
From: boost-users-bounces@lists.boost.org [mailto:boost-users-bounces@lists.boost.org] On Behalf Of Murilo Adriano Vasconcelos Sent: Friday, August 13, 2010 1:04 AM To: boost-users@lists.boost.org Subject: Re: [Boost-users] About naming of integer log in base 2 Hi Murilo
and maybe you need to document some examples of using policies that don't bother to check or at least do so quietly
(your > original idea), just quietly return -1, produce a message and return -1, or message and then throw an exception. I should have said that there is an example at \boost_1_43_0\libs\math\example\error_policy_example.cpp and you will also want to provide some Boost.Test style tests too - something like BOOST_CHECK_EQUAL(ilog2(1), 1); // Some plain values... ... BOOST_CHECK_EQUAL(ilog2(FFFFFFFF), 1); // Some cases that might cause trouble. BOOST_CHECK_EQUAL(ilog2(std::numeric_limits<unsigned_int>::max()), 1); // Other cases that might cause trouble. Cases that really will cause trouble... in this case zero. BOOST_CHECK_EQUAL(ilog2(0), -1); // default policy is to just return -1 (or whatever you decide for default). BOOST_CHECK_THROW(ilog2<policy<bad_argument_error<throw_on_error> > > (0)); // policy that throws an exception for zero. In the math library, these tests are obscured by using macros to repeat them for many functions and policies, but you will find many examples to help you. Setting this up is really mind-numbingly boring but is very well worth it to provide quality assurance for the users - Boost quality. It will make passing a review much more likely. I have added some tests: http://svn.boost.org/svn/boost/sandbox/SOC/2010/bits_and_ints/libs/integer/t... A good start (though I am not clear what using the macro ILOG2_TEST really buys you. It will be less familiar to Boost users. It isn't as if it saves you much typing? I haven't looked at the coverage of difficult cases but it's on the right track. Other tests can be added as you (or worse, other people) find bugs (features) ;-) You might like to indicate what feature is being tested as a comment - unless it's obvious. And examples: http://svn.boost.org/svn/boost/sandbox/SOC/2010/bits_and_ints/libs/integer/e... I'd avoid using BOOST_TEST in examples. The policies example isn't quite the template that users can copy and paste to get them started quickly. (And policies appear to be a bit scary - until you realise that most of time you can just ignore them - until you need a better error message etc. So the example would better show at least two cases - err silently and return -1 or whatever, and err with some informative output. ). (It took a real whizz like John Maddock to devise and apply policies. My contribution was to complain that you couldn't do some things that reasonable users would need to!). I also think it is helpful to include the output from the example as a comment. I think it often fits nicely on the end of the same line as the cout << .... (Saves moving ones eye down to notes at the end (or worse running the example to find out!). Of course if the output is long you will need another line, or it will fit better at the end. That can be found in the docs: http://svn.boost.org/svn/boost/sandbox/SOC/2010/bits_and_ints/libs/integer/d... st_integer.bits_and_ints.integer_logarithm_base_2__ilog2_function__ Looking very smart J Have fun! I am having, thank you :) J Keep up the good work! (I hope some people will start to use this collection 'in anger' soon?) PS I noted typos for 'wich' in more than one place. and legth in Not found a string of consecutive ones with legth at least 20 and lots of "it's" that should be "its" Only use "it's" if short for "it is" !! --- Paul A. Bristow Prizet Farmhouse Kendal, UK LA8 8AB +44 1539 561830, mobile +44 7714330204 pbristow@hetp.u-net.com