. Various branching situation(we can expect query on recursive branch is slower)
3. Benchmark on Space complexity.

When using those default branch names and tables names, we should avoid hard coding them..just used the constant we declared in core package. I fixed some in doop_db.go, but still there are some in the bench package.

parser (rewriter) can parse "SELECT * FROM t1" but not "SELECT * FROM t1". It needs to be fixed especially because MySQL by default adds these characters when storing the SQL schema.

UntrackDb should take an argument branch_name indicating:

• if it's empty(or it can be a special string or whatever), keep snapshot, clean all extraneous tables of all branches of this database
• if it's a correct branch name, keep that branch, clean all extraneous tables of other branches of this database.

Primary key of the relation should be detected to create v_sec as the following query is invalid (primary could be something other than id):

CREATE TABLE %s AS SELECT id FROM %s