Re: [lmi] Strategy for Halloween release

[Vadim Zeitlin]

On Sat, 10 Oct 2015 19:57:34 +0000 Greg Chicares <address@hidden> wrote:

GC> On 2015-10-09 18:12, Greg Chicares wrote:
GC> [...]
GC> > assert consistent run order across all cells:
GC> >   https://github.com/vadz/lmi/pull/1/files?w=1
GC> >   (useful starting point for ensuring input consistency for group quotes)
GC> 
GC> Committed 20151010T1930Z, revision 6346.

 Thank you, I've updated https://github.com/vadz/lmi/pull/3 and will base
the is_okay_to_run_group_quote when it's finally ready (I don't promise it
for "today" any longer) on this code too.

GC> Correct me if I'm wrong, but I thought parentheses had become unnecessary
GC> in code like
GC>   std::cout << "The number is " << (n + 1) << std::endl;

 Yes, they're unnecessary but IMHO make things more clear. Of course, I'm
also one of those strange people who find "n + 1" more clear than "1 + n"...

GC> Other than that, our implementations don't differ materially. You had
GC> hoisted cells.size() out of the loop, and I had thought of hoisting
GC> case_default["RunOrder"], but I guess even a decade-old compiler must do
GC> that sort of optimization routinely, based on your measurements here:
GC>   https://github.com/vadz/lmi/pull/1/
GC> (was that gcc-3.4.5, or some other compiler?).

 It was 3.4.5 (I also did some benchmarks with MSVC just to confirm that
there was nothing unexpected there -- and there wasn't).

GC> I created a trivial census with three cells, changing the run order in the
GC> third cell. Formerly, this would have "worked" silently. Now, running it in
GC> the GUI triggers an error message. So does running it at the command line:
GC> 
GC> /opt/lmi/bin[0]$./lmi_cli_shared --accept --ash_nazg 
--data_path=/opt/lmi/data --file=/opt/lmi/eraseme/run_order.cns
GC> Case-default run order 'Life by life' differs from run order 'Month by 
month' of cell number 3. Make this consistent before running illustrations.
GC> [file /lmi/src/lmi/illustrator.cpp, line 261]

 Yes, I should have probably mentioned that I tested this as well. I'm not
sure if I should have added (nor where/how) a unit test verifying this...

VZ