# Wednesday, 21 October 2009

I remember, back in the day(*), being so confused about NULL. At first I thought it was a keyword, a special value of some kind. Then I learned it was just the same as 0. Or maybe 0L. I worked with people who said it was wicked to use it, and you should always type the 0 yourself. I worked with people who said it was wicked not to use it, because it made your intentions clear to humans. I even worked with people who used it in for things other than pointer-setting or pointer-checking. I argued with people who thought if(p) was or was not more clear than if (p != NULL) and, late at night when we'd been drinking, about what would happen if some other value than numerically zero meant "null pointer" and what that would do to the world's code, and how to write our code so that could never hurt us.

But that was long ago. These days I've pretty much internalized NULL (and for the record, I'm an if(p) gal) and no longer think about it. One of the reasons I don't have to think about it is that I don't write a template library that's used by millions of developers. If I did, I might have to worry about template-argument-matching and how compilers are supposed to deal with being given an int and knowing to match that up to a Something* -- but then again, only if the int has the value of zero. It's been a pretty clever trick for a really long time, but with the advent of perfect forwarding in C++0x (thanks to rvalue references) it just becomes impossible to always pull off that trick. Luckily, the same language changes that led to rvalue references also held the solution - the literal nullptr that isn't an integer with the value 0, but actually represents a pointer-to-anything that isn't pointing anywhere right now.

For the gory details, a wonderfully transparent look at how the team decides what to do in a release and what to hold for later, as well as some delightful glee at being able to do stuff, some amazing understatement ("somewhat arcane" indeed) and moments later "the world explodes!" you have to watch STL on nullptr. BTW, he deliberately pronounces it null-putter so people can tell he's not saying "null pointer" which would still be NULL. He even gets into nullptr vs nullptr in the land of /clr and how the double underscore (seriously) rides to the rescue. Major whiteboard action and in the end you'll know a little more about why the magical STL (the library) works, why it's faster in VC++ 2010, how the team (including the human STL) thinks and works, and the insides of my favourite language.


* that would be before Microsoft had a C++ compiler, over 20 years ago. Four and a half years ago I reflected that I was very close to having 20 years of C++ experience.

Wednesday, 21 October 2009 09:49:40 (Eastern Daylight Time, UTC-04:00)  #    Comments [0]