Take the file.
Use the fix.

01 / Prepared geographic data

Delhi's 11 districts + Jammu district

Complete shapefile packages with geometry, attributes, projection and encoding sidecars. Extract the ZIP, keep the five matching files together, and open the .shp in your GIS.

12 geometries checked. Every coordinate and attribute compared with the official source.

Delhi onlyJammu onlyRead instructionsVerification

Source: National Water Data Portal / NWIC, District Boundary, SHP resource dated 1 May 2025. The dataset names GSI as producer; original records identify SOI, and those fields are preserved. Prepared 27 August 2026. This is the requested 11-district Delhi vintage, not a new boundary survey.

The original India NSF Lambert Conformal Conic projection is retained: coordinates are metres, not latitude/longitude. Reproject in your GIS if required. Shared with source acknowledgement under the portal's reproduction policy; source terms remain applicable.

Block-level follow-up: reconciled, not relabeled

Jammu's current LGD directory contains 20 development blocks. The official public polygon service also contains 20 Jammu geometries, but it is explicitly an LGD-2017 layer: 18 block codes align, two current blocks lack matching polygons, and two old polygons are absent from today's directory. Delhi currently returns zero development blocks, so its nearest official lower-level geometry is the 36-feature subdistrict layer.

56 geometries packaged. Current directory rows, shapefiles, GeoJSON, provenance, hashes, and every source mismatch are included.

Current directory: Ministry of Panchayati Raj LGD. Block geometry: Government of India Map Service, Block layer. Delhi subdistrict geometry: National Water Data Portal. The package does not claim that LGD-2017 geometry is a fully current 2026 boundary survey.

02 / Data collector correctness

Missing air-quality readings are not zero

A small CSV collector patch for the Goa AQI logger. It keeps real zero, decimal and scientific-notation readings numeric while leaving missing, blank and malformed values empty.

26 adversarial cases × 11 fields: zero incorrect cells after the fix. Repeat runs add no duplicate rows.

Includes the helper, minimal patch, instructions and test receipt. Node.js 20+, no npm dependencies. Tests exercise the actual pinned collector using synthetic API responses; they do not claim historical live readings were corrupt. This patch addresses CSV values, not the separate GeoJSON snapshot.

03 / Reproducible reconciliation

Put closing-stock movements in the right window

A minimal date-window correction for the public Goa mining reconstruction. It avoids counting pre-baseline movements twice and includes the whole closing day.

166 reconstructed balances matched independent date arithmetic. Ten boundary tests passed.

Includes all four changed before/after results from the inspected public snapshot and a timestamp audit of 396,563 trips. The test runs on an in-memory copy, leaving the input database unchanged. Python standard library only; source code and approximately 100 MB public database are downloaded separately.

The correction treats closing stock as end-of-day. Missing baselines, aliases and source completeness remain separate questions. In particular, the royalty stream has no 2024 baseline. Negative reconstruction is not by itself evidence of physical shortage or wrongdoing.