【问题标题】：Extract text and links from unbalanced html table从不平衡的 html 表中提取文本和链接
【发布时间】：2019-04-03 06:49:11
【问题描述】：

我有类似格式的表格...

我正在尝试使用 R 提取文本和链接。

# write the HTML code from R to reproduce 
x <- "
<html>
<head>
</head>
<body>
<table>
<tbody>
<tr>
<th>site</th>
<th>country</th>
</tr>
<tr>
<td> <a href='http://www.nbc.com'>NBC</a>   <a href='https://www.cnn.com'>CNN</a>  <a href='https://www.nytimes.com'>NY Times</a> </td>
<td> US</td>
</tr>
<tr>
<td> <a href='http://www.dw-world.de/'>DW</a>  </td>
<td> DE</td>
</tr>
<tr>
<td></td>
<td>FR</td>
</tr>
<tr>
<td> <a href='http://www.bbc.co.uk'>BBC</a>   <a href='https://www.itv.co.uk'>ITV</a></td>
<td> UK</td>
</tr>
</tbody>
</table>
</body>
</html>"
write.table(x = x, file = "table.html",  quote = FALSE,
            col.names = FALSE,
            row.names = FALSE)
file.show("table.html")

最终我想要这样整洁的数据框...

# # A tibble: 7 x 3
#   site                    site_name country
#   <chr>                   <chr>     <chr>  
# 1 http://www.nbc.com      NBC       US     
# 2 https://www.cnn.com     CNN       US     
# 3 https://www.nytimes.com NY Times  US     
# 4 http://www.dw-world.de/ DW        DE     
# 5 NA                      NA        FR     
# 6 http://www.bbc.co.uk    BBC       UK     
# 7 https://www.itv.co.uk   ITV       UK

我一直在玩rvest 函数，但我无法提取链接并记住它们来自哪一行来构建上述数据框？

library(tidyverse)
library(rvest)

h <- read_html("table.html") 

# a table without any of the links... no good
h %>%
  html_table() %>%
  .[[1]]
#                  site country
# 1 NBC   CNN  NY Times      US
# 2                  DW      DE
# 3                          FR
# 4           BBC   ITV      UK

# pulls the site urls    
h %>%
  html_nodes("a") %>% 
  html_attr("href")
# [1] "http://www.nbc.com"      "https://www.cnn.com"     "https://www.nytimes.com" "http://www.dw-world.de/" "http://www.bbc.co.uk"    "https://www.itv.co.uk"     

# pulls the site names
h %>%
  html_nodes("a") %>% 
  html_text()
# [1] "NBC"      "CNN"      "NY Times" "DW"       "BBC"      "ITV" 

# looks promising, perhaps can combine with results from html_table()
library(XML)
tables <- getNodeSet(htmlParse("table.html"), "//table")
hrefFun <- function(x){
  xpathSApply(x,'./a',xmlAttrs)  
}
readHTMLTable(doc = tables[[1]], elFun = hrefFun)
# V1     V2
# 1                                                                                         list() list()
# 2 c(href = "http://www.nbc.com", href = "https://www.cnn.com", href = "https://www.nytimes.com") list()
# 3                                                                        http://www.dw-world.de/ list()
# 4                                                                                         list() list()
# 5                               c(href = "http://www.bbc.co.uk", href = "https://www.itv.co.uk") list()

# looks promising for the rows.. don't know where to go from here
h %>%
  html_nodes("tr")
# {xml_nodeset (5)}
# [1] <tr>\n<th>site</th>\r\n<th>country</th>\r\n</tr>\n
# [2] <tr>\n<td> <a href="http://www.nbc.com">NBC</a>   <a href="https://www.cnn.com">CNN</a>  <a href="https://www.nytimes.com">NY Times</a>  ...
# [3] <tr>\n<td> <a href="http://www.dw-world.de/">DW</a>  </td>\r\n<td> DE</td>\r\n</tr>\n
# [4] <tr>\n<td></td>\r\n<td>FR</td>\r\n</tr>\n
# [5] <tr>\n<td> <a href="http://www.bbc.co.uk">BBC</a>   <a href="https://www.itv.co.uk">ITV</a>\n</td>\r\n<td> UK</td>\r\n</tr>

【问题讨论】：

向我们展示您尝试过的代码
@JackWasey 添加了一些代码

标签： r web-scraping html-table rvest

【解决方案1】：

当然还有改进的余地，但它确实有效：

extract_data <- function(url) {
  table <- read_html(url) %>%
      html_table() %>%
      data.frame()

  site <- read_html(url) %>%
    html_nodes("a") %>%
    html_text()

  site_name <- read_html(url) %>%
    html_nodes("a") %>%
    html_attr("href")

  df <- data.frame(site_name = site_name, site = site, stringsAsFactors = F)

  df$country <- table$country[apply(df, 1, function(x) grep(x[2], table$site))]

  emptyRow <- data.frame(site = rep(NA, length(table$country[table$site ==""])),
                         site_name = rep(NA, length(table$country[table$site ==""])),
                         country = table$country[table$site ==""])
  df <- rbind(df, emptyRow)
  df
}

extract_data("table.html")

#                site_name     site country
#1      http://www.nbc.com      NBC      US
#2     https://www.cnn.com      CNN      US
#3 https://www.nytimes.com NY Times      US
#4 http://www.dw-world.de/       DW      DE
#5    http://www.bbc.co.uk      BBC      UK
#6   https://www.itv.co.uk      ITV      UK
#7                    <NA>     <NA>      FR

【讨论】：

【解决方案2】：

您可以通过三个步骤来完成：

创建一个包含所有映射的表。
清理未平衡的表格并将其转换为整洁的格式。
加入这两个表

第 1 步：

#create mappings 
nodes_with_info <- read_html(x) %>%
    html_nodes("a")
map_table <- tibble(site = nodes_with_info %>%
                        html_text() %>%
                        str_trim(),
                    href = nodes_with_info %>%
                        html_attr("href"))

结果：

小数点：6 x 2

site     href                   
<chr>    <chr>                  
1 NBC      http://www.nbc.com     
2 CNN      https://www.cnn.com    
3 NY Times https://www.nytimes.com
4 DW       http://www.dw-world.de/
5 BBC      http://www.bbc.co.uk   
6 ITV      https://www.itv.co.uk

第 2 步：

my_table <- read_html(x) %>%
    html_table() %>%
    .[[1]]

#create table of desired format
structured_table <- my_table %>%
    mutate(site = str_split(site,"  ")) %>%
    unnest() %>%
    mutate(site = str_trim(site))

结果：

country     site
1      US      NBC
2      US      CNN
3      US NY Times
4      DE       DW
5      FR         
6      UK      BBC
7      UK      ITV

第 3 步：

structured_table %>%
    left_join(map_table,by = "site")

结果：

country     site                    href
1      US      NBC      http://www.nbc.com
2      US      CNN     https://www.cnn.com
3      US NY Times https://www.nytimes.com
4      DE       DW http://www.dw-world.de/
5      FR                             <NA>
6      UK      BBC    http://www.bbc.co.uk
7      UK      ITV   https://www.itv.co.uk

【讨论】：