Targeted genetic test SQL generation for the DB2 database
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Automatic Query generators have been shown to be effective tools for software testing. For the most part, they have been used in system testing for the database as a whole or to generate specific queries to test specific features with not much randomness. In this work we explore the problems encountered when using a genetic algorithm to generate SQL for testing a large database system. General random SQL generation that tests the database system as a whole using genetic algorithms is relatively simple. One would need to generate millions of test cases to have a reasonable chance of hitting specific combinations of features. In order to optimize the testing, one needs to generate targeted SQL queries that narrow the testing to specific feature areas and feature combinations but yet preserve a certain amount of randomness and exploit the strength of a genetic algorithm. To do this effectively, the test generator needs to be guided so that it does not stray too much from the goals of the more targeted test requirement. In this work we explore a genetic algorithm approach to generate test queries that exercise target sub-sequences of features. Genetic algorithm parameters such as genome representation, reproduction, fitness evaluation, and selection are described. Preliminary results obtained comparing the presented approach with a random query generator are presented and discussed. We further present the DB2 SQL Query Optimizer, the application which we are using as a case study and target queries that go through certain optimization rule sequences. This application is larger and more complex in terms of code size and data input complexity then software previously used for studying test data generation.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it