[gdal-dev] Is source Dataset array to `GDALWarp` mutable?

[Even Rouault]

Le 09/01/2024 à 19:45, Fitch, Simeon via gdal-dev a écrit :
> I have a question about the mutability of the `pahSrcDS` argument to 
> `GDALWarp`.
>
> The `GDALWarp` function has the following signature:
>
> ```
> GDALDatasetH GDALWarp(const char *pszDest, GDALDatasetH hDstDS, int 
> nSrcCount,
>                       GDALDatasetH *pahSrcDS,
>                       const GDALWarpAppOptions *psOptionsIn, int 
> *pbUsageError)
> ```
> Source: 
> https://github.com/OSGeo/gdal/blob/895e9fbd63b2aa22471f4f8c69efd5ba4e700e9c/apps/gdalwarp_lib.cpp#L1379-L1381
>
> I was curious to see that `pahSrcDS` was not `const GDALDatasetH 
> *pahSrcDS` or `const GDALDatasetH * const pahSrcDS`.
>
> What mutability and ownership conclusions should I draw from it (if any)?
>
> * Might the `GDALWarp` mutate the array and/or the pointed Datasets?
No. The const correctness could have indeed be done better (although a 
reason for that is that in C, a  "T**" is not implicitly cast as "const 
T* const*", contrary to C++)
> * Is it correct to assume the caller maintains ownership of the 
> elements of `pahSrcDS`?

Yes

Note that if the output format is VRT, then the return VRT dataset will 
reference the input dataset handle (a single one is allowed in that 
case). Hence, in that particular case (works also in the general case), 
you should close first the output dataset and then source dataset(s), as 
done by gdalwarp_bin.cpp

Even

-- 
http://www.spatialys.com
My software is free, but my time generally not.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20240109/569a40d6/attachment.htm>