Insights about project management, GIS, and how technology is transforming industries operating in the physical world.
Without a .prj, nothing knows what the numbers mean: GDAL doesn't print the CRS, QGIS asks, and the viewer assumes WGS 84 and warns about it. How to tell degrees from UTM meters by the numbers, why an exported longitude of -2171693 is actually your easting, and the two ogr2ogr commands that fix it.
No .shx and GDAL refuses while the browser opens the file; no .dbf and everything opens with zero attributes and no warning; no .shp and there is nothing to open. What each Shapefile file is, whether it can be rebuilt, and what the viewer and ogrinfo say for each missing one, measured on IBGE data.
The dBase header gives a Shapefile field name 11 bytes, one a terminator: 10 characters. GDAL warns and cuts, colliding names become nome_da_un, nome_da__1 and nome_da__2, and the viewer shows the short one. On GDAL 3.13.3 a complete .shp.xml restores the name; a partial one is ignored silently.
The .dbf has no encoding of its own; a five-byte .cpg names one. Without it every reader guesses, and GDAL and the browser guess differently: Latin-1 read as UTF-8 gives Amap�, UTF-8 read as Latin-1 gives Amapá. Both reproduced, with the one-line .cpg fix, the ogr2ogr recode and the trap.
Collect and edit in GeoPackage, publish to the web in GeoJSON, send a Shapefile only when the other side insists. The evidence: the IBGE 27-state mesh converted three ways and measured on eight criteria, from bytes loose and zipped to the .cpg encoding trap, the EPSG code and viewer load time.
A Shapefile is five files, one shape type and a .dbf from 1998. We built them with GDAL 3.13.3, deleted the sidecars one by one, read the .dbf bytes and quoted every error: names cut at 10 characters, Latin-1 by default, NULL equals empty, the .prj that puts your layer in the ocean, and the fixes.
Zip the .shp, .shx, .dbf and .prj, drop the zip on a free browser viewer, and the layer draws with its attribute table. Step by step, then the three things that go wrong — no .prj, accents mangled, two Shapefiles in one zip — each reproduced with GDAL 3.13.3 and the live viewer, outputs quoted.