One call per tablecached after the first pull
x <- rmorie::morie_load_dataset(
"chicago_crime/incidents")
rmorie::morie_hosted_datasets()rmoriebricklayer::dload(
"hacker_news/stories")rmbl data list rmbl data pull nhtsa_fars/accidents
from morie.data import load_dataset df = load_dataset("chicago_crime/incidents")
morie pull chicago_crime/incidents --out x.csv
The list is openthe tables come through the packages
curl https://data.rmorie.com/manifest.json
Table files are served to the MORIE packages, which check each one against the SHA-256 in the manifest and resume a large one where it stopped. A browser that opens a table’s address gets a short note naming the packages instead of the file. The browse page queries the SQLite databases in this browser.
Sources, versionslicences, what is kept
Sources
Every table names its source in the manifest, and each keeps its source’s own licence (Statistics Canada Open Licence, the Open Government Licences of Ontario and Canada, the BigQuery publishers’ terms). Cite the source the manifest names.
Fixed copies
Each table is a fixed copy, checked against the SHA-256 in the manifest, so a result reproduces. No project machine is in the request path.
Not served here
ICPSR studies, whose terms do not allow redistribution. Asking for one gives the address to get it from ICPSR.
What is kept
No register of users and no sign-in: only how often each table is downloaded, with no address, place or identifier. Nothing from the SQL you run here, which runs on your machine.
Fair use
Requests are rate limited per address so the service stays available to everyone.
Found by the packages
The packages read this address from the signed document at rmorie.com/.well-known/morie-services.json, so it can move without a release.