2026-09-24 — Bob prune of bob.md

Housekeeping pass by /bob update. Moved completed and dated material off the status card. No project work was done in this session; nothing below is new — each item is already recorded in the log it cites.

Removed from ## This Week (completed)

  • DONE 2026-09-09: clones current; weapon-plot arrow colour + dropped zero-count points fixed; toolbar icons hidden on all four widgets; deployed and verified live — see docs/logs/2026-09-09_weapon-plot-and-toolbar-fixes.md
  • DONE 2026-09-09 (Hunter): 2025 victim memorial roster added, scroll regenerated, deployed
  • DONE 2026-09-10: dark-theme wounded/killed dotplot for Pete (graphics bw= switch), pushed; light/dark theming logged as overhaul requirement

Removed from ## Notes (dated narrative, duplicated in Milestones + log)

  • 2026-09-10: dark-theme wounded/killed dotplot for Pete via new bw= switch in graphics plots_cartesian.R (uncommitted; prototype for the two-variant plan) — same log as above

Still open (left on the card)

  • Approval needed from Peter: untrack output/*_files/ + .DS_Store in graphics; remove defunct PROJ_activeattack_app (local + GitHub). Neither actioned.
  • Tell Hunter about the 09-09 fixes and the ggiraph 0.9.1 (Mac) vs 0.9.6 (Windows) drift; agree on pinning.

Status check at prune time: all three sibling repos assumed in place; website repo main in sync with origin/main, working tree clean.

Addendum: officer-shot phase is not in the AA data

Checked DATA_aa/input/events/2025/aa_data_2025.xlsx (634 events, 79 cols) for a variable recording when in the response an officer was shot (approach, arrival, entry, etc.). It does not exist, raw or cleaned. What exists:

  • le.shot/num.le.shot and killed/wounded counts (74 events with an officer shot); only the counts reach the finished aa_data_wide_*.rds.
  • off.duty, who.notes, resolution.notes are dropped by code_events.R; the timing block (first.report, le.arrival, entry, ended) is commented out and sparse anyway (le.arrival filled for 11 of the 74).
  • Phase is inferable from the two notes columns for roughly three quarters of the 74 (ambush, approach/initial contact, entry/shootout at scene, pursuit/standoff/later day, incidental/off duty). Would need hand coding into a side table keyed on case_code.

Decision: not worth building here. LEOKA is the source for officer-assault circumstances; use that instead.

Addendum: ggiraph pinned to 0.9.6

Graphics repo code/0_packages.R now carries GGIRAPH_VERSION <- "0.9.6" with a guard that installs the CRAN binary when the pin is the current release and falls back to remotes::install_version() when it is not. Peter’s Mac was actually on 0.9.0 (earlier logs said 0.9.1; single library, checked). Upgraded to 0.9.6 here, which also pulled in gdtools 0.5.1. The remotes source build fails on this Mac because R CMD config CXX resolves to Homebrew’s g++-16; the binary path avoids it. Verified on 0.9.6: save_girafe_widget()’s liberation-* filter still matches (four such deps), and opts_toolbar(hidden = c("selection","zoom","misc")) is accepted. Existing widgets were not re-rendered; next full render will be on 0.9.6 on both machines.

Addendum: repo cleanup (approved by Peter this session)

  • Graphics repo: git rm --cached on output/.DS_Store and the four output/*_files/ dirs (59 files, ~1.9 MB, ggiraph JS 0.8.10 and 0.9.4); .gitignore gains output/*_files/ and .DS_Store. Files remain on disk.
  • PROJ_activeattack_app: local clone inspected (clean, nothing unpushed, no stashes, last push 2024-10-21, only ignored files were .Rproj.user and .DS_Store) then deleted from ~/Documents/GitHub/. GitHub deletion failed: gh repo delete returned HTTP 403 “Must have admin rights” and the token lacks the delete_repo scope. Left as the one open This Week item.
  • No registry or code references to the app repo remained; only this card did.