grdmath does not respect registration (-rp) when using gdal:GTIFF

[anbj]

When output is =gdal:GTIFF, -rp is not respected. Is this documented?

Creating the default .nc-file, things work:

$ gmt grdmath -rp -I0.1 -R0/1/0/1 0 = trash.nc

$ gmt grdinfo trash.nc 
trash.nc: Title: Produced by grdmath
trash.nc: Command: gmt grdmath -rp -I0.1 -R0/1/0/1 0 = trash.nc
trash.nc: Remark: 
trash.nc: Pixel node registration used [Cartesian grid] <------------------------- pixel
trash.nc: Grid file format: nf = GMT netCDF format (32-bit float), CF-1.7
trash.nc: x_min: 6.93889390391e-18 x_max: 1 x_inc: 0.1 name: x n_columns: 10
trash.nc: y_min: 6.93889390391e-18 y_max: 1 y_inc: 0.1 name: y n_rows: 10
trash.nc: v_min: 0 v_max: 0 name: z
trash.nc: scale_factor: 1 add_offset: 0
trash.nc: format: classic
trash.nc: Default CPT: 

Use =gdal:GTIFF, and -rp is ignored:

$ gmt grdmath -rp -I0.1 -R0/1/0/1 0 = trash.tif=gdal:GTIFF

$ gmt grdinfo trash.tif
trash.tif: Title: Grid imported via GDAL
trash.tif: Command: 
trash.tif: Remark: 
trash.tif: Gridline node registration used [Cartesian grid] <----------------------gridline, despite -rp
trash.tif: Grid file format: gd = Import/export through GDAL
trash.tif: x_min: 0.05 x_max: 0.95 x_inc: 0.1 name: x n_columns: 10
trash.tif: y_min: 0.05 y_max: 0.95 y_inc: 0.1 name: y n_rows: 10
trash.tif: v_min: 0 v_max: 0 name: z
trash.tif: scale_factor: 1 add_offset: 0
trash.tif: Default CPT: 

[mkononets]

trash.tif: x_min: 0.05 x_max: 0.95 x_inc: 0.1 name: x n_columns: 10
trash.tif: y_min: 0.05 y_max: 0.95 y_inc: 0.1 name: y n_rows: 10

this is effectively pixel registration

NB see also gdalinfo boundaries

$ gdalinfo trash.tif 
Driver: GTiff/GeoTIFF
Files: trash.tif
Size is 10, 10
Origin = (0.000000000000000,1.000000000000000)
Pixel Size = (0.100000000000000,-0.100000000000000)
Image Structure Metadata:
  COMPRESSION=DEFLATE
  INTERLEAVE=BAND
Corner Coordinates:
Upper Left  (   0.0000000,   1.0000000) 
Lower Left  (   0.0000000,   0.0000000) 
Upper Right (   1.0000000,   1.0000000) 
Lower Right (   1.0000000,   0.0000000) 
Center      (   0.5000000,   0.5000000) 
...

[anbj]

Maybe you're right, but when I tried this for an actual problem, grdmath would not do the operation, complaining 'grids have different size' (pseudophrasing). Maybe GMT/grdmath misreads the tif as gridline registered, even though it is pixel registered?

[mkononets]

when I tried this for an actual problem, grdmath would not do the operation, complaining 'grids have different size' (pseudophrasing). Maybe GMT/grdmath misreads the tif as gridline registered, even though it is pixel registered?

Did you report the actual problem?

[anbj]

No, I constructed something that looked like it captured the problem, but maybe it didn't.
Anyway, gmt reports a file created with -rp as gridline registered, so something must not be done as it is supposed to.

[joa-quim]

trash.tif: x_min: 0.05 x_max: 0.95 x_inc: 0.1 name: x n_columns: 10
trash.tif: y_min: 0.05 y_max: 0.95 y_inc: 0.1 name: y n_rows: 10

this is effectively pixel registration

Nope. That is grid registration.

[mkononets]

Right! Without -rp gmt math produces an 11x11 grid. With -rp the grid size is 10x10.

It looks like gdal does not respect gmt's concept of gridline registration, assuming pixel reg in both cases.

[joa-quim]

GDAL reports everything in pixel registration. Even when input is grid registered (but it does the conversion, so result is not wrong, just gdalified)

[anbj]

So as soon as you touch anything that involves gdal, e.g. =gdal:<driver>, things become gdalified? If GDAL reports everything in pixel registration, why does grdinfo report gmt grdinfo trash.tif as gridline? Doesn't this go through GDAL?

[joa-quim]

From memory, grids, when coming from GDAL, are always set to grid and images in pixel. So in fact, grids are always *desgdalified* Enviado a partir do Outlook para iOS<https://aka.ms/o0ukef>

…

________________________________ De: anbj ***@***.***> Enviado: Friday, April 17, 2026 9:56:02 PM Para: GenericMappingTools/gmt ***@***.***> Cc: Joaquim Manuel Freire Luís ***@***.***>; Comment ***@***.***> Assunto: Re: [GenericMappingTools/gmt] grdmath does not respect registration (-rp) when using gdal:GTIFF (Issue #8988) [https://avatars.githubusercontent.com/u/29237824?s=20&v=4]anbj left a comment (GenericMappingTools/gmt#8988)<#8988 (comment)> So as soon as you touch anything that involves gdal, e.g. =gdal:<driver>, things become gdalified? If GDAL reports everything in pixel registration, why does grdinfo report gmt grdinfo trash.tif as gridline? Doesn't this go through GDAL? — Reply to this email directly, view it on GitHub<#8988?email_source=notifications&email_token=AAEDF2KOLIHJIWDQJJJB45D4WKK6FA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMRXGEYTQOJRGA32M4TFMFZW63VHMNXW23LFNZ2KKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-4271189107>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AAEDF2IDI4G6UIB3QHVZLET4WKK6FAVCNFSM6AAAAACX4Y4RRKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DENZRGE4DSMJQG4>. You are receiving this because you commented.Message ID: ***@***.***>

[anbj]

Thanks for the insight, @joa-quim. So no bugs, closing.

[joa-quim]

No, this is a bug. -r is not being respected