CSC/ECE 517 Fall 2020 - SQLFE. Refactor Submission.java

Introduction

The SQLFE (SQL File Evaluation) open source software system automatically grades a set of files containing SQL queries. This project, written in Java, focuses on redesigning and rewriting one large method that parses one student submission file and builds a data structure for later evaluation.

About SQLFE

SQLFE is an open source software tool to assist computer science instructors by automatically grading assignments or tests involving multiple submitted files consisting of SQL queries added to a template submission file. Developed by Paul Wagner and others at the University of Wisconsin – Eau Claire, SQLFE allows instructors to specify a weighted set of tests for each assignment question, which are then used to evaluate a submitted query and/or compare that submitted query against an instructor-supplied solution query.

SQLFE also generates a detailed output file for each submission, which can be returned as feedback. SQLFE reduces instructor grading time for SQL queries and allows quicker, more detailed feedback as compared to hand grading. SQLFE source code, sample submission and instructor-generated files, and documentation can be found at https://github.com/wagnerpj42/SQL-File-Evaluation .

SQLFE is written entirely in Java. Since SQLFE executes SQL queries as part of its evaluation, it must be connected to an SQL database management system (DBMS). Currently Oracle and MySQL DBMSs are supported.

Problem Statement

Submission.java is the main file in SQLFE responsible for parsing student submissions. As the capabilities of SQLFE have increased, so has the complexity and responsibilities of the readSubmission method. At over 200 LoC, this method has become bloated and cumbersome. The challenge for this team is to refactor readSubmission and Submission.java to maintain proper functionality with proper code structure that will lend itself to future development.

1. Restructure and rewrite the readSubmission() method in the Submission.java class, adding any new sub-methods desired, to accurately and efficiently parse any student submission file that follows the assignment instructions.
   
   1. Each question number and submitted SQL query for that question should be placed in a new QuestionAnswer object in the class-level answers arraylist variable.
      
   2. Any user SQL comments (either -- single line SQL comments or /* */ multi-line SQL comments) should be written out to the AAA_student_comments.out file through the commWriter object. Each user SQL comment line may be written out separately, though if possible it would be fine to write out any user SQL comment as a whole, especially for multi-line /* */ comments.
      
   3. Any parsing problems that cannot be handled should be identified and written out to the AAA_parse_problems.out file through the parseWriter object.
      
2. You may continue to use, as is, any of the utility methods in the Utilities.java class. However, know that these methods are used by other classes in the project. If you prefer to create new utility methods (including making modifications to existing utility methods), any new or modified methods should be given new method names if they remain in the Utilities.java class file. In other words, do not modify or remove any method in Utiilties.java.
   
3. Generally, you should decide where to place sub-methods you create as part of the new readSubmission() method functionality (that is, you may place them in Submission.java or Utilities.java as you think is best for your design).
   
4. Good functionality is the primary goal. Any new code must work at least as well as the code that you’re replacing.
   
5. Good code commenting is expected. As a former consultant at 3M for a number of years, I’ve seen that the importance of good code commenting in helping me maintain or add to other people’s code cannot be overstated.
   
6. We are also interested in any constructive suggestions for improving the Submission class overall.
   

Team Members

1. Jack Maccdonald (jmmacdo4)
2. Nick Garner (nrgarner)
3. Sumitosh Pal (spal3)
4. Abhishek Gupta (agupta38)

Initial Approach

• Make use of the BufferedReader class to parse input files on a per-character basis. This will allow us to detect things like comment syntax, improper formatting of solutions, and many other criteria to control the parsing exactly as desired by the project mentor.
• Use boolean flags to track state. One of the challenges with any parsing method is the variety of input available that can create tricky edge cases. We intend to help control this with boolean flags to keep track of what sort of text we are currently parsing. For example, if we parse the opening to a comment, we will treat all subsequent text as a comment until we parse a comment close.
• Abstract commonly used code to helper methods in Utility.java. We will be looking closely for opportunities to abstract out commonly used to code to helper functions. These can be used in conjunction with the aforementioned boolean flags to help parse large chunks of input depending on surrounding syntax.
• Create unit and user tests for Submission.java. Currently, only the classes dealing with grading have attached unit tests. We will be creating unit tests as well as user tests through the UI to help ensure that our refactor of readSubmission is working properly. More details below in Test Plan.

Rationale

Most of the requirements from Dr. Wagner are related to making the readSubmission() method in Submission.java modular i.e refactor/split the function accordingly to meet standard function requirements. These can involve using helper functions from Utilities.java as well. Secondary requirements are handling several scenarios in the submission document that might appear due to students’ mistakes or other factors.

Implementation Strategy

Initially, we would fork the repository and create another branch where we will commit our changes. Dr Wagner has agreed to add us team members as collaborators for open source SQLFE, so we can create a branch and add changes to it. We also intend to add unit tests using JUnit4. After all the changes have been reviewed we would merge the changes back to main/master.

Changes introduced

TBD

Test Plan

Currently, our main focus with testing is the creation of unit and user tests for the Submission.java class. This will involve creating test submission files with edge case text in them to try and break the parser.

Examples include:

• Student answers inline with instructor comments
• Student comments written with instructor comment syntax
• Student answers with no terminating semicolon
• Student answers with inline comments
• Student comments with no termination
• Non-ASCII characters
• Submissions with question numbering different from the assignment

In addition, we will be devising a series of blackbox user tests to ensure that our updates to readSubmission do not break any functionality in the UI. This will involve users attempting use cases such as creating an assignment, grading a single submission, and grading a batch of submissions.

Beyond testing, we will be implementing some repository features to help improve the project’s accessibility for Open Source work. This includes setting up continuous integration features like TravisCI and Coveralls, to ensure that proposed changes do not break the build or significantly lower code coverage.

Results

Impact

Refactored readSubmission from XX LoC to YY LoC Abstracted commonly used behavior to maintain DRY principle Improved speed of submission parsing by ZZ%

Added Features

Continuous integration: TravisCI and Coveralls run unit tests on every commit to maintain code quality.

Proposed Future Scope

TBD