algorithms that take ranges by reference

[Andrei Alexandrescu]

Michel Fortin wrote:
> On 2009-12-30 14:53:33 -0500, Andrei Alexandrescu 
> <SeeWebsiteForEmail at erdani.org> said:
> 
>> But they break the tradition because now an algorithm may alter or not 
>> a range, and client code must be aware of that - one more thing to 
>> worry about.
>>
>> What do you think? Should we go with by-ref passing or not? Other ideas?
> 
> I'd say go by the most efficient method. I've implemented a few text 
> parsing functions of my own and they all take the range by reference.
> 
> I think you can make things clear with proper naming. All my functions 
> that advance the range passed by reference are prefixed "consume": 
> "consumeOneChar", "consumeString", "consumeNumber", "consumeUntil", 
> "consumeWhile", etc. This makes the intent very clear.

Perfect. I'll do that.

>> while (!txt.empty) {
>>      auto c = txt.front;
>>      txt.popFront();
>>      if (c == '<') {
>>          if (skip(txt, "!--")) {
>>              // This is a comment
>>              enforce(findSkip(txt, "-->"));
>>              ...
>>           } else if (skip(txt, "script")) {
>>           ...
>>           }
>>      }
>>      ...
>> }
> 
> Are you writing a new XML parser?

No, just some html scraping. I need to parse 200 GB worth of html :o).


Andrei