[threeten-dev] [threeten-develop] DateTimeFormatters factory consolidation #210

[Stephen Colebourne]

On 29 January 2013 14:53, Roger Riggs <Roger.Riggs at oracle.com> wrote:
> On 1/28/2013 6:03 PM, Stephen Colebourne wrote:
> It might be worth having an inner class lazy load for some/all of the
> constants. The RFC constant will be slow to create with its maps and
> rarely used. The other constants might be better having their startup
> delayed as well.
>
> I would agree about delaying the initialization if possible but
> with static variables, I don't think the initialization can be lazy.

True. We'll have to test startup and other performance at some point,
but hopefully this won't be a big issue.

> It might be possible for the implementation to instantiate a
> DateTimeFormatter instance
> with only an enum of its format but not initialize the fields.  I'll take a
> look at that after
> other tasks.

It could be done like the week or localized printer/parser using
package-scoped methods. Whether it is worth worrying about is another
matter.

> It also occurs to me that some folks would benefit of documentation of the
> pattern
> used. It might provoke better recognition in addition to the name.  But it
> is not always
> possible.  Thing for a future doc update perhaps.

I took the pattern letter style documentation out as the pattern
letters were not totally clear. They could be added back as guidance
rather than spec.

Stephen