How to wrap css and xpath arguments in html_nodes in a generic function

I want to create a wrapper around html_node that is capable of reading CSS and XPATH arguments. I want to create a quoted expression that can be supplied to html_node and be evaluated at the spot. I figured out how to create the path argument for css and xpath respectively, but when I supply this expression to html_node it does not work. Why not?

page_parser <- function(dat_list, path = NULL, css = FALSE, attr = "") {
  library(rlang)
  # make css or path argument for html_nodes
  if (css == TRUE) {
    path <- expr(`=`(css, !!path))
  }else{
    path <- expr(`=`(xpath, !!path))
  }
  # extract attribute value
  map(dat_list, possibly(function(x) { html_nodes(x, !!path) %>% html_attr(attr) %>% extract(1)}, NA)) %>% 
     map(1) %>% 
     lapply(function(x) ifelse(is_null(x), "", x)) %>%
     unlist()
 }

read_html("https://www.freitag.de/autoren/lutz-herden/alexis-tsipras-fall-oder-praezedenzfall" %>% parge_parser(path = "//meta[@property='og:title']")

read_html("https://www.freitag.de/autoren/lutz-herden/alexis-tsipras-fall-oder-praezedenzfall" %>% parge_parser(path = ".title", css = TRUE)

The function should spit out the content of behind the css or xpath, no matter whether I specified a CSS or a Xpath.

Best, Moritz

Solution

In general, !! operator only works in functions that support quasiquoation. Unfortunately, rvest::html_nodes currently does not. (But since it's part of tidyverse, I wouldn't be surprised if the support is added at a later date.)

There are several ways to programmatically provide arguments to a function call, including do.call() from base R. However, given that you're using map to traverse your page, I recommend pre-setting css or xpath argument of html_nodes through purrr::partial():

page_parser <- function(dat_list, path = NULL, css = FALSE, attr = "") {
  # make css or xpath argument for html_nodes
  if (css == TRUE) {
    f_html_nodes <- purrr::partial( html_nodes, css = path )
  }else{
    f_html_nodes <- purrr::partial( html_nodes, xpath = path )
  }

  # extract attribute value
  map(dat_list, possibly(function(x) { f_html_nodes(x) %>% html_attr(attr) %>% 
                                         extract(1)}, NA)) %>% 
                  map(1) %>% lapply(function(x) ifelse(is_null(x), "", x)) %>%
    unlist()
}