RFR: 8341566: Adding factory for non-synchronized CharSequence Reader

[Chen Liang]

On Sat, 5 Oct 2024 16:32:39 GMT, Markus KARG <duke at openjdk.org> wrote:

> This Pull Requests proposes an implementation for [JDK-8341566](https://bugs.openjdk.org/browse/JDK-8341566): Adding the new method `public static Reader Reader.of(CharSequence)` will return an anonymous, non-synchronized implementation of a `Reader` for each kind of `CharSequence` implementation. It is optimized for `String`, `StringBuilder`, `StringBuffer` and `CharBuffer`.
> 
> In addition, this Pull Request proposes to replace the implementation of `StringReader` to become a simple synchronized wrapper around `Reader.of(CharSequence)` for the case of `String` sources. To ensure correctness, this PR...
> * ...simply moved the **original code** of `StringBuilder` to become the de-facto implementation of `Reader.of()`, then stripped synchronized from it on the left hand, but kept just a synchronized wrapper on the right hand. Then added a `switch` for optimizations within the original code, at the exact location where previously just an optimization for `String` lived in.
> * ...added tests for all methods (`Of.java`), and applied that test upon the modified `StringBuilder`.
> 
> Wherever new JavaDocs were added, existing phrases from other code locations have been copied and adapted, to best match the same wording.

src/java.base/share/classes/java/io/Reader.java line 152:

> 150:      * reader was being used by a single thread (as is generally the case).
> 151:      * Where possible, it is recommended that this class be used in preference
> 152:      * to {@code StringReader} as it will be faster under most implementations.

I believe this paragraph is redundant, now that `StringReader` is a somewhat-deprecated API that we aim to migrate away from. We can note these differences in the API notes of `StringReader`.

src/java.base/share/classes/java/io/Reader.java line 161:

> 159:      * {@code read(char[])}, {@code read(char[], int, int)},
> 160:      * {@code read(CharBuffer)}, {@code ready()}, {@code skip(long)}, and
> 161:      * {@code transferTo()} methods all throw {@code IOException}.

I believe these 2 paragraphs are all redundant besides the initial `The returned stream is initially open.` sentence: those are duplicating information already available from `Reader::close` specification.

Closing `StringReader` is optional but its specification doesn't mention this. I think we don't need to mention if closing is optional, so that one sentence is sufficient.

src/java.base/share/classes/java/io/Reader.java line 163:

> 161:      * {@code transferTo()} methods all throw {@code IOException}.
> 162:      *
> 163:      * <p> The {@code markSupported()} method returns {@code true}.

Suggestion:

     * <p>The returned reader supports the {@link #mark mark()} operation.

src/java.base/share/classes/java/io/Reader.java line 174:

> 172:      * @since 24
> 173:      */
> 174:     public static Reader of(CharSequence c) {

Should we give this factory a more specific name so we don't clash in the future? For example, if we add another factory `of(A a)` for interface `A`, then it would be confusing to have an instance of `interface C extends CharSequence, A` to be passed to `of`.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/21371#discussion_r1789107588
PR Review Comment: https://git.openjdk.org/jdk/pull/21371#discussion_r1789111313
PR Review Comment: https://git.openjdk.org/jdk/pull/21371#discussion_r1789108008
PR Review Comment: https://git.openjdk.org/jdk/pull/21371#discussion_r1789104510