Basic use of the {officer} R package to extract a table from a Word document

officer-scrape-word.R

# Basic use of the {officer} package to scrape a table from a Word doc
# https://davidgohel.github.io/officer/

# I've now written a couple of functions to do this task:
#  https://gist.github.com/matt-dray/d4837f106bcee80ea39235b6465a7cac
# You can read more about those in a blog post:
#   https://www.rostrum.blog/2023/06/07/rectangular-officer/

# There are other solutions. You can also try {docxtractr} by Bon Rudis
# (on CRAN), which doesn't depend on {officer}, or {officerExtras} by Eli
# Pousson (on GitHub).

# Attach package
library(officer)

# Create a test docx file
doc_test <- read_docx() |>
  body_add_par("Hello world!", style = "heading 1") |>
  body_add_par("Below is a table", style = "Normal") |>
  body_add_table(mtcars[1:3, 1:5]) |> 
  body_add_par("Below is another table", style = "Normal") |>
  body_add_table(airquality[1:3, 1:5])

# Save docx to temp location
temp_docx <- tempfile(fileext = ".docx")
print(doc_test, target = temp_docx)

# Read the file from temp path
doc_path <- list.files(tempdir(), pattern = ".docx$", full.names = TRUE)
doc_in <- read_docx(doc_path)

# Get the content of the document as a dataframe
content <- docx_summary(doc_in)

# Filter the table for a specific table
table_cells <- subset(content, content_type %in% "table cell" & doc_index == 3)

# The content of the table is in 'long' format, but we can 're-rectangularise'
table_names <- table_cells[table_cells$is_header, "text"]
table_content <- table_cells[!table_cells$is_header, "text"]
row_count <- nrow(table_cells) / nrow(table_cells[table_cells$is_header, ]) - 1
table_mat <- as.data.frame(matrix(table_content, nrow = row_count))
names(table_mat) <- table_names
table_mat
##    mpg cyl disp  hp drat
## 1 21.0   6  160 110 3.90
## 2 21.0   6  160 110 3.90
## 3 22.8   4  108  93 3.85