i.pca metadata truncated

[Barney2011]

Due to character limits the full metadata of the i.pca module is not recorded to the resulting raster files.

output should read: PC1 1170.12 ( -0.63 -0.65 -0.43 ) [ 88.07% ] instead reads: PC1 13770134.87 ( 0.1019, 0.5087, 0.1896, 0.5185, 0.4578, 0.4527, 0.107

with loss of vector values and % importance.

[hamish]

Hi,

could you upload the $MAPSET/hist/$MAPNAME file as an attachment to this bug report?

also, knowing the exact command used to run the program would help- this is for 6 input maps, right? is it LANDSAT data?

thanks, Hamish

[hamish]

more from Rebecca:

Here is the command (as created through the gui in GRASS7.0)
i.pca input=1_jan_4789.1@PERMANENT,1_jan_4789.10@PERMANENT,1_jan_4789.11@PERMANENT,\
1_jan_4789.12@PERMANENT,1_jan_4789.13@PERMANENT,1_jan_4789.14@PERMANENT,\
1_jan_4789.2@PERMANENT,1_jan_4789.3@PERMANENT,1_jan_4789.4@PERMANENT output=Jan_Pca_test

Metadata file attached.

I'm actually using the pca module for airborne spectral data (14 bands from the CASI2 sensor).

[hamish]

$MAPSET/hist/ metadata file

[hamish]

so it is getting chopped at char 80. by my calculations 14 bands would need ~140 characters.

maps written by grass 7's i.pca should be ok as they don't have the 80x40 limit. although instead of allocating 256 chars for i.pca/support.c's tmpeigen[] we might allocate (bands * 8) + 30 instead. (or perhaps just 1024 and push the problem away a few years)

as for a fix in GRASS 6, I think we need to insert a "\n" then 15 spaces after each 6th band is printed (if there's more to come). so it would look like:

PC14     67.08 ( 0.0188, 0.0040,-0.0002, 0.0096, 0.0023,-0.0056,
                 0.0126,-0.0173,-0.0002, 0.0096, 0.0023,-0.0056,
                 0.0126,-0.0173 ) [ 88.07% ]

but in the case of 14 bands it would then make the history file 50 lines long, and so more than the max 40 lines long, at which the command history would be truncated.

Hamish

[neteler]

At least for GRASS 7 the raster history limitation should be overcome.