Re: [lmi] Code review: product editor

[Evgeniy Tarassov]

On 3/1/07, Greg Chicares <address@hidden> wrote:
> (1) Class template MultiDimTableTypeTraits duplicates lmi's
> function template value_cast(). I would much rather use the lmi
[...]
> (2) I'd prefer to use standard types instead of wx types where
> possible. IIRC, wx container types are just synonyms for STL
> containers anyway, but the string types are not substitutable.

Sorry for that. MultiDimTableTypeTraits removed and value_cast used
everywhere, where a conversion is needed. wxString is replaced by
std::string where possible. There are however two occurences of
wxString in constructors left untouched for only consistency with wx.
For example rounding_view_editor.hpp
   RoundingButtons
       (wxWindow* parent
       ,wxWindowID id
       ,const wxPoint& pos = wxDefaultPosition
       ,const wxSize& size = wxDefaultSize
       ,long style = 0
       ,const wxString& name = wxPanelNameStr
       );
Do you think it also should be changed to use lmi own types? It left
untouched only for consistency with its wx base class.

A little question which is not related directly to the above.

IIRC it is not advised to add return after a call to fatal_error().
Does it apply to warning() too?
Comments from alert.hpp:
[...]
/// fatal_error: Dire runtime problems that prevent the system from
/// performing a requested action running in any reasonable manner.
/// Generally, an exception would be thrown; a gui might catch it,
[...]
/// warning: Significant runtime problems that should be brought to the
/// user's attention: the program may work, but not in exactly the way
/// the user wanted. A gui would probably use a messagebox here.

"Generally" (for fatal_error()) and "would probably" (for warning())
and exception will be thrown.

Some existing code snippets from lmi:

configurable_settings.cpp
       fatal_error() << "Instantiation failed." << LMI_FLUSH;
       throw std::logic_error("Unreachable"); // Silence compiler warning.
global_settings.cpp
       fatal_error() << "Instantiation failed." << LMI_FLUSH;
       throw std::logic_error("Unreachable"); // Silence compiler warning.
ihs_avmly.cpp
       fatal_error()
           << "Change to increasing death benefit option"
           << " not allowed on this policy form."
           << LMI_FLUSH
           ;
       return;
ihs_avstrtgy.cpp
           fatal_error()
               << "Case '"
               << a_StrategyVector[Year]
               << "' not found."
               << LMI_FLUSH
               ;
           return 0; // Bogus return--actually unreachable.

Would it be correct to say that it is advised to throw a logical error
after a fatal_error(), and a warning() should be followed by a return
statement? In both cases the reason to add something is that in the
future there could be an implementation which does not throw. The
difference is due to fatal_error() being much less likely to ever let
program flow to return from it, right?