Snowflake share

Enterprise plan. eolas publishes the entire warehouse to your Snowflake account as a Snowflake secure data share — zero-copy, no ETL on your side. You mount one database and query ~1,500 always-fresh NZ + OECD datasets as native Iceberg tables, with eolas refreshing the underlying data on each source's cadence. Your compute, our data.

This page is the end-to-end lifecycle: prerequisites → request → mount → query → refresh → offboarding.

1. Prerequisites

Requirement	Detail
eolas plan	Enterprise. Talk to us to enable it.
Cloud + region	Your Snowflake account must be on AWS `ap-southeast-2` (Sydney) — the share is region- and cloud-local.
Role	An `ACCOUNTADMIN` (or a role with `IMPORT SHARE` + `CREATE DATABASE`) to mount the share.

Not on AWS Sydney?

Snowflake secure shares only work directly when provider and consumer are in the same cloud + region. If your account is on Azure, GCP, or another AWS region, we provision a replicated reader path instead (a small surcharge covers the cross-region replication). Tell us your account's cloud + region when you get in touch and we'll confirm the approach.

Email phil@eolas.fyi with:

Your Snowflake account identifier — run this in your account and paste the result:

SELECT CURRENT_ORGANIZATION_NAME() || '.' || CURRENT_ACCOUNT_NAME() AS account_locator,
       CURRENT_REGION() AS region;

Your organisation name (for the access record).

We add your account to the share and confirm — typically within one business day. There is nothing to install and no data egress on your side.

Once we've confirmed your account is added, mount it (one-time):

CREATE DATABASE eolas FROM SHARE MRWTYQY-WZ79363.vs_warehouse_share;

eolas is just the local database name — call it whatever suits your conventions. That's the whole setup; the data is queryable immediately.

eolas  (database)
 └── <namespace>  (schema, one per data source)
      └── <dataset>  (Iceberg table)

Each schema is a source namespace and each table is a dataset — the same names you see in the dataset catalogue and the API. Discover everything from SQL:

-- every source namespace
SHOW SCHEMAS IN DATABASE eolas;

-- every dataset in a namespace
SHOW TABLES IN SCHEMA eolas.rbnz;

-- column names + types for one dataset
DESCRIBE TABLE eolas.rbnz.rbnz_b2_wholesale_rates_daily;

Licence and attribution

Raw Iceberg tables do not embed licence text in every row. Use one of these paths — all share the same citation wording via app/licensing.py so Snowflake, API, and bulk cannot drift:

1. Snowflake lookup table (recommended for warehouse consumers)

Query eolas.EOLAS_META.ATTRIBUTIONS — one row per dataset with namespace, table_name, source, licence, attribution_text, source_url, and refreshed_at:

SELECT namespace, table_name, source, licence, attribution_text
FROM eolas.EOLAS_META.ATTRIBUTIONS
WHERE namespace = 'rbnz'
ORDER BY table_name;

The table is created/upserted by infra/refresh_snowflake.py at the end of each pipeline run (Glue eolas.* properties → MERGE; schema EOLAS_META granted to the share).

2. REST API headers (same share, different access path)

GET /v1/datasets/{name}/data (and /changes, /v1/bulk/...) return:

X-Eolas-Attribution — full citation string
X-Eolas-Source, X-Eolas-Licence, X-Eolas-Source-URL, X-Eolas-Namespace

JSON clients can optionally pass ?envelope=1 to receive {"data_sources":[{...}], "data":[...]} in the body. See Getting started §5.

3. Bulk downloads

Pre-built Parquet/CSV artifacts include a NOTICE.txt sidecar with identical text.

Dataset-level metadata (GET /v1/datasets/{name}) also exposes source, licence, attribution_text, and source_url fields when stamped in Glue.

The 35 namespaces: rbnz, stats_nz, treasury, oecd, linz, statsnz_geo, lris, nzta, doc, mbie, msd, acc, police, edcounts, worksafe, charities, immigration, pharmac, eeca, geonet, plus the regional-council namespaces (akl_council, akl_transport, bay_of_plenty, colab_waikato, ecan_canterbury, hawkes_bay, manawatu_whanganui, napier_whanganui, northland, otago, southland, taranaki, top_of_south, wellington, west_coast).

5. Query

Native SQL — no special syntax:

-- NZ Official Cash Rate, last 12 observations
SELECT date, cash_rate_official_cash_rate_ocr
FROM eolas.rbnz.rbnz_b2_wholesale_rates_daily
ORDER BY date DESC
LIMIT 12;

Spatial datasets carry a geometry_wkt column (WGS84 WKT). Parse it with Snowflake's geospatial functions:

SELECT name, TO_GEOGRAPHY(geometry_wkt) AS geom
FROM eolas.linz.nz_parcels
LIMIT 100;

dbt users: point a source at the eolas database and reference tables as {{ source('rbnz', 'rbnz_b2_wholesale_rates_daily') }}.

6. Freshness

The share is live — there is no copy step, so the moment eolas refreshes a dataset it's visible in your account. Each dataset refreshes on its source's cadence (daily for FX/rates, monthly/quarterly for most official statistics, etc.); the per-dataset cadence and "data as of" are on each dataset page and in the API metadata (refresh_cadence, last_refreshed_at). New datasets we add appear in the share automatically.

Licensing

Shared datasets carry their source licence and required attribution in the dataset metadata. Most NZ government data is CC-BY 4.0 (attribution required). A small number of sources are excluded from the share for licensing reasons — those aren't granted and won't appear.

7. Offboarding

To stop, simply DROP DATABASE eolas; in your account — nothing of ours remains, since the share is zero-copy. Tell us and we'll remove your account from the share access list.

Questions, a non-Sydney account, or a custom dataset request? phil@eolas.fyi.

Snowflake share

1. Prerequisites

2. Request the share

3. Mount the share

4. What's in the share (layout)

Licence and attribution

5. Query

6. Freshness

7. Offboarding