Webscraping with R: How can I tidy my data? - r

Question

I'm building a web scraper for some News websites in Switzerland. After some trial & error and a lot of help from StackOverflow (thx everyone!), I've gotten to a point where I can get text data from all articles.
#packages instalieren
install.packages("rvest")
install.packages("tidyverse")
install.packages("dplyr")
library(rvest)
library(stringr)
#seite einlesen
apisrf<- read_xml('https://www.srf.ch/news/bnf/rss/1646')
urls_srf <- apisrf %>% html_nodes('link') %>% html_text()
zeit_srf <- apisrf %>% html_nodes('pubDate') %>% html_text()
#data.frame basteln
dfsrf_titel_text <- data.frame(Text = character())
#scrape
for(i in 1:length(urls_srf)) {
link <- urls_srf[i]
artikel <- read_html(link)
#Informationen entnehmen
textsrf<- artikel %>% html_nodes('p') %>% html_text()
#In Dataframe strukturieren
dfsrf_text <- data.frame(Text = textsrf)
dfsrf_titel_text <- rbind(dfsrf_titel_text, cbind(dfsrf_text))
}
running this gives me dfsrf_titel_text. (I'm going to combine it with the titles of the articles at some point but let that be my problem.)
however, now my data is pretty untidy and I can't really figure out how to clean it in a way so it works for me. Especially annoying is that the texts from the different articles are not really structured in that way but get a new line whenever there is a paragraph in the texts. I can't combine the paragraphs because all the texts have different lengths. (The first article, starting at point 3, is super long because it's a live ticker covering the corona crisis so don't get confused if you run my code.)
how can I get R to create a new row in my dataframe only if the text is from a new article (meaning from a new URL?
thx for your help!

Answer 1

can you provide a sample of your data? you can use the strsplit(string, pattern) function where the pattern you specify is something that only happens between articles. Perhaps the URL?
strsplit(dfsrf_text,"www.\\w+.ch")
That will split your text anytime a URL in the .ch domain is found. you can use this regular expression cheat sheet to help you identify the pattern that seperates your articles.

Answer 2

You should correct this while creating dataframe itself. Here I am binding this all the data for each article together using paste0 adding new line character between them (\n\n).
library(rvest)
for(i in 1:length(urls_srf)) {
link <- urls_srf[i]
artikel <- read_html(link)
#Informationen entnehmen
textsrf<- paste0(artikel %>% html_nodes('p') %>% html_text(), collapse = "\n\n")
#In Dataframe strukturieren
dfsrf_text <- data.frame(Text = textsrf)
dfsrf_titel_text <- rbind(dfsrf_titel_text, cbind(dfsrf_text))
}
However, growing data in a loop is highly inefficient and can slow the process terribly especially when you have large data to scrape like this. Try using sapply.
dfsrf_titel_text <- data.frame(text = sapply(urls_srf, function(x) {
paste0(read_html(x) %>% html_nodes('p') %>% html_text(), collapse = "\n\n")
}))
So this will give you number of rows same as length of urls_srf .