Emit exciption for number parsing error

[sanbee]

The code in this branch intercept number parsing error from the parser and throw a C++ exception.

Two specializations of the clError class -- clBoolParsingError and clNumParsingError are also defined. These are thrown for error in parsing bool, float or int type values. The exceptions are thrown in interactive as well non-interactive (CLI) modes.

The baseclass ErrorObj is now a derivative of the std::exception class.

The exception generation mechanism is not optimal yet, till I figure out some cmake'ism I don't quite understand.

[agawatw]

Checks that have been done:

1. I built the latest LibRA ardg/main with this parafeed branch and all LibRA unit tests passed.
2. I manually tested the roadrunner built with this parafeed, and errors are thrown as expected when

• a bool argument is given a non-bool value.
• a float/int argument is given a non-float/int value.

Issues:
Current revision does not handle incorrect input for keywords that have factory supplied options. For example, when gridder=100, roadrunner prints

roadrunner>gridder=100 ###Warning: Value did not match any factory supplied options for keyword "gridder" ###Informational: Valid options are "awphpg" "awproject" roadrunner>inp vis = imagename = datacolumn = sowimageext = complexgrid = 100 imsize = 4000 cell = 0 stokes = I reffreq = 3.0e9 phasecenter = weighting = 100 wprojplanes = 1 gridder = 100
It still sets gridder=100.

[agawatw]

The issue mentioned above is probably more of a developer's choice, instead of a bug. For example, I tested this roadrunner with

1. weighting=wrong and even though parafeed does not throw an error, roadrunner does.
2. gridder=wrong neither parafeed nor roadrunner threw an error and the run completed successfully.

@sanbee It's up to you to decide how you'd like to handle these incorrect inputs to the keywords. Making parafeed throw an error may be the easiest.

[sanbee]

I _think_ parafeed is emitting exceptions (of "Informational" or "Warning" type) for a keyword setting that has factory defined options. That's what roadrunner is capturing and reporting. I'll check and make sure this is the case -- in which case your first comment is covered (parafeed emits exceptions). If the exception type is changed (in the library) to FATAL type, the keyword value won't be changed. This is a parafeed "feature" or design decision -- allow users to hang themselves if they want, but at least give a warning.

Your second comment about setting weighting=wrong I don't see. I just tested roadrunner with this setting and it did report the problem (i.e. parafeed emitted and exception that roadrunner caught and reported as warning). Same with the gridder keyword.

sbhatnag@dhruva>/home/dhruva/disk1/sanjay/Packages/LibRA/libra/install/bin/roadrunner 
vis                  = 
imagename            = 
modelimagename       = 
datacolumn           = 
sowimageext          = 
complexgrid          = 
imsize               = 0
cell                 = 0
stokes               = I
reffreq              = 3.0e9
phasecenter          = 
weighting            = natural
wprojplanes          = 1
gridder              = awphpg
cfcache              = 
mode                 = residual
wbawp                = 1
field                = 
spw                  = *
uvrange              = 
pbcor                = 1
conjbeams            = 1
pblimit              = 0.001
usepointing          = 0
roadrunner>weighting=wrong
###Warning:       Value did not match any factory supplied options for keyword "weighting"
###Informational: Valid options are "natural" "uniform" "briggs" 
roadrunner>gridder=wrong
###Warning:       Value did not match any factory supplied options for keyword "gridder"
###Informational: Valid options are "awphpg" "awproject" 
roadrunner>

parafeed does not issue exceptions in the help=noprompt mode. The reason for it is deeper in the design of the library. Basically, in the CLI mode of the application, parafeed does not have factory setting information. And therefore can't issue exceptions. Having said that, I'll look into what it may take to have the factory setting information also available (to the internals of the library) in this mode as well. Seems useful.

[agawatw]

Question about parafeed does not issue exceptions in the help=noprompt mode has been addressed and Sanjay will look into how the factory setting information can become available to the internals of the library in this mode at a later time.

The rest of the changes seems good to be merged.