numpy - Working with numpy in PrairieLearn

One of the challenges in working with numpy in PrairieLearn is serializing numpy data structures such as numpy.ndarray instances.

Serializing is necessary when passing data through the params["data"] data structure, this is stored in JSON format.

Passing a numpy.ndarray to question.html

If you want a numpy.ndarray to show up in question.html, if you try the ordinary thing:

data["params"]["values"] = values

You get this error:

TypeError: Object of type ndarray is not JSON serializable

The reason is that the numpy library doesn’t have code to serialize numpy.ndarray objects, and PrairieLearn uses JSON serialization to pass data between server.py and question.html.

So instead, you can convert to text, and then render the text variable, like this:

data["params"]["values_as_text"] = repr(list(map(lambda x: float(x), values)))

This:

• converts every np.float64 value to a regular native Python float value map(lambda x:float(x), values)
• converts the map to a list (list())
• converts that list to it’s text representation (repr())

You can then use `` to show the values in the array in your question.html

Passing a numpy.ndarray to files in tests

Here is an example of some code from a server.py file that generates a numpy.ndarray of numpy.float64 values, and then serializes them:

import numpy as np
import random
import prairielearn

def get_array_of_values():
    return np.random.rand(random.randint(5, 10)) * random.randint(1, 10)

def generate(data):
    values = get_array_of_values()    
    data["params"]["values"] = prairielearn.to_json(values)

We need values inside the file tests/setup_code.py. To get the values out, we can use this trick:

  values = data["params"]["values"]["_value"]

This gets you only part of the way there, however, because values is not necessarily at this point what we started with; that is, it is not a np.ndarray of np.float64 values. Instead, it is an ordinary python list.

So, instead, we may need something more like this:

  raw_values = data["params"]["values"]["_value"]
  values = np.array(raw_values, dtype=np.float64)