Topic/performance and expressiveness

[wilhelmtell]

A number of slight and not-so-slight fine-tunings. Some are performance:

• use '\n' instead of std::endl. If we don't want to force a flush it's better not to, for performance.
• std::move() in initialization lists. In the case of rvalues passed at the call-site this leaves the compiler an opportunity to sidestep a copy.

Some are expressiveness-related:

• missing includes. I think it's better to be explicit about the includes: if we use something, include where it's at.
• s/size() > 0/! empty()/

And one commit for idiomatic code: if a member of a copiable object is const then the object is not assignable. I think that usually (barring ... cases I can't think of off the top of my head) copying and assigning go together: when one is allowed so should the other be.

[martinmoene]

Hi Matan,

Thanks for your suggested changes with their explanations. I agree with several of your observations, but not all.

• why allow copying but disable assigning? -- disagree: these are (intended as) immutable values, hence copyable, non-assignable.
• include cstddef for std::size_t …-- heartily agree.
• s/std::endl/\n/. why flush? -- I do prefer a test's output to be flushed to make sure it becomes visible before the next test begins. Update 5-nov-2013: If there are many errors to report, much more time is needed to fix them anyway.
• comment::operator bool expressiveness: prefer size() > 0 over ! empty(): agree.
• rm unnecessary copies in member initialization via std::move -- I agree with the observation that unnecessary copies are made, and realize that your suggestion to use std::move is correct. However one of the aims of lest is to be tiny and easy to read. I'm not sure I like the increase of wordiness in this respect; I might prefer to pass the arguments by const-reference to eliminate a copy.

Thanks to @jwakely for his help.

cheers, Martin