FIGARO (2026) (part 2)

README.md

@@ -84,7 +84,7 @@ Redistributed by OptimalSolution LLC -- see `DISCLAIMER.md`. Redistribution only
 - The axes are not purely industries and products. ind_use also carries final demand (P3_S13/S14/S15, P51G, P5M) and prd_ava also carries value added (B2A3G, D1, D21X31, D29X39, OP_RES, OP_NRES). Summing an axis without separating these adds final demand to industry output.
 - Reporting areas are Eurostat geo codes: EL is Greece, UK is the United Kingdom and WRL_REST is the rest of the world. scales/figaro_geo.csv maps them to ISO3 where one exists.
 - DOM ('Domestic country') in c_orig is NOT a partner country and does NOT duplicate the self-pair. Verified on the 2020 use table: it appears on exactly the six value-added rows (B2A3G, D1, D21X31, D29X39, OP_RES, OP_NRES) and on nothing else, with zero overlap against the 64 products -- value added has no foreign origin. So c_orig is additive; it is the PRODUCT axis that is not, because summing all 70 codes adds value added to intermediate inputs.
-- Datasets above 10M rows are written hive-partitioned by year. energyRt's en_open_dataset() CANNOT read a hive layout -- it lists non-recursively, so year=2010/ has no extension to sniff -- so getDataset() works only for the single-part datasets. Each dataset.yml records this as storage.energyrt_readable.
+- Datasets above 10M rows are written hive-partitioned by year, the rest as a single part. `storage.layout` says which, and a hive dataset must be opened as a dataset (with its year partitioning), not as one file.
 
 ## Using it
 

datasets/figaro_land_footprints/dataset.yml

@@ -94,5 +94,4 @@ bank:
     parts: 1
     compression: zstd
     bytes: 2955
-    energyrt_readable: yes
     part_sha256: ceeb54fd60394bc29e6b3f5b2046a6546c97b5ce0731adbd3c776429bb10e957

datasets/figaro_supply/dataset.yml

@@ -151,5 +151,4 @@ bank:
     parts: 1
     compression: zstd
     bytes: 2974684
-    energyrt_readable: yes
     part_sha256: 08a6873647926a635cd27d9e255e5afd407a8ed52aa5ad378b548718c9884d7a

datasets/figaro_use/dataset.yml

@@ -160,5 +160,4 @@ bank:
     parts: 15
     compression: zstd
     bytes: 142749397
-    energyrt_readable: no
     part_sha256: ~

source.yml

@@ -73,7 +73,6 @@ notes:
   overlap against the 64 products -- value added has no foreign origin. So c_orig
   is additive; it is the PRODUCT axis that is not, because summing all 70 codes adds
   value added to intermediate inputs.'
-- Datasets above 10M rows are written hive-partitioned by year. energyRt's en_open_dataset()
-  CANNOT read a hive layout -- it lists non-recursively, so year=2010/ has no extension
-  to sniff -- so getDataset() works only for the single-part datasets. Each dataset.yml
-  records this as storage.energyrt_readable.
+- Datasets above 10M rows are written hive-partitioned by year, the rest as a single
+  part. `storage.layout` says which, and a hive dataset must be opened as a dataset
+  (with its year partitioning), not as one file.