Skip to content

About

Static explorer over precomputed RaMa-Scene scenario results on EXIOBASE 2011, with every number traced to a verified run

Topics

Resources

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

155 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

RaMa-Scene Explorer

ci CodSpeed DOI OpenSSF Scorecard

A static, map-first page over precomputed RaMa-Scene engine output on the EXIOBASE 2011 base year. It shows where a country's consumption causes greenhouse-gas emissions, jobs and metal-ore extraction, split by 15 industry sections, and what one intervention would have changed: three named interventions and a grid of 150 material-efficiency scenarios. Nothing on the published page is a live calculation (an optional live mode needs a job API; see docs/page.md), and nothing is a forecast. The engine changes final demand or technical coefficients in the 2011 economy and has no projection step.

Open it at https://simonvanlierde.github.io/ramascene-explorer/.

data/results.json and data/embodied.parquet come from one precompute run on 2026-10-14, with engine split/3-performance @ d9113ca, the head of CMLPlatform/ramascene#74. The same run wrote data/regions.json and data/flows.json. The precompute checks the flows against these detail files, and the page does not read them. The Parquet file is too large for git, so it ships as a release asset (see docs/precompute.md).

Nothing else in the repository is a result. tests/fixture.json and tests/fixture-embodied.parquet are synthetic data for testing the page. For any data whose data_kind is not exiobase-v3, the page shows a red "SYNTHETIC TEST DATA: NOT A RESULT" banner.

View

The page is Svelte 5, built with Vite. It needs Node 24 and pnpm. Install once with pnpm install, and fetch the flows from the release (or with gh release download data-2011 --pattern embodied.parquet --dir data):

curl -L -o data/embodied.parquet https://github.com/simonvanlierde/ramascene-explorer/releases/download/data-2011/embodied.parquet

Then:

pnpm dev        # dev server with live reload, on http://localhost:5173/
pnpm build      # static site in dist/, with relative paths
pnpm preview    # serves dist/ on http://127.0.0.1:4180/

Open ?mode=whatif&scenario=steel-in-vehicles-europe or ?region=CN&dir=out&indicator=7&sections=2. Add data=tests/fixture.json to see the page on the synthetic fixture. The selection lives in the query string, so you can share a view by its URL.

Brightway check

brightway-check/ loads the same EXIOBASE v3.3.sm arrays into Brightway as an input-output database and scores one product four ways: Brightway, numpy, pymrio and this engine through the adapter in precompute.py. All agree to within 2.2e-13. It is a separate uv project with its own Brightway and pymrio pins.

Test

uv run pytest                          # everything except the engine adapter
uv run pytest --engine="$ENGINE_DIR"   # adds the real-engine adapter checks (768 MB)
node --test                            # the page logic (after pnpm install)
pnpm exec playwright test              # the built page in six browser setups, with axe (once: pnpm exec playwright install chromium firefox webkit)
pnpm mutation                          # mutation score of flows.js with Stryker (about 4 minutes); fails below 75%

Write --engine=PATH with the =. Without it, pytest takes the path as its root directory and loads the engine's own conftest.py.

Benchmark

pnpm bench   # the page's per-selection work, with Vitest

.github/workflows/codspeed.yml runs it on every pull request and push to main under CodSpeed's CPU simulation, which reports the change against main on the pull request.

Docs

Cite

Cite the archived release: doi:10.5281/zenodo.23461260 always points to the latest version. GitHub's "Cite this repository" button reads CITATION.cff. Cite EXIOBASE and the RaMa-Scene engine separately (see below and docs/precompute.md).

Licence

The code is under the MIT licence (LICENSE), © 2026 Simon van Lierde.

The result files in data/ are computed from EXIOBASE v3.3.sm by Franco Donati and Arjan de Koning (Zenodo, version V1, 2019-11-08, doi:10.5281/zenodo.3533196), which the Zenodo record publishes under Creative Commons Attribution 4.0 International (CC BY 4.0). The results are a changed version of that dataset: each file holds scenario totals and flows computed from its matrices, not the matrices themselves. They are shared under CC BY 4.0 too. Reuse them with credit to EXIOBASE v3.3.sm and its authors, as above, and say that they were changed.

The fonts are Public Sans and Spectral, under the SIL Open Font License 1.1, bundled from the @fontsource packages. The map geometry is Natural Earth, public domain; its source is in docs/page.md.

About

Static explorer over precomputed RaMa-Scene scenario results on EXIOBASE 2011, with every number traced to a verified run

Topics

Resources

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages