YASQLT – Yet Another SQL Tutor. A Pragmatic Approach

1DSV SU

YASQLT – Yet Another SQL Tutor

Ilia Bider & David Rogers

4th Symposium on Conceptual Modeling Education, Gifu, Japan 2016 (part of ER 2016)

A Pragmatic Approach

DSV SU 2

A Goal with Presentation

Present and discuss motivation, rational, ideas and result from a pragmatic project of creating an automated assessment tool for an introductory database course

DSV SU 3

MotivationContext: • a short introductory database course with 2-3 rounds a year with around

200 students: 5 weeks, 7,5 ECT points• SQL assignment done in groups 4-6 persons• Too much work spent on assessing assignments• The feedback to the students comes too late, after the written exam

Needed• A tool for automated assessment so that the students get early feedback

and submitted somewhat tested SQL

DSV SU 4

Details: Database Scheme

Example of Query:Get the national Id, name and city of students who have participated in at least three different courses. (group by or self-join)

DSV SU 5

Details: Typical Errors

List national ID numbers and names of students who have participated in a course with code Java1SELECT DISTINCT S.nationalidnumber, S.name FROM Student AS S, Participation AS P WHERE P.course = ‘JAVA2’ (no connection between the tables)

List national ID numbers of students who have participated in a course in room JupiterSELECT DISTINCT student FROM Participation P, Courseround C, Room R WHERE P.course = C.course AND C.room = R.id AND R.name='Jupiter'. (segmented key error)

List national ID numbers of students who have participated in a course with code Java1SELECT student FROM Participation WHERE course= 'Java1‘ (missing DISTINCT)

DSV SU 6

Why to do on our ownTried to find something in the outer world:

Message to AISworld list – no result

Contacting some authors of published work – not available

After several tries decided to do on our own

“I wonder whether anybody has a tool in operation for automated assessment of assignments in a database course. In particular, we would be interested to have such a tool for assessing assignments related to SQL, though other tools like converting UML into Relational scheme, or assessing assignments in the area of Relational Algebra would be also good to have. In connection to SQL, assessing assignments connected to learning SELECT has a priority. The tool does not need to make a sophisticated analysis of SQL statements, and it does not need to insure 100% correctness - it can miss some incorrect statements, but shouldn't mark as incorrect correct statements. The following scheme would do. A database is created in some popular DBMS, like SQLserver, or MySQL, or even Access. A teacher creates a textual question, and a corresponding SQL SELECT. The statement is run and the set of records it gives is saved. The students are not allowed to change the content of the database. When a student creates his/her own SELECT to answer the question, it is run against the same database and the set produced is compared with the one stored when the teacher ran his/her version of SELECT. If the data sets differ, the tool marks the student's SELECT as incorrect. Any help in acquiring such a tool would be much appreciated. As customary, a summary of responses will be reported back to the list.”

DSV SU 7

Pragmatic principles• No syntax control (use standard DBMS)

• Only SELECT and CREATE VIEW (intro course)

• No semantic or optimization checks. Checks are done only (a) whether the result of the query is correct and on typical errors

• No generic approach to semantic analysis

• Agile approach to development

DSV SU 8

Presumptions about students SQL • Students are using “classical” SQL without inner joins, though

inner joins are not prohibited.

• Tables listed in the FROM clause are to be joined by conditions in the WHERE clause, no Cartesian products are expected.

• Tables are joined based on foreign key - primary key relationships, directly or indirectly (transitively). Self-join can be used based on one or several columns with the same names.

DSV SU 9

User interface

DSV SU 10

Technical implementationServer• Linux (Debian), Tomcat 8 webserver (from Apache) • UCanAcess library (MS Acess) (from Marco Amadei)• Jackcess (from Health Market Science), • HyperSQL (from The HSQL Development Group)• Commons Lang, Apache Commons Logging (from Apache)• JSqlPArser (from R. Manning)

Client• HTML, CSS, Javascript and jQuery (from jQuery Foundation)

DSV SU 11

Preliminary resultsQuality of submissions

Course round Passed on first submission Spring 2015 – 1st round. No YASQLT 17% (5 of 30 groups) Spring 2015 – 2nd round. No YASQLT 0% (0 of 21 groups) Autumn 2015. No YASQLT 0% (0 of 11 groups) Spring 2016 – 1st round. First edition of YASQLT 43% (13 of 30 groups) Spring 2016 – 2nd round. Second edition of YASQLT 48% (13 of 27 groups)

DSV SU 12

Preliminary resultsStudents’ opinions on Likert scale (based on 30 answers)

YASQLT 5 - agree 4 3 2 1 - disagree helped me to prepare for the exams 20.0% 26.7% 36.7% 0.0% 16.7% helped me to learn SQL 46.7% 26.7% 13.3% 6.7% 6.7% motivated me to test more SQL 33.3% 26.7% 13.3 20% 6.7%

DSV SU 13

Learning by failingPrinciple: Try - Fail - Learn (from failure)

Requires: Prompt feedback and explanation of failures

Context: Large classes and short calendar time

Requires: Automated tools to discover and explain errors

Focus: On typical, often encountered errors (rather than full analysis)

YASQLT – an example of such a tool

Other areas: Errors in conceptual modeling (e.g. multiplicity), Translation from conceptual model to relational database model

DSV SU 14

Q & A

Thank you for your patience

Questions and comments

Contact [email protected]