A simple solution to "Twitter interview problem" (http://qandwhat.apps.runkite.com/i-failed-a-twitter-interview/, https://gist.github.com/mkozakov/59af0fd5bddbed1a0399) using numpy.

gistfile1.py

import numpy as np
from matplotlib import pyplot as plt


def waterfill(walls):

    land = np.zeros((len(walls), max(walls)), dtype=int)

    # Put the walls
    for index, value in enumerate(walls):
        land[index, 0:value] = 2

    # Drop water
    land[land == 0] = 1

    # Remove flowed-out water
    for height in range(0, max(walls))[::-1]:

        left_max = land[:, height].argmax()
        right_max = len(walls) - land[:, height][::-1].argmax()

        land[:left_max, height] = 0
        land[right_max:, height] = 0

    return len(land[land == 1])


def waterfill_with_plots(walls):

    land = np.zeros((len(walls), max(walls)), dtype=int)
    fig = plt.figure()

    # Put the walls
    for index, value in enumerate(walls):
        land[index, 0:value] = 2

    dry_land = np.copy(land)
    dry_fig = fig.add_subplot(1, 2, 1)
    dry_fig.imshow(np.transpose(dry_land),
                   origin="lower",
                   interpolation="nearest")

    # Drop water
    land[land == 0] = 1

    # Remove flowed-out water
    for height in range(0, max(walls))[::-1]:

        left_max = land[:, height].argmax()
        right_max = len(walls) - land[:, height][::-1].argmax()

        land[:left_max, height] = 0
        land[right_max:, height] = 0

    wet_fig = fig.add_subplot(1, 2, 2)
    wet_fig.imshow(np.transpose(land),
                   origin="lower",
                   interpolation="nearest")

    fig.suptitle("{} filled blocks".format(len(land[land == 1])))
    fig.show()

The graph is wrong for [7, 1, 1, 1, 7, 1, 1, 1, 1, 7], but the solution is wright 42

Well, the graph looks good to me...