mirror of
https://github.com/apache/superset.git
synced 2026-05-24 17:25:20 +00:00
1817e37b06
Move build pipeline outputs from a sibling `output/` directory into
Superset's Flask-served `superset/static/assets/country-maps/` so the
plugin can fetch them at runtime without webpack involvement (Flask
serves the tree at `/static/...` directly).
Commit the 220 generated GeoJSONs + manifest.json so a fresh
ephemeral environment can render the chart immediately, no build
step required at deploy time. Trade-off: ~17 MB of generated files
in-tree. (For comparison the legacy plugin commits ~34 MB of
GeoJSON; net change is -17 MB once we remove the legacy plugin in a
future major version.)
Files committed:
superset/static/assets/country-maps/
README.md (humans, not data)
manifest.json
ukr_admin0.geo.json 2.1 MB
ukr_admin1_<adm0_a3>.geo.json × 214 ~50 KB - 662 KB each
regional_<country>_<set>_ukr.geo.json × 4 ~30 KB each
composite_france_overseas_ukr.geo.json 322 KB
Build script changes:
- OUTPUT_DIR computed via SCRIPT_DIR.parents[3] / "superset" /
"static" / "assets" / "country-maps"
- mkdir(parents=True) so a fresh checkout works first run
- Stale `output/` entry kept in scripts/.gitignore for safety
(some local checkouts may have it from earlier iterations)
.gitignore: add re-include lines so superset/static/assets/country-maps/**
gets committed despite the broader superset/static/* exclusion.
SIP_DRAFT updated with the hosting-decision rationale.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Country Map data pipeline
This directory contains the build pipeline that turns upstream Natural Earth data into the GeoJSON files consumed by @superset-ui/plugin-chart-country-map.
It replaces the legacy scripts/Country Map GeoJSON Generator.ipynb notebook. See SIP_DRAFT.md in the parent directory for the full design rationale.
Layout
scripts/
build.sh # one-shot reproducible build
README.md # this file
config/ # declarative YAML — handles ~95% of fixes
name_overrides.yaml # typos, deprecated ISO codes, admin renames
flying_islands.yaml # repositioning + bbox drops for far-flung territories
territory_assignments.yaml # add features from sibling Admin 0 records
regional_aggregations.yaml # dissolve Admin 1 into administrative regions
composite_maps.yaml # multi-country composites (e.g. France-with-Overseas)
procedural/ # escape hatch — handles the rare 5%
README.md # when to use, when not
NN_<descriptive_name>.py # one focused script per genuine edge case
output/ # gitignored — build artifacts
Operating principles
- Default tool: declarative YAML. Most touchups are renames, repositions, dissolves, or filters — all expressible in YAML. Diffs are small, conflicts localize cleanly to one entry, contributors can submit "fix typo X" as a one-line PR.
- Escape hatch:
procedural/directory of small, named, single-purpose Python scripts for the rare cases YAML can't express cleanly. Each script has a header comment explaining why it's not in YAML. Seeprocedural/README.mdfor the bar. - Build is reproducible from a pinned NE version.
build.shrecords the NE git SHA it consumed; outputs are deterministic given inputs. - CI regenerates on schema change and opens a PR if outputs differ. Maintainers review the cartographic diff in legible GeoJSON, not opaque notebook JSON.
Workflow for adding a fix
- Identify the upstream NE issue (wrong name, missing territory, etc.).
- Try YAML first. Add the smallest possible entry to the appropriate config file with a
descriptionfield explaining the fix. - If YAML can't express it cleanly, add a numbered script in
procedural/with a header comment explaining why YAML didn't fit. - Run
build.shlocally, verify the output GeoJSON looks right. - Open PR. Reviewer sees the YAML diff (or new procedural script) plus the regenerated GeoJSON.
See also
SIP_DRAFT.md(parent dir) — design rationale, notebook audit, obsolescence checkprocedural/README.md— when to use the escape hatch