DO NOT MERGE: reportlab greenfield planner spike

[yyolk]

DO NOT MERGE. Experimental throwaway spike. This branch will not be merged and may be deleted later.

What this is

A greenfield “what if we never ported the Ruby/Typst decisions” experiment. The Typst/parch application tree is stripped. Tiny Python package writes a yearly planner PDF with open-source ReportLab only (pdfgen.canvas + bookmarkPage / linkAbsolute). No fpdf2, no Typst, no ReportLab PLUS.

Glanceable artifacts (in-repo)

Same shape as the fpdf2 spike so yolk can flip pages:

• artifacts/specimen-2026.pdf — cover + year + Q1 + January + first week + 1 Jan (6 pages)
• artifacts/specimen-2026-notes2.pdf — same + notes 1/2 and 2/2 for 1 Jan (8 pages)
• artifacts/planner-2026.pdf — full-year N=0 (436 pages, ~4.3 MB)
• Preview PNGs (pdftoppm @ 140 dpi):
  • artifacts/preview/cover.png
  • artifacts/preview/year.png
  • artifacts/preview/quarter.png
  • artifacts/preview/month.png
  • artifacts/preview/week.png
  • artifacts/preview/day.png

Cover
Year
Quarter
Month
Week
Day

Generate

python3 -m venv .venv
.venv/bin/pip install -e .
.venv/bin/press --specimen              # artifacts/specimen-2026.pdf
.venv/bin/press                          # N=0 full year
.venv/bin/press --notes-pages 2

Defaults: year 2026, Monday week start, page 106 × 144 mm portrait.

--notes-pages N: 1 daily page + N dedicated notes pages per day. N=0 (default) is the day page only, with its on-page notes panel.

Notes-pages benchmark

Full-year 2026, Monday weeks. /usr/bin/time -v around press (generate only). VM: 15 GiB RAM, 4 CPUs, no swap. Order 0 → 2 → 10 → 20 → 30 → 50 → 100. Nothing OOM’d. Last completed N=100. Died N: none. Full-year N≥2 PDFs not committed (N=100 was ~109 MB). Details: BENCH.md.

Named dests counted via ReportLab catalog /Dests → pypdf.named_destinations (keys like /day-2026-07-02).

N	pages	wall	peak RSS	PDF bytes	link annots	named dests
0	436	1.25 s	55.6 MiB	4.3 MB	18 259	436
2	1 166	1.87 s	73.1 MiB	6.2 MB	24 099	1 166
10	4 086	4.40 s	138.4 MiB	14.5 MB	50 379	4 086
20	7 736	7.76 s	224.5 MiB	25.0 MB	83 229	7 736
30	11 386	11.29 s	297.5 MiB	35.5 MB	116 079	11 386
50	18 686	17.84 s	472.7 MiB	56.5 MB	181 779	18 686
100	36 936	34.60 s	892.6 MiB	109 MB	346 029	36 936

Spot-check (pypdf): day-2026-07-02 and day-2026-07-02-notes-1 exist for N≥2; day→notes-1 and notes-1→day annots resolve (N=2: pages 618↔619; N=100: 18454↔18455).

vs fpdf2 (PR #204): same order of magnitude. ReportLab is a bit slower/hungrier at N=100 (35 s / 893 MiB vs 31 s / 629 MiB) and files are ~1.4× larger. No Typst notes-slab cliff on either path.

CI

Old Typst jobs replaced with a specimen generate stub. Do not treat as product CI.

_{To show artifacts inline, enable in settings.}

  

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}