Skip to content

About

Practice turning a table into a chart with the datascience library. Python runs in your browser.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

2 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Chart Match

Try it: https://ncssm.github.io/chartmatch/

Practice for students learning the datascience Python library (the one used in Berkeley's Data 8 and in NCSSM's Foundations of Data Science). Each problem shows the first 10 rows of a table and a target chart. Students write the code that turns the table into that chart, run it, and see their chart next to the target with feedback.

Python runs entirely in the browser with Pyodide, so there is no server and nothing to install.

How it works

  • Problems come from recipes (for example "where on a category, then group, sort, take, barh") applied to 22 datasets. The random part is the details: which column, which filter value, which group function (np.mean, np.median, max, min, sum), how many rows to keep. Filters and top-N limits that the chart can't show are stated in a note above the chart. There are 21 recipes in three levels (easy, medium, hard).
  • Grading compares what the charts show, not the code text, so any code that draws the right chart is correct. chartmatch.py reads the data behind a matplotlib figure (bar labels and lengths, histogram bins, scatter points, line values, axis labels, legend) and compares it with the target. When they differ it explains how: wrong chart type, different columns, missing color groups, too many bars, bars out of order, and so on.
  • Hints go in two steps (the chart type, then the steps in order) before the answer.
  • Scores are kept in the student's own browser (localStorage); nothing is sent anywhere.

Layout

Path What it is
docs/ The site GitHub Pages serves.
docs/chartmatch.py Problem recipes, generator, chart comparison, and feedback. Runs in CPython and in Pyodide.
docs/worker.js Starts Pyodide in a Web Worker and passes messages between the page and chartmatch.py.
docs/data/ The practice datasets, trimmed to the columns problems use.
docs/datascience-0.18.1-py3-none-any.whl The datascience package, installed into Pyodide at startup.
tools/index.template.html The page. tools/build.py fills in a pre-rendered first problem and the build settings.
tools/build.py Builds docs/index.html (GitHub Pages) or dist-artifact/ (a self-contained copy).
tools/test_chartmatch.py Stress test for the generator and grader.
tools/prep_data.py Rebuilds docs/data/ from the course repo's dataset collection.

datascience imports scipy, IPython, folium, and branca when it loads. None of them are used by the chart methods practiced here, so in the browser chartmatch.py replaces them with small stand-in modules instead of downloading them.

Working on it

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python tools/test_chartmatch.py 300    # every solution grades correct; common mistakes grade wrong
python tools/build.py pages            # rebuild docs/index.html after changing the template
python -m http.server --directory docs # then open http://localhost:8000

To add a dataset: put a CSV in docs/data/, add it to DATA_FILES in docs/worker.js, and add an entry to DATASETS in docs/chartmatch.py saying which columns can play which roles. To add a kind of problem, write a recipe function and add it to RECIPES.

python tools/build.py artifact makes dist-artifact/, a copy with Pyodide and its packages bundled in instead of loaded from the jsDelivr CDN. That is how the Claude artifact version is published: its host only serves standard web file types, so the .whl and .zip files are renamed with an extra .wasm extension and worker.js adds it back when fetching them.

Data credits

The datasets were found through Jeremy Singer-Vine's Data Is Plural newsletter. Files are trimmed to a few columns, and rows missing a needed value are removed.

File Dataset Source Data Is Plural
penguins Palmer penguins data 2020-06-17
candy Halloween candy rankings (FiveThirtyEight) data 2021-10-20
pixar Pixar films data 2024-12-04
survivor Survivor seasons data 2021-05-05
volcanoes Volcanoes of the world (Smithsonian GVP) data 2018-06-06
ramen Instant ramen ratings (The Ramen Rater) data 2018-01-24
world_cup World Cup group standings data 2022-07-20
fight_songs College fight songs (FiveThirtyEight) data 2019-09-04
jeans Jeans pocket sizes (The Pudding) data 2018-08-22
peaks Himalayan peaks (Himalayan Database) data 2018-05-09
banknotes People on banknotes (The Pudding) data 2022-05-18
nobel Nobel Prize winners data 2017-02-08
chocolate Chocolate bar ratings (Flavors of Cacao) data 2022-01-12
exoplanets NASA Exoplanet Archive data 2016-05-11
quakes USGS earthquakes, magnitude 5.5+ in 2024 data 2024-04-17
warming US warming by state (The Washington Post) data 2020-08-12
groundhogs Groundhog Day predictions data 2023-02-08
tour Tour de France winners data 2020-03-11
bob_ross Bob Ross paintings data 2020-12-09
yearbook Big hair in yearbook photos (The Pudding) data 2019-12-11
baby Child Health and Development Studies births (via Data 8) Data 8
galton Galton's family heights (via Data 8) Data 8

Each dataset remains under its original source's terms.

About

Practice turning a table into a chart with the datascience library. Python runs in your browser.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages