Emissions

[louispt1]

Context

Adds multi-year emissions support with 1990 baseline tracking and aggregation capabilities.
Also covers multi-line group handling

Implemented changes

• Added year column to emissions CSV schema
• Implemented Qernel::Emissions#sum() for subsector aggregation
• Updated EMISSIONS() GQL function to aggregate across subsectors
• Enhanced ScopedSector with year-aware scoping
• Updated atlas gem dependency (includes emissions schema changes)

Related

Goes with pull requests:

• Emissions #199
• Emissions etengine#1779
• Direct emissions chart and dashboard items, mapping file, 1990 data etsource#3463
• Direct emissions chart and dashboard items etmodel#4753
• https://github.com/quintel/etdataset/pull/1095
• Direct emissions chart and dashboard items documentation#343

Checklist

• I have tested these changes
• I have updated documentation as needed
• I have tagged the relevant people for review

[mabijkerk]

I had AI do a full review of the branches, it flagged the following:

Etsource::Sectors#build_node_index normalizes use but not sector_label
(etsource/sectors.rb):

pair = [node.sector_label, Atlas::SectorMapping.normalize(node.use)]

Atlas's row_pair normalizes both sides. This works today because sector_label is declared Symbol and every .ad happens to use a slug already — but a label written with a capital, space, or dash produces a pair matching nothing, and the node then silently vanishes from SECTOR(...) results and from the CSV export with no error. Normalizing both sides costs nothing.

[kndehaan]

Also here, I think this finding should not block the merging of the PR. I'll discuss this point with @louispt1 next week.

[louispt1]

I had AI do a full review of the branches, it flagged the following:

Etsource::Sectors#build_node_index normalizes use but not sector_label (etsource/sectors.rb):

pair = [node.sector_label, Atlas::SectorMapping.normalize(node.use)]

Atlas's row_pair normalizes both sides. This works today because sector_label is declared Symbol and every .ad happens to use a slug already — but a label written with a capital, space, or dash produces a pair matching nothing, and the node then silently vanishes from SECTOR(...) results and from the CSV export with no error. Normalizing both sides costs nothing.

The ETSource spec will catch drift between the normalised labels so there is no need to normalise in both locations. It's not a coincidence that every sector_label is written as a slug now, that's a style and code choice and the spec will fail if there's a mistake in that respect