[arunissun]

per-map-data was about 9.9 MB uncompressed with 4–6 s before the first byte, and the dashboard used about 6% of it. Only the PER dashboard reads these two endpoints, and the dashboard on develop does not use them yet.

Related: IFRCGo/go-api-artifacts#6 (schema)

per-map-data

• remove results (the dashboard works out the latest process per country itself), processes[].components, the four *_considerations_from_assessment fields and the duplicate assessment_date, lat and lon
• stop loading assessments and component responses, which only those fields needed
• add areaNum to prioritized_components
• skip the parent component through is_parent instead of the hardcoded component id 14

per-dashboard-data

• add area_num, component_letter and is_parent to the component entries, so the dashboard can use the shared PER area colours and show each component 14 sub-component as its own row ("14 A: First Aid")
• sort components by area number, component number and letter

Both endpoints

• remove the translated region_name, type_of_assessment_name and phase_display (the dashboard uses the ids and the global enums)
• gzip the responses with gzip_page;
• cache the data per language: country, component and rating names are translated, and the shared cache key served the first language's names to everyone
• cache the data on the live API too: since refactor(per): consolidate PER dashboard data endpoints #2821 the cache only worked when DJANGO_READ_ONLY was true, so the live API rebuilt the data on every request (per-dashboard-data takes about 17 s on production data). The data is now cached for 1 hour whatever DJANGO_READ_ONLY is set to, so a PER update can take up to an hour to show on the dashboard
• describe both responses in the OpenAPI schema so the web app can use generated types

Tests

• updated the map and dashboard data tests for the new shape
• added tests for the removed fields, the parent component, the component letters, gzip and the per-language cache
• the cache tests run with DJANGO_READ_ONLY=False, as on the live API

Checked locally (read-only, through the SSH tunnel)

Production data:

• per-map-data: 176 processes from 116 countries, 506 KB of JSON, 40 KB sent with gzip
• per-dashboard-data: 51 components, 2,513 KB of JSON, 278 KB sent with gzip; 16.6 s before the first byte without the cache, immediate from the cache
• French responses carry French country names (102 of 176 differ from English), so the per-language cache works
• area_id equals area_num for all 51 components; component 14 comes as the parent plus sub-components 14 A to 14 N, and no parent component appears in prioritized_components
• the web app (PER dashboard go-web-app#2555) was checked in the browser against this API