Learn

R was designed to ingest, manipulate, and emit data. The IO functions are battle-tested.

read.csv / write.csv:

df <- read.csv("data.csv")
write.csv(df, "out.csv", row.names = FALSE)

Key args:

• header = TRUE — first row is column names (default)
• stringsAsFactors = FALSE — don't auto-convert strings to factors (default in R 4.0+)
• na.strings = c("", "NA", "N/A") — values to treat as NA
• sep = "," — field separator (use \t for TSV)
• colClasses = c("character", "numeric", "Date") — explicit types

read.table is the parent — more flexible:

df <- read.table("data.tsv", header = TRUE, sep = "\t")

readr package (tidyverse) is faster and infers types better:

library(readr)
df <- read_csv("data.csv")        # tibble, no factors
write_csv(df, "out.csv")

readLines for raw text:

lines <- readLines("file.txt", warn = FALSE)
for (line in lines) cat(toupper(line), "\n")

Use "stdin" to read from standard input — that's what we've been doing.

scan for fast numeric reading:

nums <- scan("file.txt", what = numeric())

Much faster than read.csv for plain numeric data.

JSON with jsonlite:

library(jsonlite)
data <- fromJSON("data.json")
writeJSON(data, "out.json", pretty = TRUE)

Handles nested structures, simplifies arrays-of-objects to data frames.

Excel with readxl:

library(readxl)
df <- read_excel("file.xlsx", sheet = 1)

Database with DBI + driver package (RSQLite, RPostgres, etc.):

library(DBI)
con <- dbConnect(RSQLite::SQLite(), "db.sqlite")
df <- dbGetQuery(con, "SELECT * FROM users WHERE active = 1")
dbDisconnect(con)

The DBI interface is consistent across drivers — switch from SQLite to PostgreSQL just by changing the driver.

Practical patterns:

• Always check nrow(df) and summary(df) after reading
• head(df) and str(df) show structure quickly
• For huge files: data.table::fread is dramatically faster than read.csv
• For streaming: read in chunks with nrows argument or use readr::read_csv_chunked