Guess that Taylor Swift line <3

1989.style is made with Torus and blocks.css. You can try it live at 1989.style.

Cornelia was a one-night project, so for sake of time I kept the software design pretty simple. There's a Go server which serves static files, in addition to a single dynamic JSON endpoint, GET /line, which returns a new random line of lyric along with which song it was pulled from.

The client-side single page app polls this endpoint for each question and renders it, and keeps track of the score per-device, completely locally, with synchronous localStorage. The quiz is a single Torus component.

The endpoint returns data of the form

{
    line: "I fell in love with a careless man's careful daughter",
    title: "Mine",
    choices: [
        "Should've Said No",
        "Superman",
        "You Are In Love",
    ],
}

i.e. it picks a song at random from the dataset, picks a line at random from that song, and also presents three other unique songs as alternate choices for the question. The server does not verify or keep track of answers.

The dataset of lyrics is imported into the repository as a fixture and manually vetted for reasonable playability. It's stored as a flat directory of text files in data/. Each file is a song with the song title as the file name, and each line in the text file is a discrete lyric line that should be presented to the user in the quiz. This makes it trivial to add and revise the lyrics as needed.

Rather than use a structured database, for simplicity, Cornelia simply imports the dataset from the data directory on startup and commits it to a data structure in memory, from which it polls for new questions. Since there are only 120 songs, this design is just about as efficient as you can get.

This project is licensed under the included MIT License, except the lyrics under data/, whose copyright belong to the rights holders of the records.