Fast and elegant XML generation for R
xmlwriter is an R package that provides a fast and simple interface for creating XML documents and fragments from R. It has a simple elegant syntax for creating xml_fragments.
xmlwriter’s XML generation from R lists is fast, implemented in C++ using Rcpp. Curious for the benchmarks? Check the performance section.
xmlwriter can be used as a companion to R packages XML or xml2 which are both wonderful packages optimized for parsing, querying and manipulating XML documents. Both XML and xml2 provide several ways for creating XML documents, but they are not optimized for generating and writing XML.
Creating XML documents with XML and xml2 can be a bit cumbersome, mostly because it
forces the author to manipulate the XML document tree, obscuring the XML structure of the document, and making it hard to read the XML that is being generated. xml2 does provide a way to create XML documents from R data structures using nested lists which is a powerful feature, but it is not optimized for speed or readability.
xmlwriter provides an intuitive interface for creating XML documents, that mimics how XML is written in a text editor.
It has two different ways to create XML documents:
- a light weight R syntax using
tag,frag,+,/anddata_frag, creating anxml_fragment, that can be easily translated into acharacterorxml2::xml_documentobject, or be used as a flexible building block for generating a larger XML document. - an
xmlbuilderobject that allows you to create XML documents in a feed-forward manner, withstartandendmethods, giving you more control on the XML document structure, including XML comment, prolog etc.
Installation
You can install the development version of xmlwriter from:w
GitHub with:
# install.packages("devtools")
devtools::install_github("edwindj/xmlwriter")Example
Using tag, frag, +, / and data_frag:
xml_fragments allow to write XML using readable R syntax. tag and frag are function that create XML elements that can be combined using + and / operators resulting in an xml_fragment.
xml_fragments can be used to create an character with valid XML, a xml2::xml_document or as a building block for more complex XML documents. An xml_fragment is a list object that is identical to the output of xml2::as_list, and can be converted to a character or an xml2::xml_document object but xmlwriter provides a much faster implementation (see performance).
tag is a function that creates a simple xml_fragment element with a given tag name, text, and attributes. It allows for creating elements with a tag name using a character, which gives flexibility.
library(xmlwriter)
# tag creates a simple xml_fragment, with text and named attributes
tag("person", "John Doe", id = 1, state="CA")
#> {xml_fragment}
#> <person id="1" state="CA">John Doe</person>frag is a function that allows for specifying nested elements and attributes, thus creating a more complex xml_fragment.
An argument to frag is either:
- a named element in which case the name is used as the tag name for that element. The element can be a value or a nested
frag. - an unnamed element in which case the element is added as a text node.
- a
.attrargument that is used to add attributes to the parent element.
# a frag can contain multiple elements
frag(
name = "John Doe",
age = 30
)
#> {xml_fragment (2)}
#> [1]<name>John Doe</name>
#> [2]<age>30</age>
#> ...
# and can nest frags
frag(
person = frag(
# attributes on person are specified using .attr
.attr = c(id = 1, state = "CA"),
name = "John Doe",
age = 30
)
)
#> {xml_fragment}
#> <person id="1" state="CA">
#> <name>John Doe</name>
#> <age>30</age>
#> </person>data_frag is function that converts a data.frame to an xml_fragment:
data <- data.frame(
id = c("p1", "p2"),
name = c("John Doe", "Jane Doe"),
age = c(30, 25),
stringsAsFactors = FALSE
)
# create an xml_fragment from a data.frame
data_frag(
.attr = data[c("id")],
data[c("name", "age")],
row_tag = "person"
)
#> {xml_fragment (2)}
#> [1]<person id="p1">
#> <name>John Doe</name>
#> <age>30</age>
#> </person>
#> [2]<person id="p2">
#> <name>Jane Doe</name>
#> <age>25</age>
#> </person>
#> ...Or you can use it within an xml_fragment:
# but you can also use it within an xml_fragment
doc <- xml_fragment(
homeless = data_frag(
.attr = data[c("id")],
data[c("name", "age")],
row_tag = "person"
)
)
doc<homeless>
<person id="p1">
<name>John Doe</name>
<age>30</age>
</person>
<person id="p2">
<name>Jane Doe</name>
<age>25</age>
</person>
</homeless>Combine fragments with +, append or c())
xml_fragments such as tag, frag and data_frag can be combined with the + operator, which is equivalent to the append() and c() function: it creates sibling XML nodes.
library(xmlwriter)
john <- tag("person", "John", id = 1)
jane <- tag("person", "Jane", id = 2)
john + jane
#> {xml_fragment (2)}
#> [1]<person id="1">John</person>
#> [2]<person id="2">Jane</person>
#> ...
john + tag("person", "Jane", id = 2)
#> {xml_fragment (2)}
#> [1]<person id="1">John</person>
#> [2]<person id="2">Jane</person>
#> ...
john + xml_fragment(
person = frag(
.attr = c(id = 2),
"Jane"
)
)
#> {xml_fragment (2)}
#> [1]<person id="1">John</person>
#> [2]<person id="2">Jane</person>
#> ...Add child fragments with ‘/’ or add_child_fragment()
- the
/operator, which is equivalent to theadd_child_fragmentfunction which creates a child XML node of the last XML node in anxml_fragment.
tag("person", id = 1) / (
tag("name", "John Doe")
)
#> {xml_fragment}
#> <person id="1">
#> <name>John Doe</name>
#> </person>
tag("person", id = 1) / frag(
name = "John Doe",
age = 30
)
#> {xml_fragment}
#> <person id="1">
#> <name>John Doe</name>
#> <age>30</age>
#> </person>
tag("person", id = 1) |>
add_child_fragment(
name = "John Doe",
age = 30
)
#> {xml_fragment}
#> <person id="1">
#> <name>John Doe</name>
#> <age>30</age>
#> </person>Flexible XML creation…
Using the tag, frag, + and / functions, one can create complex XML documents in a very flexible manner.
# overly complex, but shows flexible construction of XML
fragment <-
tag("person", id = "1") / # adds child nodes to person
( frag(
name = "John Doe",
age = 30
) + # add an extra child node to person, with subnodes
tag("address") |> add_child_fragment(
street = "123 Main St",
city = "Anytown",
state = "CA",
zip = 12345
)
)
fragment
#> {xml_fragment}
#> <person id="1">
#> <name>John Doe</name>
#> <age>30</age>
#> <address>
#> <street>...A cleaner version of the above:
xml2 compatibility
xml_writer does not have a hard dependency on xml2, but the output of xml_writer can be converted to a xml2::xml_document object. An xml_fragment is identical to the output of xml2::as_list, so it can be converted to a xml2::xml_document object.
xml_fragment supports the following xml2 methods:
-
xml2::as_xml_document, calls thexml2::read_xmlmethod -
xml2::as_list, removes thexml_fragmentclass -
xml2::write_xml, writes the xml to a file or console
One can also use the as_xml_nodeset function to convert the xml_fragment to a xml2::xml_nodeset object.
fragment |> xml2::as_xml_document()
#> {xml_document}
#> <person id="1">
#> [1] <name>John Doe</name>
#> [2] <age>30</age>
#> [3] <address>\n <street>123 Main St</street>\n <city>Anytown</city>\n <sta ...
fragment |> as_xml_nodeset()
#> {xml_nodeset (1)}
#> [1] <person id="1">\n <name>John Doe</name>\n <age>30</age>\n <address>\n ...xml_fragment implements the xml2::write_xml method
xml2::write_xml(fragment, file="") # print to the consoleresults in:
Performance of xml generation:
xmlwriter is optimized for generating xml documents and fragments from a R list structure that is identical to the xml2::as_list output.
library(microbenchmark)
library(xml2)
library(xmlwriter)
# read in a sample 600k xml file as a R list str
doc <- xml2::read_xml("./example/DataGeneric.xml")
doc_list <- xml2::as_list(doc)
# just copy the list and set an extra attribute class="xml_fragment"
# making it a xml_fragment object
doc_fragment <- structure(doc_list, class = "xml_fragment")
# see how long it takes to create an xml document with xml2 and xmlwriter
(m <- microbenchmark(
xml2 = xml2::as_xml_document(doc_list),
xmlwriter = xml2::as_xml_document(doc_fragment),
times = 10
))
#> Warning in microbenchmark(xml2 = xml2::as_xml_document(doc_list), xmlwriter =
#> xml2::as_xml_document(doc_fragment), : less accurate nanosecond times to avoid
#> potential integer overflows
#> Unit: milliseconds
#> expr min lq mean median uq max
#> xml2 2403.61996 2430.22547 2458.59214 2453.5764 2462.63728 2591.06093
#> xmlwriter 39.91612 41.27281 43.43863 42.7238 44.96974 49.10336
#> neval
#> 10
#> 10xmlwriter is about 56.6 times faster than xml2 for creating an xml document from an R list. Note that xmlwriter includes a round trip, since xmlwriter first generates a character vector which is then read using xml2::read_xml().
Using an xmlbuilder object
xmlbuilder is an object that allows you to create xml documents in a feed-forward manner.
With xml_fragment one creates the xml document structure as one (large) R list and then converts it to a xml string or xml2::xml_document, with xmlbuilder one incremently builds the xml document, without having the whole list structure in memory. It also provides more functions for the output xml document, like adding a prolog or comment.
It provides the following methods:
-
startto start a new element -
endto end the current element -
elementto add an element with a value -
commentto add a comment -
prologto add a prolog -
fragmentto add a fragment
library(xmlwriter)
b <- xmlbuilder(allow_fragments = FALSE)
b$comment("This is an xml comment")
b$start("homeless")
b$start("person", id = "1")
b$element("name", "John Doe")
b$element("age", 30)
b$start("address")
b$element("street", "123 Main St")
b$element("city", "Anytown")
b$element("state", "CA")
b$element("zip", 12345)
b$end("address")
b$end("person")
b$start("person", id = "2")
b$element("name", "Jane Doe")
b$element("age", 25)
b$start("address")
b$element("street", "321 Main St")
b$element("city", "Anytown")
b$element("state", "CA")
b$element("zip", 54321)
b$end("address")
b$end("person")
b$fragment(
person = frag(
.attr = c(id = "3"),
name = "Jim Doe",
age = 35
)
)
b$end("homeless")
# includes a xml prolog and comment
b
#> {xmlbuilder}
as.character(b)
#> [1] "<?xml version='1.0' encoding='UTF-8'?><!--This is an xml comment--><homeless><person id=\"1\"><name>John Doe</name><age>30</age><address><street>123 Main St</street><city>Anytown</city><state>CA</state><zip>12345</zip></address></person><person id=\"2\"><name>Jane Doe</name><age>25</age><address><street>321 Main St</street><city>Anytown</city><state>CA</state><zip>54321</zip></address></person><person id=\"3\">\n <name>Jim Doe</name>\n <age>35</age>\n</person></homeless>"
# only contains the actual nodes
xml2::as_xml_document(b)
#> {xml_document}
#> <homeless>
#> [1] <person id="1">\n <name>John Doe</name>\n <age>30</age>\n <address>\n ...
#> [2] <person id="2">\n <name>Jane Doe</name>\n <age>25</age>\n <address>\n ...
#> [3] <person id="3">\n <name>Jim Doe</name>\n <age>35</age>\n</person>