Application Programming Interfaces

Lecture 13

Dr. Benjamin Soltoff

Cornell University
INFO 2951 - Spring 2025

October 7, 2025

Announcements

Homework 4 due tomorrow
Project proposals due Thursday
No class Thursday
Dr. Soltoff’s office hours cancelled tomorrow

Learning objectives

Define application program interfaces (APIs)
Explain authentication keys and demonstrate secure methods for storing these keys
Demonstrate how to use canned packages in R to access APIs
Identify methods for writing functions to interact with APIs
Access APIs directly using the {httr2} package

Reading data into R

Local data files
Databases
Web scraping
Application programming interfaces (APIs)

Application programming interface (API)

Representational State Transfer (REST)
Uniform Resource Location (URL)
HTTP methods
- GET
- POST

RESTful queries

Submit request to server via URL
Return result in a structured format
Parse results into a local format

Install and play packages

Packages with R functions written for existing APIs

Useful because

Reproducible
Up-to-date (ideally)
Provenance (can blame someone if something goes wrong)
Ease of access

API authentication

How to verify that a user or device has permission to access an API?

Different methods include:

Username/password (out of favor)
Keys (e.g. GitHub and {usethis})
OAuth tokens (e.g. what {googlesheets4} tried to do)

API keys

Random string of alphanumeric characters unique to a device (and possibly a user)
Access restrictions based on needs
Obtain key
Store key securely

Never store directly in an visible R script or Quarto document

Store in .Rprofile or .Renviron and exclude these files from a public Git repo

Edit with usethis::edit_r_profile()

`.Rprofile`

options(this_is_my_key = "value")

R script

key <- getOption("this_is_my_key")

Edit with usethis::edit_r_environ()

`.Renviron`

this_is_my_key=value

R script

key <- Sys.getenv("this_is_my_key")

Census data with {tidycensus}

API to access data from US Census Bureau
- Decennial census
- American Community Survey
Returns tidy data frames with (optional) {sf} geometry
Search for variables with load_variables()

Store API key

library(tidycensus)

census_api_key("YOUR API KEY GOES HERE")

Obtain data

usa_inc <- get_acs(
  geography = "state",
  variables = c(medincome = "B19013_001"),
  year = 2023
)
usa_inc

# A tibble: 52 × 5
   GEOID NAME                 variable  estimate   moe
   <chr> <chr>                <chr>        <dbl> <dbl>
 1 01    Alabama              medincome    62027   400
 2 02    Alaska               medincome    89336  1374
 3 04    Arizona              medincome    76872   414
 4 05    Arkansas             medincome    58773   503
 5 06    California           medincome    96334   298
 6 08    Colorado             medincome    92470   483
 7 09    Connecticut          medincome    93760   669
 8 10    Delaware             medincome    82855  1234
 9 11    District of Columbia medincome   106287  1803
10 12    Florida              medincome    71711   282
# ℹ 42 more rows

Visualize data

Obtain geometry with {tidycensus}

Simple feature geometry with {sf}

tompkins <- get_acs(
  state = "NY",
  county = "Tompkins",
  geography = "tract",
  variables = c(medincome = "B19013_001"),
  year = 2023,
  geometry = TRUE
)

tompkins

Simple feature collection with 26 features and 5 fields
Geometry type: MULTIPOLYGON
Dimension:     XY
Bounding box:  xmin: -76.69666 ymin: 42.26298 xmax: -76.23782 ymax: 42.62742
Geodetic CRS:  NAD83
First 10 features:
         GEOID                                          NAME  variable estimate   moe
1  36109000100     Census Tract 1; Tompkins County; New York medincome    40861  5663
2  36109000300     Census Tract 3; Tompkins County; New York medincome       NA    NA
3  36109002200    Census Tract 22; Tompkins County; New York medincome    62006  8155
4  36109000500     Census Tract 5; Tompkins County; New York medincome    92335  6203
5  36109001200    Census Tract 12; Tompkins County; New York medincome       NA    NA
6  36109001301 Census Tract 13.01; Tompkins County; New York medincome    43883 18383
7  36109000400     Census Tract 4; Tompkins County; New York medincome    61250 17491
8  36109002300    Census Tract 23; Tompkins County; New York medincome    83774  9843
9  36109000700     Census Tract 7; Tompkins County; New York medincome    64783 13868
10 36109001500    Census Tract 15; Tompkins County; New York medincome    80082 22146
                         geometry
1  MULTIPOLYGON (((-76.50839 4...
2  MULTIPOLYGON (((-76.48981 4...
3  MULTIPOLYGON (((-76.40229 4...
4  MULTIPOLYGON (((-76.48412 4...
5  MULTIPOLYGON (((-76.49984 4...
6  MULTIPOLYGON (((-76.48902 4...
7  MULTIPOLYGON (((-76.48973 4...
8  MULTIPOLYGON (((-76.66654 4...
9  MULTIPOLYGON (((-76.51177 4...
10 MULTIPOLYGON (((-76.53789 4...

Visualize geometry

Generative AI models with {ellmer}

{ellmer}

Store your API key in .Renviron

library(ellmer)

chat <- chat_openai(
  model = "gpt-5-mini",
  system_prompt = "You are a friendly but terse assistant.",
)

chat$chat("Why is R a useful programming language?")

Short answer: R is built for statistics and data analysis, so it makes exploratory analysis, 
modeling, visualization, and reproducible reporting fast and convenient.

Key reasons it’s useful
- Rich statistical tooling: built-in and well-tested functions for tests, regression, time series, 
mixed models, etc.
- Vast package ecosystem (CRAN, Bioconductor): thousands of packages for specialized methods and 
domains (bioinformatics, econometrics, spatial data, etc.).
- Excellent visualization: ggplot2 and related packages make high-quality, customizable plots 
straightforward.
- Data wrangling made easier: tidyverse (dplyr, tidyr) simplifies cleaning and transforming data.
- Interactive / exploratory workflow: REPL, RStudio, and notebooks (R Markdown) support iterative 
analysis and reproducible reports.
- Reproducible research and reporting: R Markdown integrates code, output, narrative and can 
produce HTML/PDF/Word.
- Rapid prototyping of statistical models and experiments: concise syntax for models and convenient
diagnostics.
- Deployment and interactivity: Shiny for building interactive web apps from R analyses with 
minimal web dev.
- Integrations: call C/C++, Python, databases, and big-data tools; good plotting and reporting 
pipelines.
- Strong community and domain adoption: particularly in academia, bioinformatics, epidemiology, and
some parts of finance.

When R might not be ideal
- General-purpose application development (GUIs, large backend systems) — languages like Python, 
Java, or Go can be better.
- Very large-scale production systems may prefer ecosystems optimized for low-latency, concurrency,
or deployment standards.
- If you already have an established Python stack, interop is possible but may duplicate effort.

If you want, I can summarize how R compares to Python for data science or suggest starter packages 
and tools.

Application exercise

Writing an API function

Open Movie Database
No R package for API
Write your own function with {httr2}

`ae-11`

Instructions

Go to the course GitHub org and find your ae-11 (repo name will be suffixed with your GitHub name).
Clone the repo in Positron, run renv::restore() to install the required packages, open the Quarto document in the repo, and follow along and complete the exercises.
Render, commit, and push your edits by the AE deadline – end of the day

Wrap up

Recap

APIs offer a set of structured HTTP requests that return JSON or XML files
Use pre-written packages in R to access APIs when available
Use {httr2} to write your own API functions
Store API keys securely in .Rprofile or .Renviron