[threeten-dev] CaseInsensitive parsing

[Stephen Colebourne]

Is the plan to add these as package-scoped classes?
Does it need the <T>? I assume it will always be a string.

Stephen


On 25 December 2012 03:47, Xueming Shen <xueming.shen at oracle.com> wrote:
> Sure here is the version that supports both case sensitive and insensitive,
> by
>
> RadixTree rtree = RadixTree.newTree();
> RadixTree rtree_ci = RadixTree.newTreeCI();
>
> The 20% slow down happens only in -client mode, does not impact the -server
> vmd.  Who cares -client these days :-)
>
> http://cr.openjdk.java.net/~sherman/jdk8_threeten/radixtree_ci/RadixTree.java
> (http://cr.openjdk.java.net/~sherman/jdk8_threeten/radixtree_ci/Main.java is
> the test)
>
> -Sherman
>
>
> On 12/24/2012 4:21 AM, Stephen Colebourne wrote:
>>
>> I think a single class that can handle cases sensitive and insensitive
>> is probably best. Bytecode size and lack of duplication probably
>> trumps absolute performance.
>>
>> I'm happy for you to integrate this and push if you want.
>>
>> Stephen
>>
>> On 24 December 2012 05:34, Xueming Shen<xueming.shen at oracle.com>  wrote:
>>>
>>> Here is the case insensitive version of the prefix tree.
>>>
>>>
>>> http://cr.openjdk.java.net/~sherman/jdk8_threeten/radixtree_ci/RadixTreeCI.java
>>>
>>> It holds performance relative good in -serve mode, about 2 x the case
>>> sensitive
>>> version (it's 6 x in -client vm)
>>>
>>>
>>> http://cr.openjdk.java.net/~sherman/jdk8_threeten/radixtree_ci/RadixTree.java
>>>
>>> I do have a combined version, but it slows down the case sensitive
>>> version
>>> about 20%.
>>> I'm not sure if we really case about this 20% in parsing.
>>>
>>> -Sherman