loop across multiple urls in r with rvest [duplica

问题:

This question already has an answer here:

• Harvest (rvest) multiple HTML pages from a list of urls 1 answer

I have a series of 9 urls that I would like to scrape data from:

http://www.basketball-reference.com/play-index/draft_finder.cgi?request=1&year_min=2001&year_max=2014&round_min=&round_max=&pick_overall_min=&pick_overall_max=&franch_id=&college_id=0&is_active=&is_hof=&pos_is_g=Y&pos_is_gf=Y&pos_is_f=Y&pos_is_fg=Y&pos_is_fc=Y&pos_is_c=Y&pos_is_cf=Y&c1stat=&c1comp=&c1val=&c2stat=&c2comp=&c2val=&c3stat=&c3comp=&c3val=&c4stat=&c4comp=&c4val=&order_by=year_id&order_by_asc=&offset=0 

The offset= at the end of the link goes from 0 up to 900 (by 100) when pages change through the last page. I would like to loop through each page and scrape each table, then use rbind to stack each df on top of one another in sequence. I have been using rvest and would like to use lapply since I am better with that than for loops.

The question is similar to this (Harvest (rvest) multiple HTML pages from a list of urls) but different because I would prefer not to have to copy all the links to one vector before running the program. I would like a general solution to how to loop over multiple pages and harvest the data, creating a data frame each time.

The following works for the first page:

library(rvest)
library(stringr)
library(tidyr)

site <- 'http://www.basketball-reference.com/play-index/draft_finder.cgi?request=1&year_min=2001&year_max=2014&round_min=&round_max=&pick_overall_min=&pick_overall_max=&franch_id=&college_id=0&is_active=&is_hof=&pos_is_g=Y&pos_is_gf=Y&pos_is_f=Y&pos_is_fg=Y&pos_is_fc=Y&pos_is_c=Y&pos_is_cf=Y&c1stat=&c1comp=&c1val=&c2stat=&c2comp=&c2val=&c3stat=&c3comp=&c3val=&c4stat=&c4comp=&c4val=&order_by=year_id&order_by_asc=&offset=0' 

webpage <- read_html(site)
draft_table <- html_nodes(webpage, 'table')
draft <- html_table(draft_table)[[1]]

But I would like to repeat this over all pages without having to paste the urls into a vector. I tried the following and it didn't work:

jump <- seq(0, 900, by = 100)
site <- paste('http://www.basketball-reference.com/play-index/draft_finder.cgi?request=1&year_min=2001&year_max=2014&round_min=&round_max=&pick_overall_min=&pick_overall_max=&franch_id=&college_id=0&is_active=&is_hof=&pos_is_g=Y&pos_is_gf=Y&pos_is_f=Y&pos_is_fg=Y&pos_is_fc=Y&pos_is_c=Y&pos_is_cf=Y&c1stat=&c1comp=&c1val=&c2stat=&c2comp=&c2val=&c3stat=&c3comp=&c3val=&c4stat=&c4comp=&c4val=&order_by=year_id&order_by_asc=&offset=', jump,'.htm', sep="")

webpage <- read_html(site)
draft_table <- html_nodes(webpage, 'table')
draft <- html_table(draft_table)[[1]]

So there should be a data frame for each page and I imagine it would be easier to put them in a list and then use rbind to stack them.

Any help would be greatly appreciated!

回答1:

You are attempting to vectorize a method that cannot take multiple items in one call. Specifically, read_html() requires one page per call since it needs to read in web data one at a time and expects a scalar value. Consider looping through the site list with lapply then bind all dfs together:

jump <- seq(0, 800, by = 100)
site <- paste('http://www.basketball-reference.com/play-index/draft_finder.cgi?',
              'request=1&year_min=2001&year_max=2014&round_min=&round_max=',
              '&pick_overall_min=&pick_overall_max=&franch_id=&college_id=0',
              '&is_active=&is_hof=&pos_is_g=Y&pos_is_gf=Y&pos_is_f=Y&pos_is_fg=Y',
              '&pos_is_fc=Y&pos_is_c=Y&pos_is_cf=Y&c1stat=&c1comp=&c1val=&c2stat=&c2comp=',
              '&c2val=&c3stat=&c3comp=&c3val=&c4stat=&c4comp=&c4val=&order_by=year_id',
              '&order_by_asc=&offset=', jump, sep="")

dfList <- lapply(site, function(i) {
    webpage <- read_html(i)
    draft_table <- html_nodes(webpage, 'table')
    draft <- html_table(draft_table)[[1]]
})

finaldf <- do.call(rbind, dfList)             # ASSUMING ALL DFs MAINTAIN SAME COLS

回答2:

You can use curl to run all of the requests at once. I Be nice to the sites that may have small servers and don't blow them up. With this code you can use the lapply at the end to clean up the table so you can stack it with do.call(rbind, AllOut) but I will leave that to you.

library(rvest)
library(stringr)
library(tidyr)

OffSet <- seq(0, 900, by = 100)

Sites <- paste0('http://www.basketball-reference.com/play-index/draft_finder.cgi?request=1&year_min=2001&year_max=2014&round_min=&round_max=&pick_overall_min=&pick_overall_max=&franch_id=&college_id=0&is_active=&is_hof=&pos_is_g=Y&pos_is_gf=Y&pos_is_f=Y&pos_is_fg=Y&pos_is_fc=Y&pos_is_c=Y&pos_is_cf=Y&c1stat=&c1comp=&c1val=&c2stat=&c2comp=&c2val=&c3stat=&c3comp=&c3val=&c4stat=&c4comp=&c4val=&order_by=year_id&order_by_asc=&offset=', OffSet)


library(curl)

out <<- list()
# This is function, function which will be run if data vendor call is successful
complete = function(res){
  # cat("Request done! Status:", res$status, "\n")
  out <<- c(out, list(res))
}

for(i in 1:length(Sites)){
  curl_fetch_multi(
    Sites[i]
    , done = complete
    , fail = print
    , handle = new_handle(customrequest = "GET")
    )
}

multi_run()

AllOut <- lapply(out, function(x){

  webpage <- read_html(x$content)
  draft_table <- html_nodes(webpage, 'table')
  Tab <- html_table(draft_table)
  if(length(Tab) == 0){
    NULL
  } else {
    Tab
  }

})