Lets copy XMP metadata #3050
Lets copy XMP metadata #3050
Conversation
Implements in GDALDrive::DefaultCreateCopy and in GTiff driver CreateCopy (indirectly also available in COG driver) gdal_translate has a new option '-noxmp' to not copy the XMP
gdal/frmts/gtiff/geotiff.cpp
Outdated
| @@ -18010,6 +18010,15 @@ GTiffDataset::CreateCopy( const char * pszFilename, GDALDataset *poSrcDS, | |||
| } | |||
| } | |||
|
|
|||
| /* -------------------------------------------------------------------- */ | |||
There was a problem hiding this comment.
This will likely result in a suboptimal file. This will cause likely the TIFF directory to be rewritten not at the beginning of the file. You could check that with tiffdump. If things are written optimally you should see "Directory 0: offset 8" towards the beginning of the output.
I believe you'd need to manually write the TIFF tag before the /* Cleanup */ of line 17796 with
something like
char **papszXMP = poSrcDS->GetMetadata("xml:XMP");
if( papszXMP != nullptr && *papszXMP != nullptr )
{
int nTagSize = static_cast<int>(strlen(*papszXMP));
TIFFSetField( m_hTIFF, TIFFTAG_XMLPACKET, nTagSize, *papszXMP );
}
There was a problem hiding this comment.
You could add a check for that in autotest with:
assert int(ds.GetRasterBand(1).GetMetadataItem('IFD_OFFSET', 'TIFF')) == 8There was a problem hiding this comment.
Thanks @rouault
Before fixing the code I added the assert you mentioned... and it does not fail.
This is the output of tiffdump
> tiffdump ./gcore/tmp/test_181_copy.tif
./gcore/tmp/test_181_copy.tif:
Magic: 0x4949 <little-endian> Version: 0x2a <ClassicTIFF>
Directory 0: offset 758 (0x2f6) next 0 (0)
ImageWidth (256) SHORT (3) 1<20>
ImageLength (257) SHORT (3) 1<20>
BitsPerSample (258) SHORT (3) 1<8>
Compression (259) SHORT (3) 1<1>
Photometric (262) SHORT (3) 1<1>
StripOffsets (273) LONG (4) 1<358>
SamplesPerPixel (277) SHORT (3) 1<1>
RowsPerStrip (278) SHORT (3) 1<20>
StripByteCounts (279) LONG (4) 1<400>
PlanarConfig (284) SHORT (3) 1<1>
SampleFormat (339) SHORT (3) 1<1>
700 (0x2bc) BYTE (1) 671<0x3c 0x3f 0x78 0x70 0x61 0x63 0x6b 0x65 0x74 0x20 0x62 0x65 0x67 0x69 0x6e 0x3d 0x22 0xef 0xbb 0xbf 0x22 0x20 0x69 0x64 ...>
33550 (0x830e) DOUBLE (12) 3<60 60 0>
33922 (0x8482) DOUBLE (12) 6<0 0 0 440720 3.75132e+06 0>
34735 (0x87af) SHORT (3) 32<1 1 0 7 1024 0 1 1 1025 0 1 1 1026 34737 21 0 2049 34737 6 21 2054 0 1 9102 ...>
34737 (0x87b1) ASCII (2) 28<NAD27 / UTM zone 11N|NAD ...>
What is wrong with the assert?
There was a problem hiding this comment.
I think I found the reason. I have to close and open the file to let the code to dump all the data.
There was a problem hiding this comment.
Or calling FlushCache(). Any preference?
There was a problem hiding this comment.
Any preference?
after close & reopen. The semantic of FlushCache() is that it should push 'dirty' blocks from the block cache to disk, but there's no guarantee that other 'structural' elements of the file format will be flushed as well. Only close() guarantees that the dataset is properly finalized
There was a problem hiding this comment.
I have added the assert, and _check_cog in cog.py where the ghost area makes the size different. (and some other minor improvements)
Check that directory offset of first band is 8 bytes in normal GTiff Improve tests, including COG test
|
There are two failing tests in Travis CI. They seem to be not related with my changes. What should I do? |
I've restarted them |
Add feature to copy XMP metadata in
CreateCopy, in particular forGTiffdriver (COGindirectly), and all drivers that useDefaultCreateCopymethod.gdal_translateadds the option-noxmpto not copy this data when desiredTasklist