Re: [lmi] New rate-table implementation

[Vadim Zeitlin]

On Thu, 19 May 2016 00:40:24 +0000 Greg Chicares <address@hidden> wrote:

GC> On 2016-03-19 19:43, Greg Chicares wrote:
GC> [...]
GC> > Therefore, 'actuarial_table.?pp' should be replaced in its entirety after
GC> > thorough testing.
GC> 
GC> Replacement candidate:
GC>   https://github.com/vadz/lmi/pull/23/
GC> committed 20160519T0031Z, revision 6603.

 Thank you for committing this, but I'm not sure if we're attaching the
same meaning to the word "replacement": I was speaking about replacing the
existing code in actuarial_table.cpp with the code in what is now known as
rate_table.cpp. So this replacement hasn't been even started yet, let alone
achieved candidate status. The question is: should I start it now?

GC> I made some superficial changes before committing:
GC>  - smite files with BDFL rod of renaming +/-7 (sign depends on your POV)
                                                                        ^^^
 Didn't you mean "alignment"?

 Anyhow, to continue my preaching to the converted, this renaming didn't
bother me at all because of Git content-addressable nature: it knows that
mort-tables:soa_database.cpp master:rate_table.cpp are the same files and
so can show me the (small) difference the two, for example. With svn it
would have been much more fun...

GC>  - use regexen instead of substrings in tests of thrown exceptions
GC>      (I strove to escape periods ('.'), but didn't exercise any real
GC>      cleverness in turning these into regular expressions)

 BTW, you may want to use C++11 raw string for regexes, not needing to
escape backslashes in regexes is really helpful.

GC>  - reformat comments explaining header inclusion (thanks for following
GC>      a style consistent with existing code...but that was Beman Dawes's
GC>      code, which I recently reformatted, so I did the same here)

 Sorry about this, I'm still not sure what the rule here is. But all in all
there are very few changes, so I'm quite happy about it ("it" being either
the success of my attempts to conform to lmi coding style or you deciding
it was not worth to deal with my failure to do it...).


GC> I clumsily revised one use of the stdout redirector because it seemed
GC> to gobble a useful error message.

 Sorry, I don't see it, which one? I'd really, really love the tests to be
less verbose and ideally output nothing at all in case of success. This
rejoins the other thread specifically about the tests, but I do believe
it's very important for the test suite to be easy to run. And for me ideal
is to do "make check" and only get any output if any tests failed.

 Anyhow, thanks again for applying this and please let me know if I should
mutilate soa_actuarial_table in actuarial_table.cpp now.

 Thanks in advance,
VZ