Breaking VPL SQL Assignment

Post created: Tue, 18 Mar 2025 21:10:00 +0800

A quick demo photo of what is the main focus of this post

How do I find it?

Background

We got a question:

Q6. [12%] Display station_ID, station_name, and rental_income for each of the charging stations in February 2024.

  • Sort the results by rental_income in descending order, then station_ID in ascending order.
  • rental_income is the sum of the payment_amount in a given period.

Normally, we’d answer this question using a (My)SQL statement:

SELECT cs.station_ID, cs.name AS station_name,
    SUM(rt.payment_amount) AS rental_income
FROM ChargingStation cs
    INNER JOIN RentalTransaction rt ON cs.station_ID = rt.station_ID
WHERE rt.end_datetime >= '2024-02-01 00:00:00' AND
    rt.end_datetime < '2024-03-01 00:00:00' AND
    rt.status = 1 -- Completed transaction
GROUP BY cs.station_ID, cs.name
ORDER BY rental_income DESC, cs.station_ID ASC;

But someone on Moodle pointed out that the database is actually SQLite (In case if you’re wondering, he is skylee). That sparked my curiosity and got me exploring ways to “complete” this question through alternate means.

The output of `.version`: which shows the SQLite database version installed on the VPL jail system.

Information Gathering

By Googling, I learned we can execute shell commands inside an SQLite query. First, I listed the files in the current directory:

.shell ls

Result of running `.shell ls` in SQLite

Obviously, we need to inspect each file to see what we’ve got:

.shell cat *

Result of running `.shell cat *` which prints the contents of all files in the current directory. On the screen, there’s one shell script used by VPL for judging purposes.

Hmm. Most of the files aren’t related to the assignment or the test cases themselves. Let’s check the *.txt and *.sql files since I strongly suspect they’re tied to this particular question:

.shell cat *.txt *.sql

Result of running `.shell cat *.txt *.sql` which attempts to print out all files related to the above-mentioned SQL question.

We can see that the text file is likely what we want—the desired output. So, simply printing it should be enough to score 100%.

Note that this probably won’t work against hidden test cases, where they might use a different text file name to differentiate between tests. But that’s it.

Fluff

If you want to make VSCode (and GitHub Copilot) work like Cursor Tab, make sure to enable this experimental setting (switch to a pre-release version of the extension first!):

VSCode Copilot Tab setting location

I recommend trying out other options too! I can’t say for sure, but some might surprise you!

Also, I used copilot-more and got banned (lol). The repository should be deleted by GitHub by now. If you’re curious: it’s a tool that exposes GitHub Copilot’s (Large Language) Models as an OpenAI-compatible API. The reason you might want to use it (at first) is that if you have GitHub Education status as a student, it’s like getting free access to something equivalent to the latest Claude 3.7 Sonnet (thinking) API!

I don’t recommend doing this. Though, if you’re wondering what happened after I used it, it’s simple: my GitHub Copilot Pro got suspended. I reached out to GitHub Support, asked for reinstatement, promised not to abuse the system anymore, and they kindly re-enabled it. That’s it. (I hope this satisfies your curiosity and stops you from making the same mistake I did and getting banned too!)