XML

#library(XML)
#library(RCurl)

xml_url <- "https://raw.githubusercontent.com/Michelebradley/DATA_607_HW/master/HW7/Books.xml"
books_xml <- getURL(xml_url)

XML_Books <- xmlParse(file = books_xml, validate=TRUE)
root <- xmlRoot(XML_Books)
xmlbookdf <- xmlToDataFrame(root)
xmlbookdf

https://stackoverflow.com/questions/16138693/r-rbind-multiple-data-sets

XML_List <- xmlToList(XML_Books)
Book1 <- as.data.frame(XML_List[1])
Book2 <- as.data.frame(XML_List[2])
Book3 <- as.data.frame(XML_List[3])

do.call("rbind", list(Book1, Book2, Book3))

JSON

#library(RJSONIO)
#library(stringr)
#library(dplyr)
#library(tidyr)

isValidJSON("https://raw.githubusercontent.com/Michelebradley/DATA_607_HW/master/HW7/Books.json")
## [1] TRUE
booksjson <- fromJSON(content="https://raw.githubusercontent.com/Michelebradley/DATA_607_HW/master/HW7/Books.json", simplify = FALSE)

https://stackoverflow.com/questions/43259380/spread-with-duplicate-identifiers-using-tidyverse-and

booksunlist <- sapply(booksjson[[1]], unlist)
booksjsondf <- as.data.frame(as.table(booksunlist))
Books_DF <- booksjsondf %>%
              select(-Var2) %>%
              group_by(Var1) %>%
              mutate(ind= row_number()) %>%
              spread(Var1, Freq)
Books_DF

HTML

#library(XML)
#library(RCurl)

html_url <- "https://raw.githubusercontent.com/Michelebradley/DATA_607_HW/master/HW7/Books.html"
books_html <- getURL(html_url)

https://stackoverflow.com/questions/4512465/what-is-the-most-efficient-way-to-cast-a-list-as-a-data-frame

html_df <- readHTMLTable(books_html, stringsAsFactors=FALSE)
# class(html_df) = "list"
do.call(rbind, lapply(html_df, data.frame, stringsAsFactors=FALSE))

Conclusions

The three dataframes are not identical. There are distinct differences between each file, (XML, JSON, and HTML) that make writing the files and importing them different.

The XML file was easy to write since it’s very readable. However, I did get a few DTD validation errors that provided an initial barrier into reading the data in. The co-author was found to be an issue, and the data structure of the XML datatable is not very pretty. But, I think if we didn’t have those technicalities, I would have preferred working with XML – xmltoDataFrame makes things easy.

The JSON file was also readable, but a little less so than the XML file. I did like the fact that there weren’t any encodings or validation errors that we had to work around. Reading the file in initially, I didn’t include a co-author for each book, but then I started having difficulty reading the data in without it, so I included a Co-Author: NA to books with one author. Worked great making it a dataframe.

HTML was the most different from XML and JSON. The text was very different to write, but after getting over the inital difference, it was really easy to create a table using HTML. Then, you could read the HTML table using readHTMLTable. Unfortunately, that came in as a list, so I had to do some digging online so as to make the list into a dataframe. There’s also this NULL value that got attached somewhere, I’m not sure how.