NEWS.md

In vennDiagramLab: Headless Venn Diagram Analysis and Rendering

vennDiagramLab — NEWS

v2.0.5 — 2026-05-12 — Fix inst/CITATION pre-install crash + finish vignette skip

Patch release addressing two issues from the CRAN auto-check on v2.0.4.

Fixed

• inst/CITATION no longer errors during CRAN incoming-feasibility check. The file previously called utils::packageDescription("vennDiagramLab"), which returns NA when the package is not yet installed (as during CRAN's pre-install incoming check). NA$Version then triggers the cryptic "$ operator is invalid for atomic vectors" error. The fix uses the meta variable that R auto-injects when parsing the CITATION file, the documented pattern from "Writing R Extensions". citation("vennDiagramLab") output is unchanged on installed copies.

Changed

• Vignette evaluation now globally gated on NOT_CRAN for every chunk in every vignette. v2.0.4 gated only the four obviously-slow chunks (render_upset, render_network, to_pdf_report, geom_venn composite), but win-builder still reported [12m] OK on the re-building of vignette outputs step. The remaining time was spent on per-vignette library(vennDiagramLab) calls (which transitively load ggplot2, ComplexUpset, ggraph, tidygraph, rsvg, patchwork, gridExtra, BiocGenerics — slow on Windows VMs) plus the lightweight analyze() / render_venn_svg() / broom chunks across eight vignettes. The setup chunk of every vignette now sets knitr::opts_chunk$set(eval = NOT_CRAN), so on CRAN the rebuild is text-only (~1 minute total). Under devtools::check() and on the package's GitHub Actions CI matrix NOT_CRAN=true and all chunks evaluate normally. Locally verified: total vignette rebuild dropped from 12 min (Windows) / ~2 min (Mac) to ~12 seconds on Mac.

No public-API or feature changes.

CRAN history

• v2.0.4 (2026-05-12): rejected on win-builder — (1) inst/CITATION packageDescription() NA crash during incoming check, (2) overall checktime 20 min > 10 min (vignettes 12 min — partial gating was insufficient).
• v2.0.5 (this release): CITATION uses auto-injected meta; all vignette chunks gated.

v2.0.4 — 2026-05-10 — DESCRIPTION quoting + vignette buildtime fix

Patch release addressing two issues from the CRAN auto-check + human reviewer feedback on v2.0.3.

Changed

• DESCRIPTION single-quotes software names per CRAN reviewer (Uwe Ligges). Quoted: 'Venn Diagram Lab', 'UpSet', 'CSV', 'TSV', 'GMT', 'GMX', 'Jaccard', 'ggplot2', 'tidygraph', 'broom'. Suppresses the "Possibly misspelled words in DESCRIPTION" NOTE (flagged GMX, Jaccard, TSV, UpSet, ggplot, tidygraph).
• Vignette buildtime cut from ~12 minutes to ~1 minute on CRAN by gating the slow rendering chunks (render_upset, render_network, to_pdf_report, geom_venn composite, rsvg_png/rsvg_pdf export) on NOT_CRAN. On dev (devtools::check() sets NOT_CRAN=true) and CI all chunks still build their figures; the full 590+ test suite + 8 vignettes still execute on the 5-cell GitHub Actions matrix. On CRAN only the lightweight analysis chunks evaluate, bringing total checktime well under the 10-minute budget.

No public-API or feature changes.

CRAN history

• v2.0.3 (2026-05-08): rejected on win-builder — 18-min overall checktime (vignettes alone ~12 min) and DESCRIPTION-quoting NOTE.
• v2.0.4 (this release): vignette NOT_CRAN gating + DESCRIPTION quoting.

v2.0.3 — 2026-05-08 — Skip slow tests on CRAN

Patch release reducing CRAN auto-check time below the 10-minute target. The full test suite (590+ tests) and integration tests against bundled samples + 44 SVG models pushed the Windows R CMD check to ~41 minutes, which auto-rejected the v2.0.2 incoming pretest.

Changed

• Slow-running tests (test-render-pdf, test-render-svg, test-render-upset, test-render-network, test-parity-with-webapp) now call testthat::skip_on_cran() at the top of every test_that() block. CRAN sees only the fast unit-test slice (test-analysis, test-classes, test-statistics, test-broom-tidy, test-ggplot-layer, test-io, test-samples, test-proportional, test-tsv-export). The skipped tests still run on the package's GitHub Actions CI matrix (5 cells × 590 tests), where NOT_CRAN=true triggers the full test suite.

No public-API or feature changes; the package's user-facing behavior is identical.

CRAN history

• v2.0.1 (2026-05-07): rejected on Windows pretest due to a CRLF byte-parity bug in cat(..., file = path) text-mode write.
• v2.0.2 (2026-05-07): the CRLF fix; rejected on Windows pretest due to 41-minute overall checktime exceeding CRAN's 10-minute auto-pass threshold.
• v2.0.3 (this release): skip_on_cran() in slow tests; expected sub-10-min pretest.

v2.0.2 — 2026-05-07 — Windows CRLF parity fix

Patch release fixing a cross-platform packaging bug that surfaced on the win-builder CRAN check for v2.0.1.

Fixed

• TSV writers now produce byte-identical output on Windows. cat(out, file = path) opens the connection in text mode by default; on Windows that converts "\n" → "\r\n", breaking the 12 byte-parity tests against the Python golden fixtures (which use "\n" only, matching the React webapp's TSV exports). New private helper .write_bytes() opens the file in "wb" (binary) mode and uses writeBin(charToRaw(x), con). Affects to_region_summary_tsv(), to_matrix_tsv(), and to_statistics_tsv(). No public-API changes; behavior on Linux/macOS unchanged.

CRAN history

v2.0.1 was submitted to CRAN on 2026-05-07; the win-builder pretest failed with the CRLF bug above before the human reviewer was reached. v2.0.2 is the resubmission with the fix.

v2.0.1 — 2026-05-06 — Pre-submission feedback fix (Marci)

Patch release fixing a packaging bug that broke analyze() for users installing via remotes::install_github().

Fixed

• analyze() no longer fails with "no file found" after remotes::install_github(..., subdir = "r"). Root cause: the bundled 44 SVG templates, 44 region JSONs, and 5 sample datasets under r/inst/extdata/ were gitignored; only the directory structure (via .gitkeep files) was tracked. The local developer populated them by running Rscript r/data-raw/sync_data.R before R CMD build, so the CRAN / Bioconductor tarballs would have shipped them — but remotes::install_github() builds from the git tree, where those files were absent. The result: install succeeds, but system.file("extdata", "models", "json", package = "vennDiagramLab", mustWork = TRUE) returns "", and the first analyze() call throws the cryptic mustWork error.

The fix tracks the synced data files in git (root .gitignore updated), so install_github now produces a working install out of the box. The synced files are regenerated and committed as part of the release workflow; the r/data-raw/sync_data.R script remains the single source of truth.

Changed

• .models_json_dir() (internal) replaced the cryptic mustWork = TRUE error with an actionable message walking the user through the clone + sync + R CMD INSTALL workflow as a fallback. This branch is now defensive — the gitignore fix removed the trigger — but mis-built installs (hand-edited package directories, missing rebuilds after a model change) will still see a useful message instead of "no file found".

No public-API changes. No new features.

v2.0.0 — 2026-05-04

First public release. Headless companion to the Venn Diagram Lab web tool and the Python venn-diagram-lab package, targeting CRAN + Bioconductor.

Core (Phase 1)

• VennDataset, RegionData, RegionResult, StatisticsResult — four S4 classes covering inputs, regions, results, and pairwise statistics.
• Loaders for CSV, TSV, GMT, GMX (load_csv, load_tsv, load_gmt, load_gmx) plus load_sample / list_samples for five bundled datasets (3 biological, 2 mock).
• analyze(dataset, model = "auto") — region enumeration via integer bitmask, model resolution mirroring the web tool.
• statistics(result) — lazy StatisticsResult with five pairwise metrics: Jaccard, Dice, overlap coefficient, fold enrichment, hypergeometric p-value (BH-FDR adjusted).
• Stateless metric helpers: jaccard(), dice(), overlap_coefficient(), hypergeometric_p_value(), fold_enrichment(), bh_fdr(), compute_pairwise().
• effective_universe(result) — universe N consistent with the web tool's binary/aggregated semantics.

TSV exports + parity (Phase 2)

• to_region_summary_tsv(), to_matrix_tsv(), to_statistics_tsv() — three byte-equivalent writers matching the React webapp's TSV exports.
• JS-style float formatting helpers under the hood (toExponential / toFixed semantics).
• Twelve byte-parity tests against the Python golden fixtures (4 samples × 3 export types). The 8-set dataset_mock_streaming_platforms is xfail-strict due to the documented row-vs-set divergence in the web tool's loader.

SVG rendering (Phase 3)

• render_venn_svg(result, ...) — templates the bundled 44-model SVG library via xml2 (Count_*, Name*, Bullet*, Shape* substitution).
• Custom set_names, colors, title, show_names, show_counts overrides per render.
• analyze(ds, model = "proportional") — area-proportional 2-set (exact) and 3-set (approximate) generator with solve_2set, solve_3set, circle_intersection_area, generate_proportional_svg helpers.

UpSet + Network rendering (Phase 4)

• render_upset(result, ...) — ComplexUpset-based UpSet plot with sort modes (size, degree), color modes (depth, heatmap, custom), and threshold / max_columns cutoffs.
• render_network(result, ...) — force-directed network via tidygraph + ggraph. Edge metric configurable (intersection, jaccard, fold_enrichment, overlap_coefficient); significance threshold colorizes edges.

PDF reports (Phase 5)

• to_pdf_report(result, path, ...) — multi-page US-Letter-landscape PDF combining venn + UpSet + statistics + network + about pages, generated via grDevices::pdf + patchwork.
• Network and About pages are toggleable via include_network and include_about.

ggplot2 + broom integration (Phase 6)

• geom_venn(data, ...) — embed a rendered venn as a ggplot2 layer (annotation_custom + geom_blank + coord_fixed).
• tidy.RegionResult(), glance.RegionResult(), augment.RegionResult() — broom-compatible S3 methods returning tibbles.
• broom and tibble are Suggests-only (registered via @exportS3Method) — users not in the tidyverse pay zero install cost.

Vignette gallery (Phase 7)

• Eight RMarkdown vignettes, all executed during R CMD check --as-cran:
• v01_quickstart — five-step intro
• v02_real_cancer_drivers — long-form biological walkthrough (Vogelstein / COSMIC / OncoKB / IntOGen)
• v03_proportional_diagrams — when to use area-proportional + low-level helpers
• v04_upset_vs_venn_vs_network — choosing the right visualization
• v05_statistics_deep_dive — Jaccard / Dice / hypergeometric / BH-FDR worked
• v06_pipeline_integration — broom + dplyr + targets pipeline sketch
• v07_pdf_reports — composite PDF generation
• v08_custom_styling_and_export — custom names/colors, geom_venn, multi-format export

Compatibility notes

• On R < 4.6, render_upset() and to_pdf_report() emit a runtime warning about an upstream ComplexUpset 1.3.3 / ggplot2 4.x incompatibility (tracking krassowski/complex-upset#213). Vignette chunks that depend on UpSet rendering are gated on eval = (getRversion() >= "4.6"). The warning + skip will be removed when ComplexUpset 1.4+ ships.

vennDiagramLab documentation built on May 19, 2026, 1:07 a.m.