8058779: Faster implementation of String.replace(CharSequence, CharSequence)

[Ivan Gerasimov]

On 02.06.2015 12:15, Paul Sandoz wrote:
> On Jun 2, 2015, at 10:56 AM, Ivan Gerasimov <ivan.gerasimov at oracle.com> wrote:
>>> You could simplify the data provider sourceTargetReplacementWithNull, there is no point testing with a null source. String.replace accepts two arguments, each can be either null or non-null. Thats 2 bits of state, so one can simply be explicit and presumably it should not matter what the non-null argument value (there are enough non-null values supplied by the other data providers):
>>>
>>>    {null, null}
>>>    {null, "foo"}
>>>    {"foo", null}
>> I agree there is not much value in testing the case source == null, so it can be removed.
>> However, I would leave other test cases.
>> Earlier in this thread Rémi spotted a bug, which had been observed in a call "bar".replace("foo", null), but not in "foo".replace("foo", null).
>>
> Ok, presumably for the same identity rather than value?

For any valid substring, so "foobar".replace("bar", null) wouldn't have 
thrown OOM, even though it should have.
Surely, the current testcases often duplicate each other, but it seems 
to be easier to process them all then to filter the duplicates out.

Sincerely yours,
Ivan

> Paul.
>
>
>
>