wei2912 / bce-simulation Goto Github PK

A full code review is required.

Looking through the code reveals many problems with the code that have resulted in it being messy.

@D10 I will discuss this with you tomorrow in real life.

Attempt to optimize code.

Ensure that server/build.png is a PNG. Using https://www.npmjs.org/package/is-png.

The current floats we use may underflow; for mathematical purposes, we need to ensure float precision.

All functions that need to be documented should have been documented by me; the only thing left is to add some simple examples (which doctest will test) and generate documentation using pydoc.

This is a serious error. Our only choice is to remove SciPy or use another free hosting system. See https://travis-ci.org/wei2912/bce-simulation/builds/44538826.

Currently all the *_{cli,graph}.py files are bloated with the get_args function. I suggest you create a new file containing all the argparse stuff and just import what you need. This would also make it a lot easier to solve the issue with the Heroku builds.

To support a very large number of trials at a fast speed, optimizations need to be made.

Possible ways:

• http://technicaldiscovery.blogspot.com/2011/06/speeding-up-python-numpy-cython-and.html
• Weave to run C code in Python
• Using https://github.com/pydata/numexpr

Research on the probability will need to finish before we can proceed.

I' m doing some epic stuff with the web UI, and I don't want to keep PRing 😉 Add me as collab, also create a Heroku app and enable Travis builds with a .travis.yml, and auto deploys to the heroku app, using the deploy configuration. I will push it over next week.

CoinPhysicsSim has a couple of functions that need testing to ensure that they are working properly.

The following error occured:

Traceback (most recent call last):
  File "utils/sims.py", line 10, in <module>
    from pyhull.convex_hull import ConvexHull
ImportError: No module named pyhull.convex_hull

Look at https://travis-ci.org/wei2912/bce-simulation/builds/27624403 for more details.

We need rate limiting, otherwise the server can be overloaded with resources and we'd have to pay for it or get it shut down...

We can try this: https://github.com/jhurliman/node-rate-limiter

kthxbye

The Quickhull algorithm for the Convex Hull problem needs to be implemented to support bigfloats.

References:

• QuickHull@Wikipedia
• http://mindthenerd.blogspot.com/2012/05/fastest-convex-hull-algorithm-ever.html (but we don't really need to prune for 5 points)

I'm going to guess that the server marks this as HTML instead of text.

Auto deploy to Heroku after Travis CI builds. Although this is enabled, it doesn't work as we don't have a buildpack for Python, matplotlib, and Node.js yet.

Frankly we don't really need the current coin simulation, since it won't answer any of our research questions.

Create a test suite!

read title

I think this is possible.

https://github.com/heroku/heroku-buildpack-python

Number of trials should be limited to ensure that users don't overload the server.

We need to document the commands.

In the following code:

def predict_prob(self):
    """
    For the variables passed into the simulation,
    predict the probability that the needle will hit
    at least one of the two parallel lines.
    """

    # area of coin / area of rectangle
    area_coin = math.pi * self.radius**2
    area = self.gap_x * self.gap_y
    if area_coin > area:
        return 1.0
    return area_coin / area

There is a small but noticeable difference in the probability that the chi-square test did not catch (as it does not have the following case scenario):

This happens when the width of the square is smaller than the coin itself and the four quadrants intersect together.

Probabilities must not go over 1.0; tests need to be established to ensure that they do not.

In the tests I see stuff like _pass_chi_test(ARGUMENTS) < MAX_STAT. Doesn’t _pass_chi_test account for that?

Test suite does not test the functions properly. The chi-square test should be applied to every case when calculating the probabilities, as only then will passing the chi-square test indicate that the probability calculation is correct.