Overstorey
  • Guides
  • Theory
  • Adaptively
  • Reproductions
  • Reference
  • Pipeline
Categories
All (10)
announcement (1)
code-review (1)
develop (3)
engineering (3)
experimental (2)
ff16 (1)
hydraulics (5)
numerics (1)
performance (3)
prototype (1)
refactor (1)
scm (1)
soil-water (3)
tf24 (5)

Adaptively

The plant notebook — release notes, case studies, and worked problems, each pinned to the version or commit it was built against.

worked against pinned versions

Posts here are reproducible against a specific state of plant. Master posts pin to a release version; develop posts pin to a branch and commit, since there’s no tag to point at yet. The table shows both — a blank version with a branch of develop means the result is tied to a commit, not a release.

Date Title branch version Reading Time
Jun 20, 2026 Moving cohort refining into C++ develop   3 min
Jun 19, 2026 Making root water uptake fast develop   6 min
Jun 19, 2026 A reference prototype for root water uptake develop   22 min
Jun 19, 2026 Reviewing the TF24 leaf model code develop   4 min
Jun 18, 2026 Optimising the FF16 hot paths develop   3 min
Jun 15, 2026 A soil-water bucket model from AWRA-L develop   15 min
Jun 10, 2026 Hydraulics inner-loop benchmark develop   2 min
Jul 29, 2025 When the Richards equation fights back     9 min
May 22, 2024 Prototyping a soil water-balance bucket model     19 min
Feb 23, 2016 Key technologies used to build the plant package     8 min
No matching items

How posts are pinned

From master (released)

---
title: "Calibrating an early-successional strategy"
date: 2026-05-02
plant-version: "2.1.0"
categories: [calibration, case-study]
---

The badge renders plant 2.1.0. Install in CI from the matching tag.

From develop (unreleased)

---
title: "Hydraulics inner-loop benchmark"
date: 2026-06-10
plant-ref: "develop"
plant-sha: "a1b2c3d4e5f6..."   # full or short; badge truncates to 7
categories: [hydraulics, performance]
---

The badge renders ● plant @develop a1b2c3d in a dashed bark-toned pill, so a reader immediately sees it’s a moving target. Because results are frozen, the figure stays faithful to that commit even after develop moves on — and the SHA tells anyone exactly what to check out to reproduce it.

Why this matters with two branches

freeze: auto means a post is only re-executed when its own source changes. So a develop post built today won’t silently re-run (and possibly break) when develop advances tomorrow. The version/commit stamped on the page is therefore always the one that produced the figures. To genuinely re-run an old develop post, check out its plant-sha, restore its lockfile, and delete its _freeze/ entry.

Pinning the environment per post

For light posts the project-level renv.lock is enough. For an expensive simulation you may want the post pinned to its own snapshot so a later project-wide upgrade can’t change its result even if you do force a re-run. See the hydraulics benchmark for the pattern: a post-local renv profile plus a recorded plant_session_info() footer.

Built with Quarto · frozen for reproducibility

 
  • Edit this page
  • Report an issue
  • License