Scraping a JavaScript object and converting to JSO

问题:

I am scraping the following website: https://www.banorte.com/wps/portal/ixe/Home/indicadores/tipo-de-cambio

I am trying to get the table of currency exchange rates into an R data frame via the rvest package, but the table itself is configured in a JavaScript variable within the HTML code.

I located the relevant css selector and now I have this:

library(rvest)    
banorte <- "https://www.banorte.com/wps/portal/ixe/Home/indicadores/tipo-de-cambio/" %>%
      read_html() %>%
      html_nodes('#indicadores_financieros_wrapper > script:nth-child(2)')

my output is now the following JavaScript script, as an XML nodeset:

<script>
$(document).ready(function(){
    var valor = '{"tablaDivisas":[{"nombreDivisas":"FRANCO SUIZO","compra":"18.60","venta":"19.45"}, {"nombreDivisas":"LIBRA ESTERLINA","compra":"24.20","venta":"25.15"}, {"nombreDivisas":"YEN JAPONES","compra":"0.1635","venta":"0.171"}, {"nombreDivisas":"CORONA SUECA","compra":"2.15","venta":"2.45"}, {"nombreDivisas":"DOLAR CANADA","compra":"14.50","venta":"15.35"}, {"nombreDivisas":"EURO","compra":"21.75","venta":"22.60"}], "tablaDolar":[{"nombreDolar":"VENTANILLA","compra":"17.73","venta":"19.15"}]}';
    if(valor != '{}'){
        var objJSON = eval("(" + valor + ")");
        var tabla="<tbody>";
        for ( var i = 0; i < objJSON["tablaDolar"].length; i++) {
            tabla+= "<tr>";
            tabla+= "<td>" + objJSON["tablaDolar"][i].nombreDolar + "</td>";
            tabla+= "<td>$" + objJSON["tablaDolar"][i].compra + "</td>";
            tabla+= "<td>$" + objJSON["tablaDolar"][i].venta + "</td>";
            tabla+= "</tr>";
        }
        tabla+= "</tbody>";
        $("#tablaDolar").append(tabla);
        var tabla2="";
        for ( var i = 0; i < objJSON["tablaDivisas"].length; i++) {
            tabla2+= "<tr>";
            tabla2+= "<td>" + objJSON["tablaDivisas"][i].nombreDivisas + "</td>";
            tabla2+= "<td>$" + objJSON["tablaDivisas"][i].compra + "</td>";
            tabla2+= "<td>$" + objJSON["tablaDivisas"][i].venta + "</td>";
            tabla2+= "</tr>";
        }
        tabla2+= "</tbody>";
        $("#tablaDivisas").append(tabla2);
    }
    bmnIndicadoresResponsivoInstance.cloneResponsive(0);
});
</script>

My question is, how do I remove almost everything (all the JavaScript functions/operators) to get only this data with the intention of converting it eventually to a JSON table like this:

{"tablaDivisas":[{"nombreDivisas":"FRANCO SUIZO","compra":"18.60","venta":"19.45"},
{"nombreDivisas":"LIBRA ESTERLINA","compra":"24.20","venta":"25.15"},
{"nombreDivisas":"YEN JAPONES","compra":"0.1635","venta":"0.171"}, 
{"nombreDivisas":"CORONA SUECA","compra":"2.15","venta":"2.45"}, 
{"nombreDivisas":"DOLAR CANADA","compra":"14.50","venta":"15.35"}, 
{"nombreDivisas":"EURO","compra":"21.75","venta":"22.60"}],
"tablaDolar":[{"nombreDolar":"VENTANILLA","compra":"17.73","venta":"19.15"}]}

In other words, I need to extract the "valor" variable from the JS script using R.

For some reason I've had trouble getting this done all within R (without having to export the variable as an external .txt file and then using a substring)

回答1:

Definitely a bit more heavyweight answer but generalizes to other, more gnarly "javascript problems".

library(rvest)
library(stringi)
library(V8)
library(tidyverse)

banorte <- "https://www.banorte.com/wps/portal/ixe/Home/indicadores/tipo-de-cambio/" %>%
      read_html() %>%
      html_nodes('#indicadores_financieros_wrapper > script:nth-child(2)')

We'll setup a javascript V8 context:

ctx <- v8()

Then:

• get the <script> content
• split it into lines
• get it into a plain character vector
• remove the cruft
• evaluate the javascript

which is not too bad:

html_text(banorte) %>% 
  stri_split_lines() %>% 
  flatten_chr() %>% 
  keep(stri_detect_regex, "^\tvar") %>% 
  ctx$eval()

Since that javascript is a JSON string, we do the eval in R vs V8:

jsonlite::fromJSON(ctx$get("valor"))
## $tablaDivisas
##     nombreDivisas compra venta
## 1    FRANCO SUIZO  18.60 19.45
## 2 LIBRA ESTERLINA  24.20 25.15
## 3     YEN JAPONES 0.1635 0.171
## 4    CORONA SUECA   2.15  2.45
## 5    DOLAR CANADA  14.50 15.35
## 6            EURO  21.75 22.60
## 
## $tablaDolar
##   nombreDolar compra venta
## 1  VENTANILLA  17.73 19.15

If there had been other, useful processing in javascript, this generalizes better.

NOTE: Google translate in my Chrome beta channel was not translating the site well but I think you're awfully close to being in violation of the spirit of item 6 on the "Términos Legales" page but until I can translate it I can't fully tell. When/if I can and it seems like you are I'll delete this.

回答2:

You could do this:

library(rvest)    
banorte <- "https://www.banorte.com/wps/portal/ixe/Home/indicadores/tipo-de-cambio/" %>%
    read_html() %>%
    html_nodes('#indicadores_financieros_wrapper > script:nth-child(2)') %>%
    as_list()

banorte_vec <- strsplit(banorte[[c(1,1)]],"\r\n")[[1]]
valor <- grep("valor = ", banorte_vec, value = T)
valor <- gsub("\tvar valor = ","",valor)
valor <- gsub("';$","",valor)
valor <- gsub("^'","",valor)

library(jsonlite)
result <- fromJSON(valor)
result

$tablaDivisas
  nombreDivisas compra venta
1    FRANCO SUIZO  18.60 19.45
2 LIBRA ESTERLINA  24.20 25.15
3     YEN JAPONES 0.1635 0.171
4    CORONA SUECA   2.15  2.45
5    DOLAR CANADA  14.50 15.35
6            EURO  21.75 22.60

$tablaDolar
  nombreDolar compra venta
1  VENTANILLA  17.73 19.15