Jux Video Dimensions

results.md

video_dimensions.R

# inspired by http://wiki.stdout.org/rcookbook/Graphs/Scatterplots%20(ggplot2)/#handling-overplotting
# to run:
#   $ R < video_dimensions.R --no-save

# install.packages("ggplot2")

dat <- read.csv("/home/jux/videos.csv")

library(ggplot2)

# Make each dot partially transparent, with 1/4 opacity
# For heavy overplotting, try using smaller values
ggplot(dat, aes(x=width, y=height, title="Dimensions of Videos on Jux")) +
    geom_point(shape=19,      # Use solid circles
               alpha=1/40)

# get a sample of the videos
samp <- dat[sample(nrow(dat), size=10000), ]

# Jitter the points
ggplot(samp, aes(x=width, y=height, title="Dimensions of Videos on Jux by Service")) +
    geom_point(shape=1,      # Use hollow circles
               position=position_jitter(width=20,height=10),
               aes(color=service))


# sort dimensions by count of occurance
dimCount <- sort(table(paste(dat$width, dat$height, sep="x")), decreasing=T)
# limit to top 20
dimCount <- head(dimCount, n=10)
barplot(dimCount, cex.name=0.6, main="Most Common Video Dimensions on Jux", xlab="Dimensions", ylab="Occurences")

video_dump.rb

require 'benchmark'
require 'csv'

CSV.open('/home/jux/videos.csv', 'wb') do |csv|
  csv << %w(id width height service)

  Mongoid.unit_of_work(disable: :current) do
    puts Benchmark.measure {
      Video.all.only(:property_set).each do |vid|
        ps = vid.property_set
        dims = ps.try :dimensions
        width = dims.try :width
        height = dims.try :height
        csv << [vid.id, width, height, ps.service] if width && height
      end
    }
  end
end

videos.csv

id,width,height,service
255835,459,344,youtube
345278,480,270,youtube
299833,480,270,youtube
257513,1280,720,vimeo
259259,640,480,vimeo
...