cleaned package review

▶ Try the Scenario Builder mockup (redesigned data entry, feedback built in) · UI/UX review of the app · review of the mockup

An adversarial review of the cleaned R package (v0.7.0, staging @ 800d53a), which powers the iCLEANED app. Four expert reviews, unit tests against the IPCC source text, proposed fixes tested on a copy, and a script that reproduces every number.

Nothing here changes the cleaned repository. Fixes are proposals for its maintainers. Every page has a comment thread (GitHub Discussions via giscus).

Start here

• 00 summary: verdict, deduplicated critical findings, fix order.
• Critical issues: one page per issue with current / fixed / IPCC values.
• 08 demo: the same, plus the effect on the two iCLEANED example studies.

Reviews

• 01 climate and GHG
• 02 livestock
• 03 land use, water, soil, N
• 04 package design and code
• 05 app relevance and workshop plan

App UI/UX and redesign

• UI/UX review of the iCLEANED web app: 88 findings from code reading and a live walkthrough, a workshop run sheet, and a feedback-capture form mapped to the GitHub issue forms.
• Scenario Builder mockup: a clickable redesign of the data-entry flow. It compiles plain-language answers into the exact cleaned input, scales from one household to a national herd, shows where every value comes from (you entered / from database / assumed / fixed), and captures feedback in place (comment mode, highlights, screengrabs, ratings, A/B votes). Opened from this site, feedback is saved on your device and exported from the Feedback screen; the shared-store version lives on claude.ai.
• Design benchmarks: patterns worth adapting from 22 farm and livestock tools.
• Review of the mockup: 93 findings from three adversarial passes (compile correctness, live UX, runtime hygiene) and what was fixed.
• Source: mockup source (vanilla JS, one build script) and Markdown.

Evidence

• Sources: every cited IPCC equation and table, linked to the PDF page, with the package's own parameter tables rendered from inst/extdata.
• 06 critical tests: tests against IPCC equations, with citations.
• 07 fix log: each proposed fix, its status, the test matrix.

Reproduce it yourself

In the browser: opens a notebook in Google Colab (R runtime) with one cell per issue and a final cell that applies the proposed fixes. Setup takes a few minutes. Notebook file.

Locally:

curl -O https://peetmate.github.io/cleaned-review/reproduce/reproduce_critical_issues.R
Rscript reproduce_critical_issues.R

The script installs any missing packages into a cleaned_review_lib folder next to it, installs cleaned at the reviewed commit there, and prints the package's value next to the IPCC value for each issue. For the fixed column:

curl -O https://peetmate.github.io/cleaned-review/fixes/proposed_fixes.R
CLEANED_FIXES=proposed_fixes.R Rscript reproduce_critical_issues.R

Sources: tests, proposed_fixes.R, diffs and test matrix, Markdown reports.