Using a ParseKit grammar to parse timestamps

Question

I've a fairly simple question about ParseKit and parsing timestamps... how do I go about forcing the symbolic-nature of a dot/period.

For example, if I am trying to parse 2008-01-25, I could use something like date = /\d{4}/ '-' /\d{2}/ '-' /\d{2}/. In fact, there is a date.grammar shipped with ParseKit that does exactly this (interestingly enough, though, the provided grammar doesn't work in the DemoApp unless you add @symbolState='-';, but I digress...)

However, what do I do if I want to parse a date with dots in it... for example, 2008.01.25 or 2008-01-25-12.34.45. I've tried added '.' to the @symbolState directive but it just keeps getting ignored. Note that I am relying on the DemoApp to test my grammars at the moment... not sure if that makes any difference.

Any thoughts would be much appreciated.

Answer 1

Developer of ParseKit here.

First, thanks for the heads up on the bug in the date.grammar file. I have fixed it.

---

As for your main question, I'm pretty sure what you are trying was not possible with ParseKit until now.

That is, ParseKit's tokenizer (PKTokenizer) was not able to produce only whole number Number tokens. Numbers were always tokenized as floating point which means it was impossible to parse input like 3.14 as three separate tokens 3 (Number) . (Symbol) 14 (Number). Rather it would always be tokenized as 3.14.

Good news: I've added this capability with a new method:

-[PKNumberState allowsFloatingPoint]

which defaults to YES.

And I added a matching Tokenizer Directive which you can use in your ParseKit Grammars like:

@allowsFloatingPoint = NO;

NOTE you'll need to checkout the latest HEAD of trunk on Google Code to see this feature.

So, here's an example date grammar which does roughly what you were asking for with the new feature:

@symbolState = '.';
@allowsFloatingPoint = NO;

@start  = date;

date    = year dot month dot day;

year    = /\d{4}/;
month   = /\d{2}/;
day     = /\d{2}/;

dot    = '.';

---

Answer 2

I'm wondering if a simpler idea might be to get parseKit to simply parse the date as a string, and then hand it off to the NSDate::dateWithNaturalLanguageString:locale: orNSDate::dateWithNaturalLanguageString: for processing.