Why a spectral distribution read from a CSV file cannot be reshaped?

[frankkjensen]

Hi,

I load a spectra with:
sds = colour.read_sds_from_csv_file('Cyan.csv', delimiter='\t', header=None, names=['WL', 'Cyan'])
And then I try to reshape it with:
colour.colorimetry.reshape_sd(sds['Cyan'], colour.SpectralShape(360, 830, 1))
But I get this error:
AssertionError: "start" attribute value must be strictly less than "780"!

What am I doing wrong?

I have attached the csv file:
Cyan.csv

-Frank

[KelSolaar]

Hi @frankkjensen,

I think it is an issue with the current code that now uses Numpy: If you look at the spectral distribution, the first wavelength is nan:

[[             nan   0.00000000e+00]
 [  3.58338471e+02   2.60061000e-04]
 [  3.58638672e+02   9.62139000e-04]
 ..., 
 [  8.31245300e+02   1.45844600e-03]
 [  8.31516602e+02   1.42960800e-03]
 [  8.31787903e+02   9.48240000e-04]]

Try to set the header directly and remove your two calls that set it for the time being:

WL	Cyan
358.038269	0
358.3384705	0.000260061
358.6386719	0.000962139

I will try to understand what is going on!

[KelSolaar]

Confirming that np.genfromtxt is the issue:

import numpy as np
print(np.genfromtxt('/Users/kelsolaar/Downloads/Cyan.csv', delimiter='\t'))

[KelSolaar]

I found out! You have 3 bytes that cannot be decoded at the start of your file, so the first wavelength ends up as nan:

[frankkjensen]

Thanks for looking into this.
The file is from SpecWin (spectrometer software from Instrument Systems). I looked up the "EF BB BF", and it should be the "byte order mark" for UTF-8 encoding, but maybe np.genfromtxt does not support that.
Anyway, as it was only 6 files, I did as you suggested and added the header in the files.

Thank for the support :-)
-Frank