On Thu, Apr 04, 2013 at 09:30:49AM +0200, Martti K??hne wrote: > On Sun, Mar 31, 2013 at 4:53 AM, Calvin Morrison <mutanttur...@gmail.com> > wrote: > > Forget this whole LINE_MAX thing, I am looking at your suggestion in > > the other message. > > For line buffering, I also use char arrays
I tend to use either a char array (and long line -> error) or a 'double on full' malloc'd char 'vector'. That's somewhat inefficient with the memory but seems to be a good approach when you don't know how big the line or bloc is going to be. It only has to move memory log(n) times, which is not too bad. Moving total cost is O(n). But there can be high latency for appending a single char if it needs to realloc. Can also use some list of buffers ala readv / writev, but then it's not cstr compatible and have to write new functions for simple things like searching in it. Kernighan and Pike implement that 'double on full' approach in 'the Practise of Programming' for one of their examples, so I guess it's legit. This approach could be used for C++ std::vector, requires append to be O(1) on average, not for each append individually. Sam
