Scraping with rvest - complete with NAs when tag is not present

Question

I want to parse this HTML: and get this elements from it:

a) p tag, with class: "normal_encontrado".
b) div with class: "price".

Sometimes, the p tag is not present in some products. If this is the case, an NA should be added to the vector collecting the text from this nodes.

The idea is to have 2 vectors with the same length, and after join them to make a data.frame. Any ideas?

The HTML part:

<html>
<head></head>
<body>

<div class="product_price" id="product_price_186251">
  <p class="normal_encontrado">
    S/. 2,799.00
  </p>

  <div id="WC_CatalogEntryDBThumbnailDisplayJSPF_10461_div_10" class="price">
    S/. 2,299.00
  </div>    
</div>

<div class="product_price" id="product_price_232046">
  <div id="WC_CatalogEntryDBThumbnailDisplayJSPF_10461_div_10" class="price">
    S/. 4,999.00
  </div>
</div>
</body>
</html>

R Code:

library(rvest)

page_source <- read_html("r.html")

r.precio.antes <- page_source %>%
html_nodes(".normal_encontrado") %>%
html_text()

r.precio.actual <- page_source %>%
html_nodes(".price") %>%
html_text()

Something like this might be helpful - [R dataframe from xml when values are multiple or missing](http://stackoverflow.com/questions/17349630/r-dataframe-from-xml-when-values-are-multiple-or-missing) — thelatemail, Oct 21 '15 at 06:29

G. Grothendieck · Answer 1 · 2015-10-26T14:47:52.087

Using the XML package parse the input with xmlTreeParse and then use xpathSApply to interate over the product_price class div nodes. For each such node the anonyous function gets the value of the div and p subnodes. The resulting character matrix m is reworked into a data frame DF and the columns are cleaned removing any character that is not a dot or digit and also removing any dot followed by a non-digit. Copnvert result to numeric. Note that no special processing for the missing p case is needed.

# input

Lines <- '<html>
<head></head>
<body>

<div class="product_price" id="product_price_186251">
  <p class="normal_encontrado">
    S/. 2,799.00
  </p>

  <div id="WC_CatalogEntryDBThumbnailDisplayJSPF_10461_div_10" class="price">
    S/. 2,299.00
  </div>    
</div>

<div class="product_price" id="product_price_232046">
  <div id="WC_CatalogEntryDBThumbnailDisplayJSPF_10461_div_10" class="price">
    S/. 4,999.00
  </div>
</div>
</body>
</html>'

# code to read input and produce a data.frame

library(XML)
doc <- xmlTreeParse(Lines, asText = TRUE, useInternalNodes = TRUE)

m <- xpathSApply(doc, "//div[@class = 'product_price']", function(node) {
  list(p = xmlValue(node[["p"]]), div = xmlValue(node[["div"]])) })

DF <- as.data.frame(t(m), stringsAsFactors = FALSE) # rework into data frame
DF[] <- lapply(DF, function(x) as.numeric(gsub("[^.0-9]|[.]\\D", "", x))) # clean

The result is:

> DF
     p  div
1 2799 2299
2   NA 4999

score 3 · Answer 2 · answered Oct 25 '15 at 02:53

Go one level up from your target and lapply over each parent element:

library(xml2)
library(rvest)

pg <- read_html('<html>
<head></head>
<body>

<div class="product_price" id="product_price_186251">
  <p class="normal_encontrado">
    S/. 2,799.00
  </p>

  <div id="WC_CatalogEntryDBThumbnailDisplayJSPF_10461_div_10" class="price">
    S/. 2,299.00
  </div>    
</div>

<div class="product_price" id="product_price_232046">
  <div id="WC_CatalogEntryDBThumbnailDisplayJSPF_10461_div_10" class="price">
    S/. 4,999.00
  </div>
</div>
</body>
</html>')

prod <- html_nodes(pg, "div.product_price")
do.call(rbind, lapply(prod, function(x) {
  norm <- tryCatch(xml_text(xml_node(x, "p.normal_encontrado")),
                   error=function(err) {NA})
  price <- tryCatch(xml_text(xml_node(x, "div.price")),
                    error=function(err) {NA})
  data.frame(norm, price, stringsAsFactors=FALSE)
}))

##                     norm                  price
## 1 \n    S/. 2,799.00\n   \n    S/. 2,299.00\n  
## 2                   <NA> \n    S/. 4,999.00\n

I have no idea if you wanted the strings trimmed or anything else done, but those machinations are pretty easy.

BjaRule · Accepted Answer · 2015-10-29T14:16:36.293

If the tag is not found, rvest returns a character(0). So assuming you will find at most one current and one regular price in each div.product_price, you can use this:

pacman::p_load("rvest", "dplyr")

get_prices <- function(node){
  r.precio.antes <- html_nodes(node, 'p.normal_encontrado') %>% html_text
  r.precio.actual <- html_nodes(node, 'div.price') %>% html_text

  data.frame(
    precio.antes = ifelse(length(r.precio.antes)==0, NA, r.precio.antes),
    precio.actual = ifelse(length(r.precio.actual)==0, NA, r.precio.actual), 
    stringsAsFactors=F
  )

}

doc <- read_html('test.html') %>% html_nodes("div.product_price")
lapply(doc, get_prices) %>%
  rbind_all

Edited: I misunderstood the input data, so changed the script to work with just a single html page.

The more clear method, thanks. I also like Grothendick method, but I've never used the XML package. — Omar Gonzales, Nov 01 '15 at 02:20

cocquemas · Answer 4 · 2015-10-24T16:24:01.957

It may not be the most idiomatic way to do this, but you can use lapply over the .product_price nodes like this:

r.precio.antes <- page_source %>% html_nodes(".product_price") %>%
  lapply(. %>% html_nodes(".normal_encontrado") %>% html_text() %>% 
     ifelse(identical(., character(0)), NA, .)) %>% unlist

This will return NA whenever the .normal_encontrado element is not found.

r.precio.antes
# [1] "\n                    S/. 2,799.00\n                "
# [2] NA  

length(r.precio.antes) # 2

If I wanted to develop the code to make it clearer, first I isolate the .product_price nodes:

product_nodes <- page_source %>% html_nodes(".product_price")

Then I could use lapply in more traditional way:

r.precio.antes <- lapply(product_nodes, function(pn) {
  pn %>% html_nodes(".normal_encontrado") %>% html_text()
})
r.precio.antes <- unlist(r.precio.antes)

Instead I'm using the magrittr syntax for lapply, see e.g. the end of the Functional sequences paragraph here.

One final hurdle is that if the element is not found, this will return character(0) rather than NA like you wanted. So I'm adding ifelse(identical(., character(0)), NA, .)) to the pipe inside the lapply to fix that.

Would you mind explaining the code? Especially this part: `lapply(. %>% html_nodes(".normal_encontrado")` why is the "." there (after lapply)? And also: `(function(x) ifelse(identical(x, character(0)), NA, x)) )`. Thanks. — Omar Gonzales, Oct 24 '15 at 15:34
Actually, I realized you could just do `ifelse(identical(., character(0)), NA, .))` instead of the `(function(x) ...)` syntax. I've developed the code and explanations. Is this clearer? — cocquemas, Oct 24 '15 at 16:25

Scraping with rvest - complete with NAs when tag is not present

4 Answers4

Linked

Related