Jeff Garland wrote:
Jason Dolan wrote:
Jason Dolan wrote:
I'm looking to take a string and convert it to a date. The only problem is the string can be one of many patterns. i.e. ("%Y%m%d", "%Y-%m-%d", "%d/%m/%Y", etc...). It is also possible that the given string will fail all pattern matches, and thus return false. And there's nothing to indicate which pattern it might be? Nope. I'm basically allowing the user to input a date and time *almost* anyway they want. What I want to do is test that string against my list of patterns(i.e. known ways to write a date and time) to try and parse
Jeff Garland wrote: the date.
Well, again some of them are going to be ambiguous. And I'll just say, this is a big (read not doable) undertaking. Are all these valid? They certainly are to someone somewhere:
june 5 2006 june 5, 2006 june 5, 06 jun 5 //just assume the current year junio 5, 2006 //spanish 5-jun-2006 5-june-2006 05-June-2006 05 JUNE 2006 7/5/6 7/6/5 5/6/7 07/05/06 05-07-06 05.07.06 050706 Yikes! Maybe I'll be a little more restrictive... like force 4 digit years, etc...
...snip...
You could refactor your code so you don't reallocate the stringstream and facet (just reset the strings and formats). But if you really need a high level of efficiency, then you're going to just have to bite the bullet and write a custom parser. The iostreams solution will always be inherently less efficient than custom solutions.
Thats why I was looking into it this way, I was hoping not to have to create my own parser. I'm actually kinda of surprised that there isn't a open source natural language date string parser out there already.
Jeff