textread() barfs silently when file contains "nan" or "inf"

[John W. Eaton]

On 13-Aug-2008, Rob Mahurin wrote:

| running octave 3.0.1.
| 
| file "bug.txt" contains
| 0.090   015.0   013.9   20.456  15.3822 60.7908
| 0.088   015.0   NaN     20.486  15.3438 60.7438
| 0.086   015.0   013.8   20.494  15.2289 60.5916
| 
| from octave:
| [c1,c2,c3,c4,c5,c6] = textread("bug.txt","%f %f %f %f %f %f");
| [c1,c2,c3,c4,c5,c6]
| ans =
| 
|      0.090000   15.000000   13.900000   20.456000   15.382200    
| 60.790800
|      0.088000   15.000000   15.000000   15.000000   15.000000    
| 15.000000
|      0.086000   15.000000   13.800000   20.494000   15.228900    
| 60.591600
| 
| So, reading "nan" or "inf" (or other invalid input) stops processing  
| for the current line without throwing an error.
| 
| Quick fix: detect and complain about error
| Better fix: support nan and inf, document
| 
| I forget whether textread() is an octave-forge function, and whether  
| this is the right mailing list if so.  Sorry.

It is part of the Octave Forge io package.

jwe