Parity checklist

This page is a translation guide for the operational parity between the R and Python versions of csiapps. Each row is a capability; the R and Python columns give the function(s) that provide it. An empty cell marks a capability that exists in only one language (see the notes below for why).

The two packages have full feature parity — every capability is available in both — with two exceptions that are intentional and explained under Intentional divergences.

Configuration

Capability	R	Python
Set the target institute	`set_institute()`	`set_institute()`
Check whether sandbox mode is on	`is_sandbox_mode()`	`is_sandbox_mode()`
Force sandbox mode on/off	`options(csiapps.sandbox = )`	`set_sandbox_mode()`
Check auth environment / secrets	`check_secrets()`	`check_secrets()`

REST API client

Capability	R	Python
Authenticated request to any endpoint	`make_request()`	`make_request()`
Fetch organisation options	`fetch_org_options()`	`fetch_org_options()`
Fetch profiles (auto-paginating)	`fetch_profiles()`	`fetch_profiles()`
Fetch a single profile by id	`fetch_profile()`	`fetch_profile()`
Flatten a profile into a table row		`flatten_profile()`

Sandbox

Capability	R	Python
Register a JSON schema for a source	`register_sandbox_schema()`	`register_sandbox_schema()`
Create a dummy sport organization	`create_sport_org()`	`create_sport_org()`
Create dummy athlete profiles	`create_profile()`	`create_profile()`
Clear sandbox state	`clear_sandbox()`	`clear_sandbox()`
Open the sandbox payload folder	`browse_sandbox()`	`browse_sandbox()`

Shiny app wrappers

Capability	R	Python
Wrap app UI (navbar/footer/auth status)	`ui_wrapper()`	`ui_wrapper()`
Wrap app server (OAuth2 / sandbox login)	`server_wrapper()`	`server_wrapper()`
Reactive login guard	`token_ready()`	`token_ready()`
Make globals visible to the server	`global_wrapper()`

Intentional divergences

These are the only differences in the public surface, and each exists for a concrete reason:

global_wrapper() is R-only. It evaluates a code block in R's global environment so top-level objects are reachable from the server function — an R Shiny scoping concern. In Shiny for Python a module-level assignment is already captured by the server closure, so no equivalent is needed.
flatten_profile() is Python-only. Shiny for Python's render.data_frame rejects the nested dicts fetch_profiles() returns, so this helper flattens one profile into scalar fields for a table. R users flatten inline with data.frame() / do.call(rbind, ...), so no dedicated helper ships in R.

Beyond the public surface, a few behaviours differ by language idiom or as a deliberate hardening — same capability, adapted shape:

Sandbox toggle mechanism. R uses the csiapps.sandbox option (options(csiapps.sandbox = FALSE)); Python has no options system, so it uses the set_sandbox_mode() function. Both also respect CSIAPPS_ENV=production.
fetch_org_options() return shape. R returns label/value pairs for selectInput(); Python returns a {value: label} dict for ui.input_select(choices = ). Same data, framework-native shape.
fetch_profile() argument order. R is fetch_profile(token, profile_id); Python puts the required argument first: fetch_profile(profile_id, token). Equivalent when called by keyword.
Schema validator engine. Sandbox ingestion validates with Ajv (via jsonvalidate) in R and jsonschema (Draft 7) in Python. Error wording and some format-keyword edge cases differ.
fetch_profiles() pagination guard. The Python port adds a max_pages bound with cycle detection and a truncation warning; the R version paginates until the API stops returning a next link. Python is a hardened superset.
Per-session token storage. Both packages store the logged-in user's access token per session (never a process-global) and resolve it reactively, so make_request() / fetch_*() gate themselves until login completes and re-run automatically once it does. R keeps the token in a reactiveVal on session$userData; Python keeps it in a reactive.value keyed on the session. R additionally seeds the CSIAPPS_ACCESS_TOKEN environment variable with a non-secret placeholder so older apps that guard on nzchar(Sys.getenv("CSIAPPS_ACCESS_TOKEN")) keep working; Python needs no such shim, as it never wrote the token to the environment. See Waiting for login.

Function reference

R: R reference (pkgdown)
Python: Python reference (on this site)