IN_PROGRESS: Mutating/Creating nodes

[jimhester]

Clearly this work isn't done, but this PR will be a good place for discussion.

[jimhester]

@hadley I think this is mostly in a good state now and modifying existing nodes has most of the features I think we want (with possible exception of setting the namespace of a node). I wrote a simple vignette to explain the functionality but you can also look at the tests to get a feel for the API.

Questions

• What should the modification/replacement functions do with xml_missing objects, throw an error or return them unchanged? If the latter should we issue a warning?
• How do we want to support creation of new nodes? Currently you can use read_xml() with character vector input and use the modification functions to insert the node into a tree. But this is clearly less efficient than creating the nodes directly.
• xml_text()<- uses xmlNodeSetContentLen internally, which means that if it is called on a parent with non-text children they will be destroyed. This is the expected behavior for xmlNodeSetContentLen, but it could cause confusion or errors for users if they have objects pointing to the child nodes.

[hadley]

I think these look better with {}

[hadley]

For creation, I'm imagining something like:

xml_new_node("nodename", child1, child2, attr1 = "foo", attr3 = "bar")

I'm not sure how namespaces would work here. One option would be to have special name and attribute constructors:

xml_new_node(nsName("url", "nodename"), child1, child2, nsAttr("attr1", "url", "foo"))

And instead of url could you use a prefix, if you also supplied the ns map.

[hadley]

Need to check that we can solve http://stackoverflow.com/questions/32939229/creating-xml-in-r-with-namespaces/32941524#32941524 reasonably elegantly.

[jimhester]

After a81f82c you can now do the following to get the desired XML from the SO post.

Maybe it would be nicer to have the second argument to xml_new_namespace() be a character vector of namespaces to add, with the names as the prefixes.

d <- xml_new_document(xml_new_node("sld", version = "1.1.0", xml_new_node("layer", xml_new_node("Name"))))
sld <- xml_find_one(d, "//sld")
name <- xml_find_one(d, "//Name")
xml_new_namespace(sld, "http://www.o.net/sld")
xml_new_namespace(sld, "http://www.o.net/ogc", "ogc")
xml_new_namespace(sld, "http://www.o.net/se", "se")
xml_set_namespace(name, "se")
cat(as.character(d))
#> <?xml version="1.0"?>
#> <sld xmlns="http://www.o.net/sld" xmlns:ogc="http://www.o.net/ogc" xmlns:se="http://www.o.net/se" version="1.1.0"><layer><se:Name/></layer></sld>

[hadley]

Yeah, I like the idea to add multiple namespaces at once.

Could we make something like this work?

xml_new_document() %>% 
  xml_add_child("sld", version = "1.1.0") %>% 
  xml_add_namespace(
    "http://www.o.net/sld", 
    ogc = "http://www.o.net/ogc",
    se = "http://www.o.net/se"
  ) %>% 
  xml_add_child("layer") %>% 
  xml_add_child("se:Name")

[hadley]

Or maybe even:

xml_new_document() %>% 
  xml_add_child("sld", 
    version = "1.1.0"
    ns = "http://www.o.net/sld", 
    "ns:ogc" = "http://www.o.net/ogc",
    "ns:se" = "http://www.o.net/se"
  ) %>% 
  xml_add_child("layer") %>% 
  xml_add_child("se:Name")

(That would require more work on the backend to parse the attribute names and create ns objects where necessary)

[jimhester]

Yeah I can get that working. The main issue is the current xml_add_* functions all expect the value argument to be a xml_node, so I either need to define a xml_add_new_* functions to create a new node with the specified name or have different behavior if value is a character(1). The current implementation was the simplest path forward with the current API, but I see the appeal of the proposed interface.

I think a format like the following will be possible (ns needs to be .ns to avoid classes with an attribute named ns).

xml_new_document() %>% 
  xml_add_child("sld", 
    version = "1.1.0"
    .ns = c("http://www.o.net/sld", 
      ogc = "http://www.o.net/ogc",
      se = "http://www.o.net/se")
  ) %>% 
  xml_add_child("layer") %>% 
  xml_add_child("se:Name")

[hadley]

Doing S3 dispatch on character vs. xmlNode seems reasonable to me. But I'm not sure what using .ns buys us over ns, ns:ogc etc.

[jimhester]

How do you create an attribute named 'ns' otherwise? The argument would
have to be named 'xmlns' to avoid a conflict.
On May 6, 2016 6:14 PM, "Hadley Wickham" [email protected] wrote:

Doing S3 dispatch on character vs. xmlNode seems reasonable to me. But
I'm not sure what using .ns buys us over ns, ns:ogc etc.

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#76 (comment)

[hadley]

Oh yeah, I meant xmlns instead of ns everywhere (I forgot what ns attributes look like)

[jimhester]

as of c2fcf36 you can now do

d1 <- xml_new_document(
  xml_new_node("sld",
    version = "1.1.0",
    xmlns = "http://www.o.net/sld",
    "xmlns:ogc" = "http://www.o.net/ogc",
    "xmlns:se" = "http://www.o.net/se") %>%
  xml_add_child("layer") %>%
  xml_add_child("se:Name"))
as.character(d1)
#> [1] "<?xml version=\"1.0\"?>\n<sld xmlns=\"http://www.o.net/sld\" xmlns:ogc=\"http://www.o.net/ogc\" xmlns:se=\"http://www.o.net/se\" version=\"1.1.0\"><layer/><se:Name/></sld>\n"

Using the exact syntax from above for xml_new_document() is tricky as xml_document objects expect to have a root node defined. I am still working on a good way to implement it.

[jimhester]

As of 2c3c002 you can use the following

d1 <- xml_new_document()
xml_add_child(d1, "sld",
    version = "1.1.0",
    xmlns = "http://www.o.net/sld",
    "xmlns:ogc" = "http://www.o.net/ogc",
    "xmlns:se" = "http://www.o.net/se") %>%
  xml_add_child("layer") %>%
  xml_add_child("se:Name")
#> {xml_node}
#> <Name>
as.character(d1)
#> [1] "<?xml version=\"1.0\"?>\n<sld xmlns=\"http://www.o.net/sld\" xmlns:ogc=\"http://www.o.net/ogc\" xmlns:se=\"http://www.o.net/se\" version=\"1.1.0\"><layer><se:Name/></layer></sld>\n"

The only thing preventing you doing a full chain like the original example is that xml_add_child() returns the child node rather than the parent so you have way to access the whole document otherwise.

[hadley]

Perfect!!

@jennybc how does this look to you?

[jennybc]

I will try this out in the next day or two.

[codecov-io]

Current coverage is 59.45%

Merging #76 into master will increase coverage by +9.45%

@@             master        #76   diff @@
==========================================
  Files            29         30     +1   
  Lines           852       1233   +381   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
+ Hits            426        733   +307   
- Misses          426        500    +74   
  Partials          0          0          

1. 2 files in src were modified. more
   • Misses -14
   • Hits +14
2. 4 files in R were modified. more
   • Misses -6
   • Hits +6

Powered by Codecov. Last updated by bbb035c...61735cd

[jennybc]

Is there an equivalent of XML::xmlNode(), i.e. a function to create a single new node de novo? I'm trying to translate something currently implemented with XML. I create a gazillion nodes as components of a list with purrr::pmap() and XML::xmlNode(), based on a data frame where each row provides info needed to populate one node. Then I add them to a (new) xml document en masse via XML::addChildren(kids = LIST_OF_GAZILLION_NODES).

It feels like xml2::xml_add_child() can play the role XML::addChildren(), once I figure out how to create my gazillion nodes as a nodeset (?). But how am I supposed to do that?

Am I overlooking some suitably wrapped version of the unexported function create_node()?

[jimhester]

@jennybc we removed support for creating free nodes, as it makes it impossible to assign a namespace to the nodes on creation.

For your use case can't you just pmap() directly with xml2::xml_add_child(), to create and add the nodes in one operation?

If you can point to your XML example code I can also take a look at how it would be converted.

[jimhester]

@jennybc Can you do something like the following?

``` r
mtcars$name <- gsub(" ", "-", row.names(mtcars))
d <- xml_add_child(xml_new_document(), "cars")
add_child <- function(...) {
  x <- purrr::map(list(...), as.character)

  purrr::invoke(xml_add_child,
    c(list(.x = d, .value = x$name), x[names(x) != 'name']))
}
purrr::pwalk(mtcars, add_child)
d
#> {xml_document}
#> <cars>
#>  [1] <Mazda-RX4 mpg="21" cyl="6" disp="160" hp="110" drat="3.9" wt="2.62 ...
#>  [2] <Mazda-RX4-Wag mpg="21" cyl="6" disp="160" hp="110" drat="3.9" wt=" ...
#>  [3] <Datsun-710 mpg="22.8" cyl="4" disp="108" hp="93" drat="3.85" wt="2 ...
#>  [4] <Hornet-4-Drive mpg="21.4" cyl="6" disp="258" hp="110" drat="3.08"  ...
#>  [5] <Hornet-Sportabout mpg="18.7" cyl="8" disp="360" hp="175" drat="3.1 ...
#>  [6] <Valiant mpg="18.1" cyl="6" disp="225" hp="105" drat="2.76" wt="3.4 ...
#>  [7] <Duster-360 mpg="14.3" cyl="8" disp="360" hp="245" drat="3.21" wt=" ...
#>  [8] <Merc-240D mpg="24.4" cyl="4" disp="146.7" hp="62" drat="3.69" wt=" ...
#>  [9] <Merc-230 mpg="22.8" cyl="4" disp="140.8" hp="95" drat="3.92" wt="3 ...
#> [10] <Merc-280 mpg="19.2" cyl="6" disp="167.6" hp="123" drat="3.92" wt=" ...
#> [11] <Merc-280C mpg="17.8" cyl="6" disp="167.6" hp="123" drat="3.92" wt= ...
#> [12] <Merc-450SE mpg="16.4" cyl="8" disp="275.8" hp="180" drat="3.07" wt ...
#> [13] <Merc-450SL mpg="17.3" cyl="8" disp="275.8" hp="180" drat="3.07" wt ...
#> [14] <Merc-450SLC mpg="15.2" cyl="8" disp="275.8" hp="180" drat="3.07" w ...
#> [15] <Cadillac-Fleetwood mpg="10.4" cyl="8" disp="472" hp="205" drat="2. ...
#> [16] <Lincoln-Continental mpg="10.4" cyl="8" disp="460" hp="215" drat="3 ...
#> [17] <Chrysler-Imperial mpg="14.7" cyl="8" disp="440" hp="230" drat="3.2 ...
#> [18] <Fiat-128 mpg="32.4" cyl="4" disp="78.7" hp="66" drat="4.08" wt="2. ...
#> [19] <Honda-Civic mpg="30.4" cyl="4" disp="75.7" hp="52" drat="4.93" wt= ...
#> [20] <Toyota-Corolla mpg="33.9" cyl="4" disp="71.1" hp="65" drat="4.22"  ...
#> ...

[jennybc]

I am trying to whittle my problem down to something I can either (a) solve or (b) inflict on you @jimhester.

[jennybc]

I choose (b) let @jimhester figure it out. Sorry this is not very minimal, but the examples above and the tests, etc., really don't get into the bits that feel hard about my actual problem. So I present in its full glory.

https://gist.github.com/jennybc/37398c989852aa0c7abd9c5cea75e4f6

There's an R script, corresponding md, input as csv, and XML output.

[jennybc]

Some features of the XML creation style of XML that seem to be missing in what's proposed for xml2:

• The ability to create free nodes, which is actually encouraged in XML. I liked that. Then again, I didn't like XML's memory leak problems and I suppose those are probably related 😬.
• The ability to add lots of children at once and to define children recursively.

I think anyone translating from XML code will need to know how they're supposed to do these things with xml2. Or what to do instead.