形 kata-dev

This small repo provides information for developers of kata challenges.

The kata server itself is private, because it essentially contains the solutions to the challenges. But the demo folder contains everything you need to provide for a challenge. This demo challenge is itself live at kata.scienxlab.org — check it out. Have a go at solving it, and you'll soon get the hang of what the challenges are like (although most of them are a bit harder than the demo!).

Basic information

• Users are anonymous and the server does not track or log any information at all.
• There is no UI, and the server only responds to GET requests. Forming and submitting requests is part of the learning for new Pythonistas.
• Users can respond as many times as they like, but wrong answers are throttled (one minute for the first, then 2 minutes, then 3, etc). The time penalty pertains to a key, not a user, and resets each day.
• Users can get hints by giving the server a key and question number, but no answer.

Guidelines for making challenges

• Challenges should be related to subsurface datasets or workflows, but can be quite abstract or toy-like.
• Your code must generate pseudorandom datasets of a reasonable size (usually 1000's of records), given a random seed.
• The generation should take less than about 250 ms; there is no caching on the server. (If you'd like to help implement this, please get in touch!).
• Challenges must have 3 or 4 questions.
• The responses to questions should be integers or strings. Floating point numbers are not ideal and would certainly need rounding.
• The first question should be very easy and really only check that the person has parsed the data correctly.
• The last question should be somewhat related to a real-world problem or challenge.
• Questions between the first and last should ideally build in some way towards the last.
• In general, the puzzles can be solved in any language, so we try to avoid Python-specific language or tools. But most people do them in Python so sometimes it seems humane to do it anyway.

Check out the two files in the demo folder and see if you can cook something up. The best way to share your challenge with us is probably by email at hello@scienxlab.org or via DM in the Software Underground slack (sign up at softwareunderground.org. Remember not to share it publicly; it needs to stay secret.