[PATCH] improve javadoc in TreeSet#add api documentation

[Stuart Marks]

On 11/1/18 8:12 PM, Kishor Gollapalliwar wrote:
> Thank you for providing this opportunity. I'm up for this challenge and I'd
> love to take this task. The only huddle would be, how to proceed, ie planning
> things step by step, as this is my first time. If you can help me with the
> planning, I'll do the rest and fix these issues.

Hi, sorry for the delay. I had a couple conferences and then some vacation.

This issue comes up periodically. Consider this Stack Overflow question and this
answer in particular:

     https://stackoverflow.com/a/53375284/1441122

It quotes the relevant part of the TreeSet spec (which is also in the SortedSet
interface):

> Note that the ordering maintained by a sorted set (whether or not an explicit
> comparator is provided) must be consistent with equals if the sorted set is
> to correctly implement the Set interface. (See the Comparable interface or
> Comparator interface for a precise definition of consistent with equals.)
> This is so because the Set interface is defined in terms of the equals
> operation, but a sorted set performs all element comparisons using its
> compareTo (or compare) method, so two elements that are deemed equal by this
> method are, from the standpoint of the sorted set, equal. The behavior of a
> sorted set is well-defined even if its ordering is inconsistent with equals;
> it just fails to obey the general contract of the Set interface.

What's missing here are 1) a clear statement that membership in a SortedSet is 
determined by the comparison method, and 2) a crisp definition of "comparison 
method".

Various places in the spec mention something like "the Comparator's compare() 
method, if there is a comparator method, otherwise the compareTo() method if the 
set is using the elements' natural order...." The term "comparison method" 
should be defined early on so that it can be used in later parts of the spec, 
avoiding the comparator/natural-order awkwardness.

Then, it should be specified that the comparison method 1) determines membership 
in the set as well as 2) ordering of set iteration, subsetting, etc. This 
overlaps some with the first two paragraphs of the SortedSet class spec.

Regarding membership, the Set interface says:

> sets contain no pair of elements e1 and e2 such that e1.equals(e2)

SortedSet should have something similar, e.g.,

> SortedSets contain no pair of elements e1 and e2 such that e1 CMP e2 == 0

where "CMP" is the "comparison method". (You don't have to use this notation, 
but the idea is that there is no pair of elements in the set for which the 
comparison method returns zero.)

Only at this point should the "consistent with equals" discussion be introduced. 
The problem with the existing text is that it introduces the "consistent with 
equals" topic, and then only incidentally mentions that set membership is 
determined by "equality" according to the comparison method instead of equals().

That's the first step, which basically amounts to rewriting the first three 
paragraphs of the SortedSet class specification. The subsequent steps are:

2) Reconcile class doc of SortedSet subtypes, e.g. NavigableSet, TreeSet, 
possibly ConcurrentSkipListSet.

3) Audit all method specs of all classes and reconcile them with the class 
specs. A starting point for methods to look at is in this bug:

https://bugs.openjdk.java.net/browse/JDK-8190545

4) 5) 6) Similar steps as above for the SortedMap family.

**

I'd suggest just starting off with the first step instead of trying to do it all 
at once. Don't worry about posting webrevs or specdiffs yet; just send a patch 
or even plain text of the draft to the list and I'll start reviewing it.

Thanks,

s'marks