The information available from the Flickr API is incredibly rich. This Atlantic article This is the World on Flickr motivated me to open up R and do some analysis on the Flickr Explore list. As you might expect, I'll be using my new favorite tools rCharts and slidify.

ISO Speed Popularity

I have always wondered what ISO speeds occur most frequently on Explore. I never imagined that I could answer my question with R. As usual, we will start by loading all the necessary packages.

Load Our Packages

# analyze EXIF data for interesting list
library(httr)
#updated to use version 0.4 devtools::install_github("renkun-ken/pipeR/0.4")
library(pipeR)  
library(jsonlite)

If you do not have a free noncommercial API key, apply for one here. Trust me it is very easy, so don't let this be an excuse not to try it out. I put mine in a little secrets.Rdata file that I will load with following code and then start a session.

Authorize Our flickr

load("secrets.Rdata")

flickr.app <- oauth_app("r to flickr",api_key,secret)
flickr.endpoint <- oauth_endpoint(
  request = "https://www.flickr.com/services/oauth/request_token"
  , authorize = "https://www.flickr.com/services/oauth/authorize"
  , access = "https://www.flickr.com/services/oauth/access_token"
)

tok <- oauth1.0_token(
  flickr.endpoint
  , flickr.app
  , cache = F
)

Since this is more a proof of concept rather than an ambitious scientific study, I'll just look back three days.

#use this to specify how many days to analyze
daysAnalyze = 3

My code gets a little sloppy here but it does work. Originally I was forced to use a whole lot of lapply, but httr plays nicely with JSON, and jsonlite converts it to traditional R data stuctures. I hope my comments will help you understand each of the steps.

Get the Interesting

This will give us a list interesting with about 100 photos for each day. We could get up to 500 per day if we are ambitious with the per page API option.

#get a list and then make it a data frame 
interesting <- lapply(1:daysAnalyze, function(i){
    GET(url=sprintf(
      "https://api.flickr.com/services/rest/?method=flickr.interestingness.getList&api_key=%s&date=%s&format=json&nojsoncallback=1"
      , api_key
      , format( Sys.Date() - i, "%Y-%m-%d")
      , tok$credentials$oauth_token
      )
    ) %>>%
      content( as = "text" ) %>>%
      jsonlite::fromJSON () %>>%
      ( .$photos$photo ) %>>%
      ( data.frame(
        date = format( Sys.Date() - i, "%Y-%m-%d")
        ,.
        ,stringsAsFactors=F
      )) %>>%
      return
  }
) %>>%
  # combine all the days into a data frame
  ( do.call(rbind, .) )

Get EXIF for the Interesting

Now that we have some interesting photos, we can use flickr.photos.getExif for all sorts of meta information embedded in the EXIF.

  #for each photo try to get the exif information
  #Flickr allows users to block EXIF
  #so use try to bypass error
  #in case you are wondering, yes we could use dplyr here
  exifData <- lapply(
    1:nrow(interesting)
    ,function(photo){  # now we will process each photo
      exif <-
        GET(url=sprintf(
          "https://api.flickr.com/services/rest/?method=flickr.photos.getExif&api_key=%s&photo_id=%s&secret=%s&format=json&nojsoncallback=1"
          , api_key
          , interesting[photo,"id"]
          , interesting[photo,"secret"]
          )
        ) %>>%
          content( as = "text" ) %>>%
          jsonlite::fromJSON ()
    }
  )# %>>% could chain it, but want to keep exifData

  #now that we have a list of EXIF for each photo
  #use another lapply
  #to extract the useful information
  iso <- exifData %>>% 
    # some photos will not have exif if their owners disable it
    # and the api call will give us a stat "fail" instead of "ok"
    list.map(
      f(p,index) -> {
        ifelse (
          p$stat == "ok"
          , p$photo$exif %>>%
              (.[which(.[,"label"]=="ISO Speed"),"raw"])  %>>% 
              as.numeric 
          , NA
        ) %>>%
          {
            data.frame(
              interesting[ index, c( "date", "id" )]
              , "iso" = .
            )
          }
      }
    ) %>>%
    list.table( date, iso ) %>>%
    data.frame( stringsAsFactors = F)

Plot Our Results

Now that we have a data.frame with ISO speeds, let's use rCharts to analyze it. I will use dimplejs.

# Thanks to http://tradeblotter.wordpress.com/
# Qualitative color schemes by Paul Tol
tol4qualitative <- c("#4477AA", "#117733", "#DDCC77", "#CC6677")

require(rCharts)
iso %>>% ( dPlot(
  y = "Freq",
  x = "iso",
  groups = "date",
  data = .,
  type = "bar",
  height = 400,
  width =600
  ,xAxis = list( orderRule = sort( .$iso ) )
  ,defaultColors = tol4qualitative
) ) %>>% ( .$show("inline") )

#  using {} instead of () for our enclosure
#  might be more understandable
iso %>>% {
  dPlot(
    y = "Freq",
    x = c("iso","date"),
    groups = "date",
    data = .,
    type = "bar",
    height = 400,
    width =600
    ,xAxis = list( orderRule = sort( .$iso ) )
    ,defaultColors = tol4qualitative
  )
} %>>% ( .$show("inline") )

iso %>>% ( dPlot(
  y = "Freq",
  x = c("date","iso"),
  groups = "date",
  data = .,
  type = "bar",
  height = 400,
  width =600
  ,xAxis = list( grouporderRule = sort( .$iso ) )
  ,defaultColors = tol4qualitative 
) )  %>>% ( .$show("inline") )

iso %>>% ( dPlot(
  y = "Freq",
  x = "iso",
  groups = "date",
  data = .,
  type = "line",
  height = 400,
  width =600
  ,xAxis = list( orderRule = sort( .$iso ) )
  ,defaultColors = tol4qualitative   
) )  %>>% ( .$show("inline") )

iso %>>%  ( dPlot(
  y = "Freq",
  x = c("date","iso"),
  groups = "date",
  data = .,
  type = "area",
  height = 400,
  width =600
  ,xAxis = list( grouporderRule = sort( .$iso ) )
  ,defaultColors = tol4qualitative 
) )  %>>% ( .$show("inline") )

As you might already know, I love R, especially with rCharts and slidify.

Now I can add that I love httr and pipeR.

Thanks to all those who contributed knowingly and unknowingly to this post.