Remove unread inst/extdata/context_kotl.gpkg (4.9 MB)

data-raw/example_context_fwcp_peace.R

@@ -1,6 +1,6 @@
 # Fetch spatial context layers for FWCP Peace vignette from fwapg
 #
-# Same recipe as example_context_kotl.R but tuned for the much larger
+# Same recipe as example_context_kootenay_lake.R but tuned for the much larger
 # Peace region: bigger lake threshold, higher minimum stream order to
 # keep map density reasonable, town list relevant to the Peace.
 #

planning/archive/2026-10-issue-112-context-kotl/README.md

@@ -0,0 +1,14 @@
+## Outcome
+
+`inst/extdata/context_kotl.gpkg` shipped in every install with no reader. Its only consumer, the KOTL vignette, was removed in #50, and `context_kootenay_lake.gpkg` (#57) holds the same five layer kinds plus wsgs and ecoregions over an area that contains KOTL. The file was deleted. Its producer `data-raw/example_context_kotl.R` was deleted too, rather than marked superseded, so nothing can recreate it; git history keeps the recipe. `example_aoi_kotl.gpkg` and its producer stay because the README quick-start reads them. Three code-check rounds came back Clean: code readers, including an org-wide `gh` code search; the build/publish file set; and the supersession claim and doc truth. One leftover: the two remaining `data-raw/example_context_*` headers now cite each other as "same recipe". Both reviewers who noticed judged it harmless.
+
+## Measurement
+
+R 4.5.2. `R CMD build --no-build-vignettes --no-manual` on `git archive` of `048f93a` vs the change, each installed with `R CMD INSTALL -l` into a scratch library:
+
+- **Tarball:** 7,146,394 → 3,830,880 B (−46%), 101 → 100 files.
+- **Installed `cd/`:** 12,776 → 7,948 KB (−38%).
+
+The issue's "loses 4.9 MB" holds for installed size. The gpkg compresses, so the download shrinks by 3.3 MB. After-tarball `R CMD check --no-manual --ignore-vignettes`: Status OK, 0 NOTEs. `devtools::test()`: PASS 436, FAIL 0. The reproduce commands are in `findings.md`.
+
+Closed by: commit 1d17163 / PR #115

planning/archive/2026-10-issue-112-context-kotl/findings.md

@@ -0,0 +1,46 @@
+# Findings — inst/extdata/context_kotl.gpkg (4.9 MB) ships but nothing reads it (#112)
+
+## Issue context
+
+**If we do it:** the installed package loses 4.9 MB, about 40% of its 12.5 MB installed size. **If we never do:** every install carries a file nothing reads.
+
+## Problem
+
+`inst/extdata/context_kotl.gpkg` is 4.9 MB. Nothing in `R/`, `tests/`, `vignettes/` or `README.md` references it. The only mention is its producer, `data-raw/example_context_kotl.R`, plus a comment in `data-raw/example_context_fwcp_peace.R`. It dates from `3ece27c` (2026-04-07, before the Kootenay Lake vignette). The vignette added in #57 uses `context_kootenay_lake.gpkg`, which appears to supersede it. `example_aoi_kotl.gpkg` is still used by the README example, so keep it.
+
+Found by a code-check reviewer while verifying #100's tarball contents, and confirmed with `git grep -n context_kotl`.
+
+## Proposed Solution
+
+- Confirm `context_kootenay_lake.gpkg` supersedes it. Then delete `inst/extdata/context_kotl.gpkg`, and either delete `data-raw/example_context_kotl.R` or mark it superseded.
+- Rebuild and compare the tarball size before and after.
+
+## Supersession (plan-mode exploration, 2026-10-02)
+
+- `git grep context_kotl` outside `planning/` hit only the producer and one comment in `data-raw/example_context_fwcp_peace.R`.
+- The file's consumer, the KOTL vignette, was removed in #50 (NEWS.md, v0.2.x entry). That entry's "KOTL polygon assets stay" refers to `example_aoi_kotl.gpkg`, used at README.md:23 and kept.
+- `context_kootenay_lake.gpkg` (#57) layers: lakes, rivers, streams, highways, wsgs, ecoregions, towns over KOTL+LARL+DUNC+SLOC. `context_kotl.gpkg` layers: towns, lakes, rivers, streams, highways over KOTL only. Superset.
+- Code-check round 1 widened this: no `list.files()` over `inst/extdata`, no pattern-built filenames, and an org-wide `gh` code search found no other repo reading the file.
+
+## Measurement
+
+R 4.5.2, `R CMD build --no-build-vignettes --no-manual` on `git archive` of HEAD (`048f93a`) vs the staged tree (`git write-tree`), then `R CMD INSTALL -l` into a scratch library:
+
+| | before | after | change |
+|---|---|---|---|
+| tarball | 7,146,394 B | 3,830,880 B | −3.3 MB (−46%) |
+| tarball files | 101 | 100 | only `inst/extdata/context_kotl.gpkg` gone |
+| installed `cd/` | 12,776 KB | 7,948 KB | −4.7 MB (−38%) |
+| installed `extdata/` | 11,796 KB | 6,968 KB | −4.7 MB |
+
+The gpkg compresses in the tarball (4.9 MB raw → ~3.3 MB of tarball), so the issue's "loses 4.9 MB" holds for the installed size, not the download.
+
+After-tarball `_R_CHECK_FORCE_SUGGESTS_=false R CMD check --no-manual --ignore-vignettes`: **Status: OK** (0 NOTEs; #111 had already cleared the last two). `devtools::test()`: FAIL 0 | WARN 6 | SKIP 0 | PASS 436. `system.file("extdata", "example_aoi_kotl.gpkg", package = "cd")` resolves in the installed after-library and reads 1 feature.
+
+Reproduce (per side; `<tree>` is `HEAD` or `$(git write-tree)`):
+
+```bash
+git archive <tree> | tar -x -C src && R CMD build --no-build-vignettes --no-manual src
+wc -c < cd_0.5.8.tar.gz; tar -tzf cd_0.5.8.tar.gz | wc -l
+R CMD INSTALL -l lib cd_0.5.8.tar.gz && du -sk lib/cd lib/cd/extdata
+```

planning/archive/2026-10-issue-112-context-kotl/progress.md

@@ -0,0 +1,13 @@
+# Progress — inst/extdata/context_kotl.gpkg (4.9 MB) ships but nothing reads it (#112)
+
+## Session 2026-10-02
+
+- Plan-mode exploration — phases approved by user (delete producer script too, not mark superseded)
+- Created branch `112-inst-extdata-context-kotl-gpkg-4-9-mb-shi` off main
+- Scaffolded PWF baseline from issue #112 with approved phases
+- Next: start Phase 1
+- Phase 1: baseline tarball 7,146,394 B / installed 12,776 KB from `git archive HEAD`
+- Phase 2: `git rm` of `inst/extdata/context_kotl.gpkg` + `data-raw/example_context_kotl.R`; repointed `example_context_fwcp_peace.R:3` to `example_context_kootenay_lake.R`; `git grep context_kotl` outside planning/ empty
+- Phase 3: after tarball 3,830,880 B / installed 7,948 KB; R CMD check Status OK; `devtools::test()` PASS 436 FAIL 0; README AOI resolves (numbers in findings.md)
+- /code-check: 3 rounds (code readers, build/publish file set, supersession + doc truth), all Clean — `review-round{1,2,3}.md`
+- Next: /planning-archive, /gh-pr-push

planning/archive/2026-10-issue-112-context-kotl/review-round1.md

@@ -0,0 +1,14 @@
+# Code-check review round 1 — #112 (staged diff)
+
+## Clean
+No issues found.
+
+### What was verified
+- `git grep -i kotl` outside `planning/archive/`: no reader of `context_kotl` in `R/`, `tests/`, `vignettes/`, `scripts/`, `.github/`, `README.md`, `_pkgdown.yml`, `DESCRIPTION`, `.Rbuildignore`, or other `inst/` files. Remaining `kotl` hits are `example_aoi_kotl.gpkg` (README.md:23, kept), its kept producer `data-raw/example_aoi_kotl.R`, WSG code strings, and NEWS history.
+- Indirect readers: every `list.files()` in `R/` and `scripts/` targets cache / grib / cog / monthly dirs, never `inst/extdata`; all `system.file("extdata", ...)` calls use literal names of files that still exist; no `paste0("context_", ...)` style construction anywhere.
+- Ecosystem: local grep over `~/Projects/repo` and `gh` code search (`org:NewGraphEnvironment context_kotl`, default branches) find only cd's own producer, the edited comment, and planning/archive + sred evidence dumps (history). No external package reads `system.file(..., "context_kotl.gpkg", package = "cd")`.
+- Edited comment (`data-raw/example_context_fwcp_peace.R:3`): `example_context_kootenay_lake.R` exists, uses the same recipe (frs_db_query towns/lakes/rivers/streams/highways/wsgs/ecoregions, same st_write layering), and its thresholds (lakes > 200 ha, stream order >= 5) are lower than Peace's (> 1000 ha, >= 7), so "bigger lake threshold, higher minimum stream order" remains accurate. That file's header reciprocally says "Same recipe as example_context_fwcp_peace.R".
+- Deleting the producer removes the only writer of `inst/extdata/context_kotl.gpkg`; nothing left can regenerate the deleted file.
+
+### Pre-existing, not introduced by this diff (FYI only)
+- `data-raw/example_context_fwcp_peace.R:8` lists output layers as "towns, lakes, streams, highways" but the script also writes rivers, wsgs, ecoregions. Untouched line; not a failure.

planning/archive/2026-10-issue-112-context-kotl/review-round2.md

@@ -0,0 +1,15 @@
+# Review round 2 — #112 (build/publish side)
+
+## Clean
+No issues found.
+
+Checked:
+- `.github/workflows/pkgdown.yaml`: no step copies, size-checks, globs or uploads `inst/extdata`. The allowed-pages gate looks at root markdown pages, not extdata.
+- `.github/workflows/climate-update.yml`, `update-citation-cff.yaml`: no extdata references.
+- Vignettes: `kootenay-lake.Rmd` loads `example_aoi_kootenay_lake.gpkg` and `context_kootenay_lake.gpkg`, and `peace-fwcp.Rmd` loads the fwcp_peace files and the commentary CSV. Both are still present. Neither vignette uses `kotl`.
+- `R/cd_stac_catalog.R:37` lists only `\.tif$` in `cog_dir`. The `tests/testthat/test-cd_stac_catalog.R` expectations are `>= 1` row/item checks over `example_climate.tif`, so removing a gpkg cannot change them.
+- `list.files` / `dir_ls` across R/, tests/, vignettes/, data-raw/ and scripts/: none of them point at extdata as a directory, apart from the tif-filtered STAC call above.
+- No reference to `context_kotl` or `example_context_kotl` is left outside planning/ and .git. README still uses `example_aoi_kotl.gpkg`, which is kept.
+- The new referent `data-raw/example_context_kootenay_lake.R` exists and describes the same fwapg recipe.
+
+Note (not a defect): `example_context_kootenay_lake.R:3` and `example_context_fwcp_peace.R:3` now each say "same recipe as" the other. The pointer is circular but both files exist, so it is harmless.

planning/archive/2026-10-issue-112-context-kotl/review-round3.md

@@ -0,0 +1,22 @@
+# Review round 3 — #112 supersession claim and documentation truth
+
+## Clean
+No issues found.
+
+Checked (outside planning/archive/ and docs/):
+
+- (a) Supersession: `context_kootenay_lake.gpkg` writes towns, lakes, rivers, streams,
+  highways (the deleted file's five layers) plus wsgs and ecoregions, over an AOI
+  (KOTL+LARL+DUNC+SLOC) that contains KOTL. Nothing — README, CLAUDE.md, vignettes,
+  data-raw headers, .github, tests, R/, man/, _pkgdown.yml — promises a KOTL context
+  layer. No `research/` directory exists.
+  NEWS.md:169 ("KOTL polygon assets stay in inst/extdata/ because the README quick-start
+  still uses them") is a v0.1.4 release-history entry; its premise still holds for the
+  asset the README uses (README.md:23 reads `example_aoi_kotl.gpkg`, which is kept). Not
+  made false in any way a reader would act on; history is immutable per repo convention.
+- (b) No live doc or comment names `context_kotl.gpkg` or `example_context_kotl.R`
+  (only NEWS.md:149 lists `context_kootenay_lake.gpkg`, which exists).
+- (c) `example_context_fwcp_peace.R:3` and `example_context_kootenay_lake.R:3` now cite
+  each other as "same recipe". Both scripts are self-contained and the comparisons
+  ("much larger" Peace, ~73,000 vs ~24,200 km^2) are true; the mutual reference conveys
+  a sibling relationship and would not lead a reader into wrong action.

planning/archive/2026-10-issue-112-context-kotl/task_plan.md

@@ -0,0 +1,38 @@
+# Task: inst/extdata/context_kotl.gpkg (4.9 MB) ships but nothing reads it (#112)
+
+`inst/extdata/context_kotl.gpkg` is 4.9 MB. Nothing in `R/`, `tests/`, `vignettes/` or `README.md` references it. The only mention is its producer, `data-raw/example_context_kotl.R`, plus a comment in `data-raw/example_context_fwcp_peace.R`. It dates from `3ece27c` (2026-04-07, before the Kootenay Lake vignette). The vignette added in #57 uses `context_kootenay_lake.gpkg`, which appears to supersede it. `example_aoi_kotl.gpkg` is still used by the README example, so keep it.
+
+Found by a code-check reviewer while verifying #100's tarball contents, and confirmed with `git grep -n context_kotl`.
+
+## Context (from plan-mode exploration)
+
+`inst/extdata/context_kotl.gpkg` ships in every install and nothing reads it. Exploration confirmed the issue's claim and the supersession it asked about:
+
+- `git grep context_kotl` outside `planning/` hits only its producer `data-raw/example_context_kotl.R` and a "Same recipe as example_context_kotl.R" comment at `data-raw/example_context_fwcp_peace.R:3`. Nothing in `R/`, `tests/`, `vignettes/`, `README.md`.
+- Its consumer, the KOTL vignette, was removed in v0.2.x (#50, NEWS.md:169). That NEWS line says the KOTL *polygon* assets stay for the README — that is `example_aoi_kotl.gpkg` (README.md:23), which this work keeps.
+- `context_kootenay_lake.gpkg` (2.2 MB, #57) covers the KOTL+LARL+DUNC+SLOC superset with the same layer kinds (towns, lakes, rivers, streams, highways) plus wsgs and ecoregions. Superseded.
+- The only test touching `extdata` as a directory (`test-cd_stac_catalog.R`) lists `*.tif` only, so removing a gpkg cannot affect it.
+
+**Decision taken in this plan (recommended): delete the producer script too**, rather than mark it superseded. A kept script for a deleted output is a trap — running it recreates the 4.9 MB file under `inst/extdata/` — and git history keeps the recipe. `example_aoi_kotl.R` stays (README asset).
+
+## Phase 1: Baseline measurement
+- [x] Build tarball from `git archive HEAD` (`R CMD build --no-build-vignettes --no-manual`), record tarball size and installed `inst/extdata` size — same method as #100's archive README
+
+## Phase 2: Remove the asset and its producer
+- [x] `git rm inst/extdata/context_kotl.gpkg data-raw/example_context_kotl.R`
+- [x] `data-raw/example_context_fwcp_peace.R:3`: repoint "Same recipe as example_context_kotl.R" at `example_context_kootenay_lake.R` (the live sibling)
+- [x] `git grep -n context_kotl -- ':(exclude)planning'` returns nothing
+
+## Phase 3: Verify
+- [x] Rebuild tarball, compare size before/after; record both in `findings.md`
+- [x] `devtools::test()` passes
+- [x] `R CMD check --no-manual --ignore-vignettes` on the tarball (`_R_CHECK_FORCE_SUGGESTS_=false`) — no new NOTEs vs v0.5.8
+- [x] README example still resolves `example_aoi_kotl.gpkg` via `system.file()`
+
+## Validation
+- [x] Tests pass
+- [x] `/code-check` clean on each commit
+- [x] PWF checkboxes match landed work
+- [x] `/planning-archive` on completion (README carries the before/after measurement)
+
+NEWS line and version bump are left to `/gh-pr-merge` (patch release).