dcjohnston / deadleaves

Emojis and deadleaves.

deadleaves

Emojis and deadleaves.

1. Clone this!
2. cd deadleaves && npm install && bower install && pip install -r requirements.txt
3. grunt
4. Navigate here (localhost:3000)

deadleaves's Issues

Allow the user to create GPG/SSH keys based off randomness of the returned image.

Why not.

Infrastructure

• Set up hourly cron job to remove stale svgs that were left over due to failure.
• Set up provisioning in code (probably using puppet-agent single node conf)
• Set up reverse proxy w/ apache to serve static assets (also use minified deps).
• Set up express.js cluster for multi-thread processing on a single AWS box.

UI Improvements

• Add a timed update to rasterinzing messages (This may take up to 1 minute.)
• Add color picker for background
• Add social media integration (last, after infrastructure is ready)
• Add common screen sizes as options.
• Also make sure form inputs are full width on mobile. Is not currently stacking.
• Disable submit button (fix hiding messages bug) [BUG]

Issue #1: Implement a power-law random number generator

yo @dcjohnston:

The scales of the emojis will be drawn from a power-law distribution. Because JS, AFAIK, does not provide support for sampling from this distribution 😢, we need to implement it ourselves.

We can do this using the inverse of the cumulative density function (CDF) to map from probabilities to values. The cumulative density function at x records the probability that a random variable will take on a value less than or equal to x. For well-behaved distributions, this function is monotone increasing and bounded by 0 and 1, so it produces an invertible mapping between the set of possible values and the unit interval [0,1].

Therefore, we simply need to generate random values between 0 and 1 using Math.Random, then map them using the inverse CDF to generate our scales. 👍

The following python code defines a (higher-order) function that returns inverse CDFs for power-laws with different exponents alpha. Roughly speaking, a higher power means that smaller values are produced. We also need to define a minimum value for the scale, here xmin, so that we don't find ourselves in the awkward position of rendering emoji of zero size! 🙅

def getInversePowerCDF(alpha,xmin):
    def inversePowerCDF(p):
        return alpha*xmin*p**(-1/(alpha+1))
    return inversePowerCDF

Once we've defined the function inversePowerCDF based on a user-provided alpha, we need to produce a list of scales:

scales = [] #initialize
for eachEmoji in range(numEmojis):
    p = Math.Random()
    scales.append(inversePowerCDF(p))

This might also be a convenient time to produce a list of x and y locations - tuples in [0,1] x [0,1] - using Math.Random as well. If there's an efficient vectorization procedure for RNGs in JS, we can cheaply produce the 3 by numEmojis array of random numbers all at once! 💯 👌 👍 mmhmm

Gallery

Add a gallery page that stores (statically) 8 or so of our favorite emoji piles.