I checked the extent of the tiles, and 'NHS-D0309_50M_E25833.tif' has a different extent that is incompatible with the other tiles. All the other tifs are lined up with coordinates ending in 00, 25, 50 and 75; 'NHS-D0309_50M_E25833.tif' does not.
This explains the not aligned-stuff, and if I remove that tile, no misalignment notices are given. This does not solve my problem regarding the data looking strange in QGIS, unfortunately. I still get an image like the one in merge_raster2pgsql.png. I also tried with a (more) recent version of PostgreSQL/PostGIS and raster2pgsql, but the issue is still there: # SELECT version(); PostgreSQL 16.3 [..] # SELECT PostGIS_full_version(); POSTGIS="3.4.2 POSTGIS_REVISION" [EXTENSION] (liblwgeom version mismatch: "3.4.2 c19ce56") PGSQL="160" GEOS="3.13.0-CAPI-1.19.0" (compiled against GEOS 3.12.2) PROJ="9.5.1 NETWORK_ENABLED=OFF URL_ENDPOINT=https://cdn.proj.org USER_WRITABLE_DIRECTORY=/var/lib/postgres/.local/share/proj DATABASE_PATH=/usr/share/proj/proj.db" GDAL="GDAL 3.10.0, r eleased 2024/11/01" LIBXML="2.13.5" LIBJSON="0.18" LIBPROTOBUF="1.5.0" WAGYU="0.5.0 (Internal)" (core procs from "3.4.2 c19ce56" need upgrade) RASTER (raster procs from "3.4 .2 c19ce56" need upgrade) $ raster2pgsql RELEASE: 3.4.2 GDAL_VERSION=310 (POSTGIS_REVISION) Another observation is that the choice of tile size varies (I'm using -t auto): With vrt as input, always uses 128x128: INFO: Using computed tile size: 128x128 With *.tif as input the tile size varies depending on which tile is processed first, e.g.: INFO: Using computed tile size: 215x274 or INFO: Using computed tile size: 97x46 or INFO: Using computed tile size: 18x10 Just testing and thinking out loud! Best, Andreas On Sun, Dec 15, 2024 at 12:35 AM Regina Obe <[email protected]> wrote: > Sorry haven’t had a chance to look thru this yet. Going to be tied up for > the next two weeks so I might not get a chance until much later. > > > > If any others are interested in investigating, please don’t wait for me. > > > > Thanks, > > Regina > > > > *From:* Andreas B <[email protected]> > *Sent:* Friday, December 13, 2024 6:48 AM > *To:* [email protected] > *Subject:* Re: Importing tiles with raster2pgsql > > > > Regina, > > > > I'm linking to a folder with a zip containing a subset of tiles that > demonstrates my problem, a script to run the commands, and two screenshots. > Note that I'm a PostGIS beginner, so it's possible I'm doing something > wrong! > > The issues encountered differs a bit from what I described yesterday, but > the essence is the same. > > > > Link to folder: > https://www.dropbox.com/scl/fo/bk8cwktahry0oj0ab2rsz/AHdTD73unb4Efxtgf7exfm0?rlkey=esjtw2iyzysan2hdqklwmj5e3&st=hupv5tsr&dl=0 > > > > Best, > > Andreas > > > > On Thu, Dec 12, 2024 at 3:26 PM Regina Obe <[email protected]> wrote: > > Andreas, > > > > Yes I would expect them to give the same result. > > > > What version of raster2pgsql are you running? It should tell you if you > run raster2pgsql without any args. > > > > Also what platform are you on? Any chance you have some of those tiles > available so we can check it out? > > > > *From:* Andreas B <[email protected]> > *Sent:* Thursday, December 12, 2024 6:17 AM > *To:* [email protected] > *Subject:* Importing tiles with raster2pgsql > > > > Hi all, > > I have a dataset with several hundred tiles (GeoTIFFs) that make up a > digital terrain model. > Each tile is 0.5d x 0.5d. > > I tried to load these GeoTIFFs to PostGIS with: > > $ raster2pgsql -s 32631 -I -M -F -C -t auto -d -l 2,4,8,16,32 tif/*.tif (1) > > The data was loaded, but when viewing in QGIS, it looked like jagged lines > with areas of white overlain by pieces of neighboring tiles, etc. > This didn't look right. > I also noticed that for each tif, the constraints were printed. > I expected this to be done once, after the last tile was loaded. > > I then created a virtual raster with gdalbuildvrt: > > $ gdalbuildvrt merge.vrt tif/*.tif (2) > > And then used raster2pgsql to load in the vrt: > > $ raster2pgsql -s 32631 -I -M -F -C -t auto -d -l 2,4,8,16,32 merge.vrt | > psql (3) > > The data was loaded, and looked good in QGIS. > > Shouldn't commands (1) and (3) give the same results, or am I > misunderstanding? > > Best, Andreas > >
