Btibert3
12/8/2015 - 2:48 PM

Testing stattleship, weird error

Testing stattleship, weird error

stattle = function(token, 
                    sport="hockey", 
                    league = "nhl", 
                    ep="stats", 
                    query=list(), 
                    version=1, 
                    walk=F,
                    page=NA,
                    verbose=F) {
  ## if na, set page to 1 for consistency
  if (is.na(page)) page = 1
  
  ## if page is supplied, add it to the list
  if (!is.na(page) & is.numeric(page) & page >= 1) {
    query = c(query, page=page)
  }
  
  ## get the first request
  tmp = queryAPI(TOKEN, sport, league, ep, query, verbose=T)
  
  ## simple alert
  if (tmp$response$status_code != 200) {
    message("API response was something other than 200")
  }
  
  ## create the response list
  response = list()
  
  ## set the original parsed response to the first element
  response[[1]] = tmp$api_json
  
  ## if walk, parse here and send into respose[[i]]
  ## NOT FINISHED -- below is under dev
  if (walk) {
    ## check to see if paging is necessary
    total_results = as.numeric(tmp$response$headers$total)
    rpp = as.numeric(tmp$response$headers$`per-page`)
    pages = ceiling(total_results / rpp)
    
    ## the first page was already retrievedd, only care 2+
    if (pages >= 2) {
      for (p in 2:pages) {
        p_query = list(page=p)
        tmp_p = queryAPI(TOKEN, sport, league, ep, query=p_query, verbose=T)
        
        ## check to make sure 200
        if (tmp$response$status_code != 200) {
          message("the pages>2 loop requested a page that was not 200")
        }
        
        ## add as an element into the response container
        response[[p]] = tmp_p$api_json
      }
      
    }#endif(pages)
    
  }#endif(walk)
  
  ## return the list of data results
  ## list of lists
  return(response)
  
}
#' Interface with the Stattleship API
#' 
#' A simple, generic function to query data from the API
#' 
#' @param token character. A valid token for the API
#' @param sport character. The sport, such as hockey, basketball, football
#' @param league character. NHL, NBA, etc.
#' @param ep character. The endpoint
#' @param query A list that defines the query parameters
#' @param version The API version. Current version is 1.
#' @param walk logical.  NOT YET IMPLEMENTED
#' @param page numeric. The page number to request
#' @param verbose logical. For debugging, returns response and parsed response.
#' 
#' @examples 
#' \dontrun{
#' TOKEN = "aklsjdlfkajsfdas"
#' results = queryAPI(TOKEN, 
#'                    sport="hockey", 
#'                    query=list(player_id="akjasdf")
#'                    version = 1,
#'                    ep = "stats",
#'                    verbose = F)
#' 
#' @export
queryAPI = function(token, 
                    sport="hockey", 
                    league = "nhl", 
                    ep="stats", 
                    query=list(), 
                    version=1, 
                    walk=F,
                    page=NA,
                    verbose=F) {
  
  ## TODO:  walk the results if > 20 entries
  ## TODO:  test to validate data types
  ## TODO:  best practices on walking the data? Have # entries paramter? 
  
  ## packages :  doesnt feel like this is the right way to do it
  library(httr)
  
  ## build the URL and the endpoint
  URL = sprintf("https://www.stattleship.com/%s/%s/%s", sport, league, ep)
  
  ## the accept parameters.  Is there a better way to do this?
  ACCEPT = sprintf("application/vnd.stattleship.com; version=%d", version)
  
  ## if page is supplied, add it to the list
  if (!is.na(page) & is.numeric(page) & page >= 1) {
    query = c(query, page=page)
  }
  
  ## test the body to see if it is a list and has values
  ## if not, just return an empty list
  ## todo: test to ensure that query is a list if !is.na
  ## get the request from the API
  resp = GET(URL,
             add_headers(Authorization =TOKEN, 
                         Accept = ACCEPT, 
                         `Content-Type`="application/json"), 
             query=query)
  
  ## walk the content if true
  
  ## convert response to text first, do not use baseline httr::content default
  api_response = content(resp, as="text")
  
  ## use jsonlite::fromJSON
  api_response = jsonlite::fromJSON(api_response)
  
  
  ## if verbose = T, return a list that includes the parsed results
  ## and the original request
  if (verbose) {
    api_response = list(response =  resp,
                        api_json = api_response)
  }
  
  ## return the data
  return(api_response)
}

About

Updated the queryAPI and created a new function, stattle that wraps it and makes walking easier. jsonlite is great, but I started to bump into some issues with dplyr::bind_rows. Right now these two functions play nice together, but needs to be further tested.

Load/Install

## source the functions
u = "https://gist.githubusercontent.com/Btibert3/8f3671af30c8d1afc555/raw/4d620b99cd6c8df3aa17a6c84ac07056a5d479f3/queryAPI.r"
devtools::source_url(u)
u = "https://gist.githubusercontent.com/Btibert3/8f3671af30c8d1afc555/raw/4d620b99cd6c8df3aa17a6c84ac07056a5d479f3/stattle.r"
devtools::source_url(u)

Requirements

We will handle these better shortly, but the code requires:

  • dplyr
  • httr
  • jsonlite

Basic Usage

nhl_teams = stattle(TOKEN, ep='teams', walk=T)
length(nhl_teams)
names(nhl_teams[[1]])

The nhl_teams object returns a list of lists. In the example above, there were two hits to the API, so length(nhl_teams) is 2. Nested below each of these parent lists are the data.

Collapse the team data into a dataframe.

dat = data.frame()
for (i in 1:length(nhl_teams)) {
  tmp_dat = nhl_teams[[i]]$teams
  dat = bind_rows(dat, tmp_dat)
  rm(tmp_dat)
}

TODO

  • put this into a more formal package
  • leverage way better version control