Skip to content

Latest commit

 

History

History
87 lines (63 loc) · 2.43 KB

README.md

File metadata and controls

87 lines (63 loc) · 2.43 KB
Raw

litmus

A system for distinguishing candidate SQL queries using system-suggested tuples.

System Setup

Create a virtual environment and activate it:

cd python
virtualenv venv
source venv/bin/activate
pip install -r requirements.txt

Also:

cd python
cp config.ini.example config.ini

Then make sure the variables set in config.ini are what you want (in particular, the MySQL credentials must be set correctly!).

Datasets

Download the following datasets and load into MySQL:

Load Datasets

For datasets you want to use, run the following:

cd python
python setup_db.py <db_name>
python build_aig.py <db_name>
python tqc.py <db_name>

Run Experiments

For each experiment, run:

cd python
python main.py <arguments>

Example: python main.py mondial greedyall --tq_rank=equal

Required Arguments:

  • db: database name
  • mode: type of algorithm to use, select from topw, greedyall, greedybb, greedyfirst, l1s

Optional Arguments:

  • tq_rank: target query rank using weighting scheme described in the paper; select from equal, 1, q1 (n/4), half (n/2), q3 (3n/4), n. Default is equal.
  • qid: if you only want to run one task in the dataset, specify the ID of the task to run
  • info: either type or range. Just use range which uses both information sources (data types and intersecting values) described in the paper.
  • email: specify an email address to have the system email you when it's done with the task or if there's an error

Logging

Logs for tasks are generated in python/log/<db>_<mode>_tq<tq_rank>.

Results

After all tasks are completed, results are pickled into results/<db>_<mode>_tq<tq_rank>.pkl.

Analyzing Results

Analyzing results can be done using:

cd python
python analysis_single.py <arguments>

Required Arguments:

  • db: database name
  • mode: type of algorithm used (see list above)
  • tq_rank: see list above

Optional Arguments:

  • qid_min: only evaluate tasks with ID greater than or equal to this number
  • qid_max: only evaluate tasks with ID less than or equal to this number
  • tqc_min: only evaluate tasks with TQC value greater than this value
  • tqc_max: only evaluate tasks with TQC value less than or equal to this value