RFR: 8197594 - String and character repeat

[Jim Laskey]

Thank you Roger.

> On Feb 15, 2018, at 4:38 PM, Roger Riggs <Roger.Riggs at oracle.com> wrote:
> 
> Hi Jim,
> 
> Its cleaner to do the API (CSR) review before and without the implementation.
> It helps keep the issues separate.

This was on advice from a member of the core libs team. He can speak up if he wants.

> 
> Don't make statements about count == Integer.MAX_VALUE / 2.
> There is no point, unless it should throw IAE.
> 
> On 2/15/2018 3:16 PM, Jim Laskey wrote:
>>> On Feb 15, 2018, at 4:04 PM, Remi Forax <forax at univ-mlv.fr> wrote:
>>> 
>>> I'm not sure we need 4, it's just a convenient method that may be slower than if the user code calls toString() (because of profile pollution),
>>> so i'm not sure i pull it's own weight.
>>> 
>>> And about adding a default method into CharSequence, the default method should return a CharSequence or String ?
>> If you mean each class returns an instance of its class, I think that overlaps Appendable..
> Beware thinking about default methods on interfaces; the dragons will get you.
> Recently, this was discussed in the context of Reader.transferTo.
>>> and what about the other implementations, AbstractStringBuilder and CharBuffer at least ?
>> This falls into the Appendable.append( T t, int count) realm mentioned originally.
>> 
>> Long term this could be a goal, and maybe defaulting CharSequence#repeat returning a string would be shortsighted.
>> 
>> But, I think having instance String#repeat returning a CharSequence would limit its use (methods expecting strings.)
> There is no point in returning CharSequence, a String is a CharSequence and can be used anywhere
> a CharSequence can.
> 
> $.02, Roger
> 
>> 
>> — Jim
>> 
>>> Rémi
>>> 
>>> ----- Mail original -----
>>>> De: "Jim Laskey" <james.laskey at oracle.com>
>>>> À: "Brian Goetz" <brian.goetz at oracle.com>
>>>> Cc: "core-libs-dev" <core-libs-dev at openjdk.java.net>
>>>> Envoyé: Jeudi 15 Février 2018 18:34:19
>>>> Objet: Re: RFR: 8197594 - String and character repeat
>>>> Very reasonable approach.
>>>> 
>>>> 
>>>>> On Feb 15, 2018, at 1:31 PM, Brian Goetz <brian.goetz at oracle.com> wrote:
>>>>> 
>>>>> I suggest merging 1 and 4 by making it an instance method on CS with a default
>>>>> in CS and an override on string and string builder?
>>>>> 
>>>>> Sent from my MacBook Wheel
>>>>> 
>>>>>> On Feb 15, 2018, at 9:20 AM, Jim Laskey <james.laskey at oracle.com> wrote:
>>>>>> 
>>>>>> This is a pre-CSR code review [1] for String repeat methods (Enhancement).
>>>>>> 
>>>>>> The proposal is to introduce four new methods;
>>>>>> 
>>>>>> 1. public String repeat(final int count)
>>>>>> 2. public static String repeat(final char ch, final int count)
>>>>>> 3. public static String repeat(final int codepoint, final int count)
>>>>>> 4. public static String repeat(final CharSequence seq, final int count)
>>>>>> 
>>>>>> For the sake of transparency, only 1 is necessary, 2-4 are convenience methods.
>>>>>> In the case of 2, “*”.repeat(10) performs as well as String.repeat(‘*’, 10).
>>>>>> 3 and 4 convert to String before calling 1.
>>>>>> 
>>>>>> Performance runs with jmh (results as comment in [2]) show that these
>>>>>> methods are significantly faster that StringBuilder equivalents.
>>>>>> - fewer memory allocations
>>>>>> - fewer char to byte array conversions
>>>>>> - faster pyramid replication vs O(N) copying
>>>>>> 
>>>>>> I left StringBuilder out of scope. It falls under the category of
>>>>>> Appendables#append with repeat. A much bigger project.
>>>>>> 
>>>>>> All comments welcome. Especially around the need for convenience
>>>>>> methods, the JavaDoc content and expanding the tests.
>>>>>> 
>>>>>> — Jim
>>>>>> 
>>>>>> [1] webrev: http://cr.openjdk.java.net//oj/home/jlaskey/8197594/webrev-00
>>>>>> [2] jbs: https://bugs.openjdk.java.net/browse/JDK-8197594
>>>>>> 
>