samuelbharti · Oct 10, 2026 · Sep 22, 2026
diff --git a/.github/pull_request_template.md b/.github/pull_request_template.md
@@ -31,4 +31,6 @@
 
      Anything that calls NCBI goes through net() or skip_without_ncbi(), so a
      bad day at NCBI skips instead of failing the build, while a real bug still
-     fails. See tests/testthat/helper-ncbi.R. -->
+     fails. See tests/testthat/helper-ncbi.R. A test file that calls NCBI also
+     starts with skip_on_cran(), so CRAN and r-universe run only the tests
+     that need no network. -->
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -24,7 +24,7 @@ Imports:
     httr (>= 0.5),
     jsonlite (>= 0.9)
 Suggests:
-    testthat,
+    testthat (>= 3.2.3),
     knitr,
     rmarkdown
 URL: https://docs.ropensci.org/rentrez/, https://github.com/ropensci/rentrez/

diff --git a/tests/test-all.R b/tests/test-all.R
@@ -1,10 +1,7 @@
 library("testthat")
 
-#All of the tests rely on the API existing and behaving as documented. However,
-#the API occasionally falls over or stops working which lets to errors on CRAN.
-#Because we use travis CI we will hear about any test failures as soon as they
-#happen. So, let's skill all tests on CRAN:
-
-if(identical(Sys.getenv("NOT_CRAN"), "true")){
-    test_check("rentrez")
-}
+#Most test files call the live NCBI API and start with skip_on_cran(), so CRAN
+#and r-universe, which leave NOT_CRAN unset, run only the tests that need no
+#network. devtools::check() and the r-lib GitHub Actions set NOT_CRAN=true and
+#run the whole suite (#221).
+test_check("rentrez")
diff --git a/tests/testthat/test_citmatch.r b/tests/testthat/test_citmatch.r
@@ -1,4 +1,5 @@
 context("Cite matching")
+skip_on_cran()
 test_that("Citation matching works",{
      ex_cites <- c("proc natl acad sci u s a|1991|88|3248|mann bj|test1|",
                    "science|1987|235|182|palmenberg ac|test2|")

diff --git a/tests/testthat/test_docs.r b/tests/testthat/test_docs.r
@@ -2,6 +2,7 @@
 # suite. Note, the final example of the README makes a lot calls to NCBI, so is
 # not included here
 context("documentation")
+skip_on_cran()
 
 test_that("The pubmed example in the docs works", {
     hox_paper <- net(entrez_search(db="pubmed", term="10.1038/nature08789[doi]"))

diff --git a/tests/testthat/test_errors.r b/tests/testthat/test_errors.r
@@ -165,3 +165,10 @@ test_that("the parse gate turns away rettypes that are not XML", {
         expect_false(rentrez:::is_xml_record(rt, ""))
     }
 })
+
+# NCBI serves version 1.0 summaries only as XML, so entrez_summary() refuses a
+# json request for one before it sends anything. No network.
+test_that("version 1.0 summaries cannot be requested as json", {
+    expect_error(entrez_summary("pubmed", id = 1:10, version = "1.0", retmode = "json"),
+                 "only available as xml")
+})
diff --git a/tests/testthat/test_fetch.r b/tests/testthat/test_fetch.r
@@ -1,4 +1,5 @@
 context("fetching records")
+skip_on_cran()
 
 
 pop_ids = c("307082412", "307075396", "307075338", "307075274")

diff --git a/tests/testthat/test_httr.r b/tests/testthat/test_httr.r
@@ -1,4 +1,5 @@
 context("httr option passing")
+skip_on_cran()
     #most config options  don't produce capture-able output, so instead
     # we will test if we raise an error when we us a non-existant proxy to
     # connect to the internet

diff --git a/tests/testthat/test_httr_post.r b/tests/testthat/test_httr_post.r
@@ -1,4 +1,5 @@
 context("POST (the HTTP verb)")
+skip_on_cran()
 
 #setup (guarded so transient NCBI problems skip rather than abort the file)
 ncbi_ok <- tryCatch({

diff --git a/tests/testthat/test_info.r b/tests/testthat/test_info.r
@@ -1,4 +1,5 @@
 context("einfo functions")
+skip_on_cran()
 
 #setup (guarded so transient NCBI problems skip rather than abort the file)
 ncbi_ok <- tryCatch({

diff --git a/tests/testthat/test_link.r b/tests/testthat/test_link.r
@@ -1,4 +1,5 @@
 context("elink")
+skip_on_cran()
 
 #setup (guarded so transient NCBI problems skip rather than abort the file)
 message("(this may take some time, have to download many records)")

diff --git a/tests/testthat/test_link_byid.r b/tests/testthat/test_link_byid.r
@@ -3,6 +3,7 @@ context("elink by_id")
 # In by_id mode the result should always be a list (one elink per id), even for
 # a single id, and the call should not emit a spurious "invalid id" warning.
 test_that("by_id=TRUE returns a one-element list with no spurious warning", {
+    skip_on_cran()
     #only the call is wrapped. A testthat failure is an error condition too, so
     #an expectation inside this tryCatch would be reported as a skip instead.
     seen <- NULL

diff --git a/tests/testthat/test_net.r b/tests/testthat/test_net.r
@@ -1,4 +1,5 @@
 context("Network")
+skip_on_cran()
 test_that("The NCBI is contactable from this comptuter /",{
     response <- net(httr::GET("https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi"))
     skip_if(response$status_code >= 400, "NCBI returned an error status")

diff --git a/tests/testthat/test_parse.r b/tests/testthat/test_parse.r
@@ -1,4 +1,5 @@
 context("result-parsers")
+skip_on_cran()
 
 #setup (guarded so transient NCBI problems skip rather than abort the file)
 ncbi_ok <- tryCatch({

diff --git a/tests/testthat/test_post.r b/tests/testthat/test_post.r
@@ -1,4 +1,5 @@
 context("entrez_post")
+skip_on_cran()
 
 prot_ids = c(15718680,157427902)
 

diff --git a/tests/testthat/test_query.r b/tests/testthat/test_query.r
@@ -1,4 +1,5 @@
 context("query")
+skip_on_cran()
 test_that("Query building functions work", {
 
     setup <- tryCatch({

diff --git a/tests/testthat/test_search.r b/tests/testthat/test_search.r
@@ -1,4 +1,5 @@
 context("search")
+skip_on_cran()
 
 #setup (guarded so transient NCBI problems skip rather than abort the file)
 ncbi_ok <- tryCatch({

diff --git a/tests/testthat/test_summary.r b/tests/testthat/test_summary.r
@@ -1,4 +1,5 @@
 context("fetching and parsing summary recs")
+skip_on_cran()
 
 fake_ids <- sample(1e5, 501)
 pop_ids = c("307082412", "307075396", "307075338", "307075274")
@@ -66,12 +67,6 @@ test_that("JSON and XML objects are similar", {
 })
 
 
-test_that("Error whent tring to fetch 1.0 summaries as json", {
-      expect_error(
-        entrez_summary("pubmed", id = fake_ids[1:10], version="1.0", retmode="json")
-      )
-})
-
 test_that("We can print summary records", {
       skip_without_ncbi(ncbi_ok)
       expect_output(print(pop_summ_json), "List of  4 esummary records")

diff --git a/tests/testthat/test_webenv.r b/tests/testthat/test_webenv.r
@@ -1,4 +1,5 @@
 context("WebEnv")
+skip_on_cran()
 test_that("Searches using WebEnv features work", {
     #setup
     web_env_search <- net(entrez_search(db="nuccore",

diff --git a/tests/testthat/test_zz_reachable.r b/tests/testthat/test_zz_reachable.r
@@ -1,4 +1,5 @@
 context("live API reachable")
+skip_on_cran()
 
 # The rest of the suite skips its tests when NCBI is unavailable. That keeps a
 # bad day at NCBI from failing the build, but it also means a run in which every