Namespace aliases

[jeroen]

In this example borrowed form the rversions package lp1: and lp3: namespace are aliases of D: however xml_ns does not list them at all, making it difficult to query the document.

library(curl)
library(xml2)
h <- handle_setheaders(new_handle(customrequest = "PROPFIND"), Depth="1")
req <- curl_fetch_memory("http://svn.r-project.org/R/tags/", handle = h)
doc <- read_xml(rawToChar(req$content))
prop <- xml_find_all(doc, ".//D:propstat/D:prop", xml_ns(doc))

# does not work
xml_find_one(prop, ".//lp1:getetag", xml_ns(doc))

# works
xml_find_one(prop, ".//D:getetag", xml_ns(doc))

[hadley]

Why does that make it hard to navigate the document? The goal of xml_ns is to provide a unique abbreviation for each namespace.

[jeroen]

I found it confusing that the node in the xml is named lp1:getetag but searching for this node gives:

> xml_find_one(prop, ".//lp1:getetag", xml_ns(doc))
xmlXPathEval: evaluation failed
Error: No matches
In addition: Warning message:
In node_find_one(x$node, x$doc, xpath = xpath, nsMap = ns) :
  Undefined namespace prefix [1219]

Only by manually inspecting the xml I found that lp1: is an alias of D: in this case.

[hadley]

That's how xml namespaces work though - the prefix is completely arbitrary, it's only the url that identifies it.

But maybe xml_ns() shouldn't just return unique prefixes? Would need some testing - I'm not sure if the libxml's xpath code will allow the same url to be mapped to multiple prefixes.

[jimhester]

Having aliases to the same URI seems to work OK in limited testing. I think we should not return only unique prefixes.

library(curl)
library(xml2)
h <- handle_setheaders(new_handle(customrequest = "PROPFIND"), Depth="1")
req <- curl_fetch_memory("http://svn.r-project.org/R/tags/", handle = h)
doc <- read_xml(rawToChar(req$content))
ns <- xml_ns(doc)

# Add the additional alias
ns[["lp1"]] <- "DAV:"

prop <- xml_find_all(doc, ".//D:propstat/D:prop", ns)
xml_find_one(prop, ".//lp1:getetag", ns)
#> {xml_nodeset (141)}
#>  [1] <lp1:getetag>W/"70574//tags"</lp1:getetag>
#>  [2] <lp1:getetag>W/"10156//tags/R-before-GenGC-merge"</lp1:getetag>
#>  [3] <lp1:getetag>W/"371//tags/R-0-60"</lp1:getetag>
#>  [4] <lp1:getetag>W/"1452//tags/R-0-62-1"</lp1:getetag>
#>  [5] <lp1:getetag>W/"477//tags/R-0-61"</lp1:getetag>
#>  [6] <lp1:getetag>W/"1721//tags/R-0-62-2"</lp1:getetag>
#>  [7] <lp1:getetag>W/"1447//tags/R-0-62"</lp1:getetag>
#>  [8] <lp1:getetag>W/"2012//tags/R-0-62-3"</lp1:getetag>
#>  [9] <lp1:getetag>W/"2621//tags/R-0-62-4"</lp1:getetag>
#> [10] <lp1:getetag>W/"2838//tags/R-0-63"</lp1:getetag>
#> [11] <lp1:getetag>W/"4084//tags/R-0-64"</lp1:getetag>
#> [12] <lp1:getetag>W/"5635//tags/R-0-65"</lp1:getetag>
#> [13] <lp1:getetag>W/"8828//tags/R-1-0-1"</lp1:getetag>
#> [14] <lp1:getetag>W/"62482//tags/R-3-0-0"</lp1:getetag>
#> [15] <lp1:getetag>W/"62744//tags/R-3-0-1"</lp1:getetag>
#> [16] <lp1:getetag>W/"35751//tags/R-2-2-0"</lp1:getetag>
#> [17] <lp1:getetag>W/"36814//tags/R-2-2-1"</lp1:getetag>
#> [18] <lp1:getetag>W/"63988//tags/R-3-0-2"</lp1:getetag>
#> [19] <lp1:getetag>W/"18029//tags/R-1-4-1"</lp1:getetag>
#> [20] <lp1:getetag>W/"65127//tags/R-3-0-3"</lp1:getetag>
#> ...

# specifying the URI explicitly also works
xml_find_one(prop, ".//*[local-name()='getetag' and namespace-uri()='DAV:']")
#> {xml_nodeset (141)}
#>  [1] <lp1:getetag>W/"70574//tags"</lp1:getetag>
#>  [2] <lp1:getetag>W/"10156//tags/R-before-GenGC-merge"</lp1:getetag>
#>  [3] <lp1:getetag>W/"371//tags/R-0-60"</lp1:getetag>
#>  [4] <lp1:getetag>W/"1452//tags/R-0-62-1"</lp1:getetag>
#>  [5] <lp1:getetag>W/"477//tags/R-0-61"</lp1:getetag>
#>  [6] <lp1:getetag>W/"1721//tags/R-0-62-2"</lp1:getetag>
#>  [7] <lp1:getetag>W/"1447//tags/R-0-62"</lp1:getetag>
#>  [8] <lp1:getetag>W/"2012//tags/R-0-62-3"</lp1:getetag>
#>  [9] <lp1:getetag>W/"2621//tags/R-0-62-4"</lp1:getetag>
#> [10] <lp1:getetag>W/"2838//tags/R-0-63"</lp1:getetag>
#> [11] <lp1:getetag>W/"4084//tags/R-0-64"</lp1:getetag>
#> [12] <lp1:getetag>W/"5635//tags/R-0-65"</lp1:getetag>
#> [13] <lp1:getetag>W/"8828//tags/R-1-0-1"</lp1:getetag>
#> [14] <lp1:getetag>W/"62482//tags/R-3-0-0"</lp1:getetag>
#> [15] <lp1:getetag>W/"62744//tags/R-3-0-1"</lp1:getetag>
#> [16] <lp1:getetag>W/"35751//tags/R-2-2-0"</lp1:getetag>
#> [17] <lp1:getetag>W/"36814//tags/R-2-2-1"</lp1:getetag>
#> [18] <lp1:getetag>W/"63988//tags/R-3-0-2"</lp1:getetag>
#> [19] <lp1:getetag>W/"18029//tags/R-1-4-1"</lp1:getetag>
#> [20] <lp1:getetag>W/"65127//tags/R-3-0-3"</lp1:getetag>
#> ...

[hadley]

Yeah, that sounds fine to me.