DateTimeFormatter parsing issues with patterns "EEE MMM dd HH:mm:ss yyyy" and "EEE MMM dd HH:mm:ss ZZZ yyyy" about joda-time

[jodastephen]

August the 11th 2013 is not a Tuesday, its a Saturday. The code parses August the 11th, and then sets the day-of-week to be Tuesday, which is the 6th.

Not a bug I'm afraid.

[marccarre]

Thanks for your quick reply Stephen.
However, it still looks weird to me: wouldn't you typically expect one of the below results?

• 1. an exception, since the date doesn't exist
• 2. Sun Aug 11 21:41:19 2013 (day-of-month driving day-of-week, since the later one is "smaller"?)

Or, if a "lenient" parsing mode was available:

• 3. Tue Aug 11 21:41:19 2013 even if this is a non-existing date

Thoughts?
Also, would you have a list of all similar conventions that Joda follows?

Cheers,

M.

[jodastephen]

Joda-Time's parsing is essentially simple. It starts with the largest field and sets them one at a time on top of 1970-01-01. This is similar to the JDK, and it means that Joda does not "understand" the fact that it is erroneous. While it could be better, I don't plan to improve it. Essentially, any patch would probably be more about special casing, rather than a general fix.

JSR-310 handles cases like this much better, because it has a better architecture.