Restore NAMESPACE exports wiped by the v1.2.5 regenerate commit

[vertesy]

Restore NAMESPACE exports wiped by the v1.2.5 "regenerate package files" commit

Commit b164f79 ("Bump to v1.2.5 and regenerate package files") reduced
NAMESPACE from 84 export()/exportPattern lines down to just the
"Generated by roxygen2" header comment - every single function in the
package stopped being exported, even though all ~80 functions in
R/Stringendo.R still carry their @export roxygen tag untouched. Nothing
in the source changed; NAMESPACE was left empty by whatever ran that day
(most likely roxygen2::document() erroring out partway through, or being
run against a build directory that failed to load, and silently writing
an incomplete file instead of failing loudly).

Impact: this doesn't just affect Stringendo's own users calling
library(Stringendo) - it's a hard, immediate crash for every downstream
@vertesy package that Depends on or Imports from Stringendo. Concretely:
CodeAndRoll2 (Depends: Stringendo) does
importFrom(Stringendo, percentage_formatter) in its own NAMESPACE; with
that export gone, requireNamespace("CodeAndRoll2") itself fails with
"object 'percentage_formatter' is not exported by 'namespace:Stringendo'"

• so CodeAndRoll2 cannot load, which means UVI.tools (Depends:
  CodeAndRoll2) cannot load either. Confirmed this is not an isolated
  example: the diff below restores 82 export() lines that were present
  before b164f79 and match the @export tags still in R/Stringendo.R today.

Fix: ran devtools::document(roclets = c("namespace", "rd")) against the
current source (no source changes were needed or made - only NAMESPACE
was regenerated).

Verification:

• getNamespaceExports("Stringendo") includes percentage_formatter
  again after installing this fix locally
• Confirmed the full downstream chain: requireNamespace("CodeAndRoll2")
  now succeeds, and devtools::load_all() on UVI.tools (which Depends on
  CodeAndRoll2) now succeeds too - both were failing before this fix
• devtools::check() run on this branch: 1 ERROR, 3 WARNINGs, 1 NOTE, none
  of which are caused by this change or touch NAMESPACE/exports - the
  ERROR is a pre-existing, unrelated bug in the deprecated parFlags()
  ("object 'namez' not found"), and the WARNINGs/NOTE are pre-existing
  Rd/documentation issues (%!in%'s Rd \name, a stale idate.Rd
  cross-reference, an undocumented parsepvalue() argument, and some
  "no visible global variable" NOTEs) that predate this commit and are
  unrelated to exports
• lintr run on the changed file

Note: there are several other open PRs against this repo (#43-#50)
addressing narrower, unrelated bugs in individual functions - this PR is
scoped only to restoring the wiped NAMESPACE and doesn't overlap with
any of them.

Co-Authored-By: Claude Sonnet 5 noreply@anthropic.com

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0e8b2a86b8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Repair generation before restoring NAMESPACE

Problem: This commit restores only the generated NAMESPACE; it does not fix the generation path that erased every export in b164f79. Trigger: The next run of PackageTools::document_and_create_package() in Development/Create_the_Stringendo_Package.R can overwrite these entries again, leaving all public Stringendo functions unavailable to dependent packages. Fix: Correct the upstream generation step, then regenerate and commit NAMESPACE from the roxygen annotations.

AGENTS.md reference: AGENTS.md:L34-L38

Useful? React with 👍 / 👎.

[vertesy]

@claude
could you please address if u agree"

1
address comment:
" Repair generation before restoring NAMESPACE
Problem: This commit restores only the generated NAMESPACE; it does not fix the generation path that erased every export in b164f79. Trigger: The next run of PackageTools::document_and_create_package() in Development/Create_the_Stringendo_Package.R can overwrite these entries again, leaving all public Stringendo functions unavailable to dependent packages.
Fix: Correct the upstream generation step, then regenerate and commit NAMESPACE from the roxygen annotations."

2
review The PR is still relevant since many things have changed in the repository.

[claude]

Claude finished @vertesy's task in 3m 17s —— View job

---

Summary

• Gather context (PR diff, AGENTS.md, generation script, current dev/main state)
• Address Codex review comment: add a fail-loud safeguard in the build script
• Check whether the PR is still relevant given repo changes
• Push changes

---

1. Codex comment — "Repair generation before restoring NAMESPACE"

I agree with the underlying concern, but with a scope caveat:

• The actual generation call, PackageTools::document_and_create_package(), lives in the PackageTools package — a separate repo, not part of Stringendo. It isn't checked out here, so I can't inspect or fix its internals from this PR.
• What is in this repo is the call site: Development/Create_the_Stringendo_Package.R, which invokes that function twice (lines 38 and 112) with no check afterward — so a silent partial failure (the actual root cause per the PR body: roxygen2::document() erroring out or running against a build dir that failed to load) just writes an empty NAMESPACE and the script carries on to git add commit push.

Since I can't touch the upstream package, I added the next best thing directly in this repo: a stopifnot() guard right after each document_and_create_package() call that fails loudly if the regenerated NAMESPACE has zero export() lines, e.g.:

PackageTools::document_and_create_package(repository.dir, config_file = 'config.R')
namespace.lines <- readLines(file.path(repository.dir, "NAMESPACE"))
stopifnot("NAMESPACE regeneration produced no exports (roxygen2::document() likely failed silently) - fix before committing" = any(grepl("^export", namespace.lines)))

This directly targets the "Trigger" Codex flagged — the next run of that script can no longer silently overwrite all exports; it will stop before the commit step. Pushed as 6822e04.

(Note: I could not run lintr or R itself in this sandbox — no R installation is available here — so this wasn't linted in this session. Please double check syntax before merging.)

2. Is this PR still relevant?

I checked the current tips of both target branches (fetched as part of the checkout, timestamps ~2026-08-30 22:4x, i.e. very recent):

• origin/dev (this PR's actual base): NAMESPACE already has 84 lines, and it is byte-identical to what this PR's branch contains. Whatever fixed it there, dev is currently not broken.
• origin/main: NAMESPACE is still just the 2-line roxygen2 header — i.e. main still has the wiped/empty NAMESPACE, same broken state as b164f79.

So: as opened (base = dev), this PR's diff would currently be a no-op against dev — the fix already landed there some other way. The real remaining gap is on main, not dev. Worth deciding explicitly:

• if dev → main gets released/merged normally, this PR may simply be safe to close as superseded, or
• if main needs an immediate, independent fix (e.g. it's what gets released to CRAN/users directly), this PR should probably be retargeted at main instead of dev.

I did not close or retarget the PR myself — that's a call for you to make, and retargeting/closing isn't something I can do here.

Branch: fix-empty-namespace-export-wipe