xml_url does not work (at least in combination with base_url)

[mwaldstein]

The behavior of xml_url appears to have changed pretty drastically in v1.3.0.

Previously, passing base_url via read_html resulted in it setting the url returned by xml_url

These are deeply simplified examples but highlight the change in behavior. I can work around the change from "NA" to "<CHARSXP: NA>", but the loss of the xml_url is a big loss which will require a bit of re-architecting.

This breaks edgarWebR (currently off CRAN due to vignettes making remote API calls)

Using string input

Example 1

require(xml2)
doc <- read_html("<html/>", base_url = "http://test.com")
xml_url(doc)

On v1.2.5 the output was "http://test.com"
On v1.3.1 the output is "UTF-8"

Example 2

require(xml2)
doc <- read_html("<html/>")
xml_url(doc)

On v1.2.5 the output was "NA"
Ov v1.3.0 the output is "UTF-8"

Using httr response

Example 3

require(xml2)
require(httr)
href <- "https://www.sec.gov/cgi-bin/cik_lookup?company=cloudera"
res <- GET(href)
doc <- read_html(res, base_url = href)
xml_url(doc)

On v1.2.5 the output was "https://www.sec.gov/cgi-bin/cik_lookup?company=cloudera"
Ov v1.3.0 the output is "<CHARSXP: NA>"

Example 4

require(xml2)
require(httr)
href <- "https://www.sec.gov/cgi-bin/cik_lookup?company=cloudera"
res <- GET(href)
doc <- read_html(res)
xml_url(doc)

On v1.2.5 the output was "NA"
Ov v1.3.0 the output is "<CHARSXP: NA>"

[jimhester]

Thanks, it was a bug introduced when converting the code not to use Rcpp, should now be fixed. We will be doing another xml2 release in the near future. If you notice any additional regressions please let us know!

[mwaldstein]

Thanks! Appreciate the responsiveness