RFR: 8341566: Adding factory for non-synchronized CharSequence Reader [v4]

[Chen Liang]

On Tue, 8 Oct 2024 12:08:06 GMT, Markus KARG <duke at openjdk.org> wrote:

>> src/java.base/share/classes/java/io/Reader.java line 174:
>> 
>>> 172: 
>>> 173:         return new Reader() {
>>> 174:             private final int length = cs.length();
>> 
>> Hello Markus, as far as I can see, a  `CharSequence` is allowed to have a non-fixed `length()` (typically allowed to increase?). Is there a reason why the length is captured at construction time instead of being evaluated during the read operations of the `Reader`?
>
> As the anonymous class MUST NOT be used with multiple threads, I always have seen the `CharSequence` as *fixed/static* text in the moment the `Reader` is getting used. But indeed, technically one could interleave `Reader::read()` invocations by `CharSequence.append()` (or even worse, `CharSequence.delete()`) invocations. The question is: Would that make *any* sense in the end? I mean, what happens if one has `read()` text that in the next step gets `delete()`'d? I cannot image *any* scenario where such a program would result in *useful* outcome.
> 
> <fun>The fact that nobody so far (before you) brought up this question seems to proof that nobody (besides you) would write such a program. 😄 </fun>
> 
> So I would plea for clearly saying in the JavaDocs that `cs` MUST NOT be modified before `close()` is called. Every other solution implies strange side effects and slower and error-prone implementation of both, anoynous reader *and* test.
> 
> @AlanBateman WDYT?

I would treat this specific scenario as one of the "no concurrent usage" examples. Note that by this principle, mutable objects like `StringBuilder` should not override object comparison methods as these states can change, but they do :(

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/21371#discussion_r1791787167