Merge my fork? - #76
Open
friendly wants to merge 111 commits into
Open
Conversation
The old man/figures/logo.png had a solid white background that looked bad in dark mode. Replaced with a transparent-background PNG at the same resolution.
Picks up the transparent hex logo, plus a newer-pkgdown template refresh (Bootstrap 5.3.8, font-awesome/jquery/headroom/search dep bumps, llms.txt/*.md mirrors) since the installed pkgdown is newer than whatever last built this site. Reference example plots also re-rendered.
Move examples/ggvector.R to R/ggvector.R, rewrite it around ggarrow::geom_arrow_segment(), and export it as a user-facing helper for drawing labeled vectors from a common origin (e.g. supplementary variables added to an existing biplot). Fixes two latent bugs in the old prototype: the scale argument was declared but never applied, and label angle/hjust ignored a non-default origin. ggbiplot() now calls ggvector() at both of its variable-vector call sites (arrows before the ellipse layer, labels after, preserving the original z-order) instead of the inline geom_segment()+grid::arrow() code. ggarrow is now a hard dependency (Imports), since it's the default renderer, not an opt-in extra. Verified with devtools::check() (0 errors/warnings/notes) under both ggarrow 0.1.1 (CRAN) and 0.2.0.9000 (GitHub dev) -- the geom_arrow_segment()/arrow_head_wings() API used here is unchanged between them. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GUuPdPVD5x2719qyDNxdY8
varname.gap renames ggvector()'s resect arg to gap and wires it up as a ggbiplot()-level argument, pulling variable-vector arrowheads back (in mm) from their true endpoint to clear the correlation circle or crowded labels. vector.args is a generic named-list passthrough to the arrow-drawing ggvector() call, addressing the observation that ggarrow's default arrowhead ornament renders noticeably heavier than the old grid::arrow() one (e.g. vector.args = list(linewidth = 0.5)); anything not matched by a ggvector() argument flows on to ggarrow::geom_arrow_segment() (justify, force_arrow, sep, distort, ...). Adds a top-level issues/ dir (Rbuildignore'd) for this kind of notes/test script going forward. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MpEQGVbJfVdBnssW7J5QvL
ggarrow's default arrow_head_wings() ornament reads visually heavier than the old grid::arrow()-based triangular arrowhead at the same linewidth, so the previous default of 1.4 (matched to the old shaft width) looked noticeably thicker. Comparing panels on both wine and crime data settled on keeping arrow_head_wings() (preferred over arrow_head_line()) with the shaft linewidth lowered to 0.9. vector.args = list(linewidth = 1.4) restores the old weight. issues/arrow-tests.R rebuilds the old vs. new arrow comparison on the crime dataset (pulling the pre-ggarrow ggbiplot() from git history), matching what README.Rmd's crime-biplot1 example actually shows rather than wine. Also flags (not fixed here) two unrelated bugs found while building it: README.Rmd's ellipse.level should be ellipse.prob, and its crime-biplot1 chunk needs geom.ind = "text" to keep showing state labels under the current geom.ind default. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MpEQGVbJfVdBnssW7J5QvL
crime-biplot1 passed ellipse.level (not a ggbiplot() argument; silently absorbed by ...) instead of ellipse.prob. Both crime chunks relied on labels = crime$st to show state abbreviations, but geom.ind defaults to "point" (since v0.6.4) and no longer auto-switches to text, so replaying them showed plain colored points with no labels. Added geom.ind = "text" to both. Re-knit via devtools::build_readme() (installs the current dev version first so packageVersion() picks up 0.6.5 instead of the stale 0.6.2 in the system library) — also refreshes the reference figures with the new linewidth = 0.9 default arrows. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MpEQGVbJfVdBnssW7J5QvL
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TCpb3UU6Rebmr2saN5Lkn6
- Bump stale Date field, clearing the "Date field is over a month old" R CMD check NOTE. - Rewrite cran-comments.md for this submission (was still describing the earlier 0.6.2 release). - Convert the varname.gap/vector.args roxygen blocks in ggbiplot.R from Rd macros to markdown syntax, matching the rest of the package's documentation style. - Fix millimetres -> millimeters (Language: en-US) in ggbiplot.R and ggvector.R docs. - Add newly-introduced technical terms to inst/WORDLIST so spell_check() stays clean. devtools::check(cran = TRUE, remote = TRUE): 0 errors, 0 warnings, 0 notes. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TCpb3UU6Rebmr2saN5Lkn6
- Add ggvector() to the reference index (was exported but missing, which broke pkgdown::build_site() outright). - Fix _pkgdown.yaml's site url to https, matching DESCRIPTION's URL field (was flagged by pkgdown's sitrep as a URL mismatch). - Rebuilt site picks up the 0.6.2 -> 0.6.5 version bump on the home page badge, previously stale. - Add "pre" to inst/WORDLIST (spelling::spell_check_package() now reports no errors). - Ignore the release-check helper's .release_check_result.rds artifact (myutil::release_check()) in .gitignore/.Rbuildignore. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TCpb3UU6Rebmr2saN5Lkn6
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TCpb3UU6Rebmr2saN5Lkn6
Matches the convention used in Michael's other packages (heplots, Guerry, etc.). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TCpb3UU6Rebmr2saN5Lkn6
release_revdep() left an ~11MB revdep/checks + revdep/library + data.sqlite behind. Add revdep/.gitignore (standard revdepcheck pattern) and ^revdep$ to .Rbuildignore so only the small summary files (README.md, cran.md, problems.md, failures.md) are tracked. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TCpb3UU6Rebmr2saN5Lkn6
ggvector()'s text labels always angled along each vector's own direction, with no way to override it -- forcing a workaround of drawing labels separately with a plain geom_text() when horizontal (or any fixed-angle) labels were wanted (see Vis-MLM-book's diabetes-mds.R). angle = NULL keeps the existing per-vector behavior; a fixed number draws every label at that angle instead. Exposed on ggbiplot() as varname.angle, threaded to the text-layer call only (matches how varname.gap threads to the arrow-layer call). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TCpb3UU6Rebmr2saN5Lkn6
ggplot2 computes axis limits from each layer's anchor coordinates only, not the rendered text's extent, so a long variable name (or one pushed outward by a large varname.adjust) can extend past the panel edge and get silently truncated under the old clip = "on" default. "off" lets that overflow draw into the margin instead. Pass clip = "on" to restore the old behavior.
…abeling Finishes the ggplot2-layer mechanism sketched but never completed in heplots/dev/stat_noteworthy.R, wrapping the already-shipped heplots::noteworthy() for point selection. Verified the draft's documented "DOESN'T WORK" bug (stat_noteworthy(method=, label=)) does not actually reproduce under current ggplot2 -- see dev/test-noteworthy.R for the non-confounded repro and 14 worked examples (every noteworthy() method, level filtering, multiple geoms, composition on top of a real ggbiplot()). dev/stat_noteworthy.R: the StatNoteworthy ggproto + stat_noteworthy() constructor, with a production-style roxygen doc block. dev/test-noteworthy.R: bug check + examples, split out separately. Prototype only -- no ggbiplot()/ggvector() argument added yet, no version bump, no NEWS.md entry (dev/ is .Rbuildignore'd).
Matches the convention used for heplots/vcdExtra/etc. Moved the completed ggarrow integration plan out to issues/task-ggarrows.md (kept for reference, no longer an active task). TASKS-all.md's ggbiplot section now just points here.
test-noteworthy.R: added section testing whether point selection is correctly scoped per facet_wrap() panel (yes) vs. per aes(color=) group sharing one panel (no, pooled globally -- known compute_panel() vs compute_group() tradeoff). Also print each example plot. ggextenders-noteworthy.md: notes from the original ggplot-extension-club discussion #91 (opened by friendly) that the StatNoteworthy design was based on -- facet/group scoping design menu, a sharper root-cause diagnosis of the label= bug than previously documented, a real repel/label-blanking gap for dense plots, and prior-art pointers.
…and scale arg Existing prcomp/princomp/PCA/lda behavior is unchanged (pure refactor of the old if/else chain into one method per class). New reflect.data.frame()/reflect.matrix()/reflect.list() negate the given columns (by name or index) directly, so other packages/objects can register their own reflect.<class>() method instead of needing changes here -- matches what candisc's own notes wanted this for. All methods also gain a scale argument (default 1): selected columns become -scale * columns instead of just -1 * columns, e.g. to stretch or shrink variable vectors in a biplot at the same time as reflecting them (a negative scale scales without reflecting). Verified all methods + error paths manually; devtools::check() 0/0/0. No version bump (folded into the still-unreleased 0.6.6).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
OK, here's all the work I've done since I started on this.
If you prefer, you could merge this into a new branch
friendly, but that might be more work.You can comment on this PR before you accept it & merge