RFR: JDK-8312201: Clean up common behavior in "page writers" and "member writers"

[Pavel Rappo]

On Thu, 27 Jul 2023 01:44:36 GMT, Jonathan Gibbons <jjg at openjdk.org> wrote:

> Please review a medium change that is a follow-up to recent work to clean up the builders and writers in the doclet.
> 
> This change is primarily focussed on cleaning up the code to generate summary tables for each kind of member. The bulk of the code is moved from `ClassWriter` to `AbstractMemberWriter`, with just a top-level loop remaining in `ClassWriter` to iterate over the different kinds of members. In addition, there is some cleanup in `WriterFactory`, renaming the factory methods and ensuring that those that are declared are used.
> 
> One result of the cleanup is that each member writer now defines and provides two methods, `buildSummary` and `buildDetails` that each create content to go on the page of the enclosing type.
> 
> This work uncovered two issues which are _not_ addressed here, and which may be addressed in future work.
> 
> 1. There is a discrepancy between the lists of `VMT.Kind` used in `AbstractMemberWriter` (for the complete lists used to generate pages and the lists of `VMT.Kind` defined in `VMT.Kind` itself, which are indirectly used to generate entries in the sub-navbar for the pages of type elements.  Those lists (defined as `Set`s, indirectly rely on the order in which the `VMT.Kind` members are declared, although the order itself is not specified.
> 
> 2. There is an inconsistency in the subtypes of `HtmlDocletWriter`, which all generate pages, as to how those classes are invoked. It is about 50/50 whether the classes declare a `build()` method to be called from outside, or whether the classes declare a `static void generate()` method to be used instead.
> 
> --
> 
> The initial commit contains two changes to tests, copied from PR #15046.  These changes will go away once that PR is approved, integrated and merged with this work.

> 8284447

You added 8284447: Remove the unused NestedClassWriter interface. Is it this PR where this ~interface~ class becomes used or https://github.com/openjdk/jdk/pull/14913, which merges the class with the interface while also moving it?

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

PR Comment: https://git.openjdk.org/jdk/pull/15049#issuecomment-1655719330