dkhalansky / stepik Goto Github PK

For step types Test, Reordering, and Multiple selection, tasks are generated automatically. Currently we provide no interface for receiving these generated task statements, only for submitting the solutions for them.

The lessons is much slower than they can be, because some of the requests to the Stepik server can be run in parallel, but now are executed linearly.

Due to it being written in a somewhat sloppy python, it crashes with a backtrace all the time in case there is even a small error. That is, we practically don't validate any user input, and if what the user wanted works, it works, otherwise it crashes with an unhelpful error message.

This problem is somewhat mitigated by the man page being quite short and supposedly readable. Therefore, it should be easy to learn what to do to achieve a desired result without running into a backtrace.

Due to how slow stepik's server is (on my machine it's 2.2 seconds on average to get a list of courses one is enrolled in), zsh completion doesn't feel fluent and is pretty uncomfortable to use. Maybe some strategic caching could fix this, but then there would be a whole new world of problems with cache invalidation.

Now one can only pipe text to the program, but it would be nice to send files along with it.

All step types that are listed as experimental or complex, as well as the "Table" step type, are completely unsupported. They don't seem to be common, so it's probably not a big deal, but the support is limited.

Due to how slow the Stepik's server is (2.3 seconds on average for performing a single query for receiving a list of courses), the time investment in creating completion functions doesn't look like it will pay off: bash completion isn't supposed to take 2.3 seconds, and would be nearly useless if we implemented it naively.