Built on 1.18.3. Offline tests and the QGIS 4.2.2 live checks in CI pass; not hand-tested in desktop QGIS unless an item says so. Full list: the v1.18.4 block in metadata.txt and CHANGELOG.md. Damage score ranks the most damaged district first: calculate_damage_exposure_severity took the average of the SIGNED change (after minus before), so a district that lost value had the lowest number and scored 0 while unchanged districts scored 1. It now averages the size of the change. A second run on the same layer also read the first run's old column (QGIS wrote the new one as chg_mean_2); each run now reads the column it just made. Distances in degrees are caught: find_nearest_features said its distance was in metres but wrote degrees for a latitude/longitude layer; it now also adds distance_m in real metres. The blur tool (radius, grid cell), hotspot_analysis, simplify_geometry and interpolate_surface accepted 1000 on such a layer and used it as 1000 degrees; a value of 1 degree or more is now refused with the conversion (0.01 is about 1.1 km). A named data source gets its own tool: A request such as 'search HDX for health facilities datasets' or 'conflict events from UCDP' matched a 'Map health facilities' or 'Map conflict events' task whose tools could not do it, and a confirmation card for the wrong task appeared. Naming UCDP, GDACS, USGS, EONET, FIRMS or FTS, or searching datasets, now gives no task directive. IPC import names the area: import_humanitarian_table reported problem rows by number only; each issue now carries the area (P-code or name), and admin1_name, admin2_name, adm1_name and adm2_name columns are recognised as the area name. Clearer errors: obfuscate_sensitive_points given a plain table says it has no geometry and how to load points. A GDACS HTTP 400 is retried once with the documented query parameters, and any GDACS failure names the site and the kind of failure without quoting the raw error. Known and open: The elevation-profile print item sometimes renders empty in CI (#379); the cause is not found. In the 1.18.3 hand run the model did not call the survey tool (D3), exported instead of writing the field (D4) and planned an analysis instead of refusing a trigger with no threshold (S5); the tool router already offers those tools, so these are model choices. The change-layer legend (D6) needs a screenshot. Issues #362 and #380 stay open for their hand rows. Not hand-tested: Everything above is checked by the offline suite and, for the QGIS parts, by CI on QGIS 4.2.2; none of it has been hand-tested in desktop QGIS.
yes
baronala
2026-10-27T10:07:33.245287+00:00
4.2.0
4.99.0
None
no
Plugin Tags