Fix treap map/list bugs, tighten Treapability rule, expand tests - #29

Raise test coverage and record known collection bugs

getShallowMerger treated a null value as a missing key, so a key present with a null value was merged twice and could be stored twice in the bucket.

Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>

Fix union merge of hash-treap null values

The null-key, hash-key, copy-constructor, and iterator-removal cases were the same assertions with different arguments. Each scenario still runs, through one helper instead of a copied method.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: ericeil <7407587+ericeil@users.noreply.github.com>

The size was copied once in the constructor, so add and clear left it stale while isEmpty read the updated list.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>

A hashCode body that returns a property directly was skipped because the reference has no parent call. Object diagnostics also named a utils.treapHashObject helper that does not exist.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>

set replaced the element in the builder but left the iterator on the previous treap, so previous still returned the old value.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>

fallbackUnion and fallbackIntersect forced the stored value with !!, so a null value threw NullPointerException before the merger ran.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>

parallelMerge, parallelUnion, and parallelIntersect subtracted one before the size check, so a threshold of 0 asked for a negative log and threw.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>

* Seed TreapList priorities so test coverage is reproducible

Node priorities came from Random.Default, so each JVM run built a different
tree and JaCoCo instruction coverage for getNodeAt and computeHashCode moved
between runs. Tests now install a seeded generator, with one fixed tree for
the branches a single random shape can miss. Production still uses
Random.Default when the override is null.

Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>

* Keep TreapList priority seeds on the test thread

The shared Random required tests in one JVM not to overlap, and ProductBugTest
still built lists from Random.Default. Each test method that builds a TreapList
now gets its own seeded generator on the thread that runs it.

Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>

---------

Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Eric Eilebrecht <ericeil@users.noreply.github.com>

…layout (#10)

* Reset TreapListBuilder iterator state after add and remove.

The builder's list iterator kept lastReturnedIndex after add() and
remove(), so a following set() or remove() acted on whichever element
had shifted into that index instead of throwing IllegalStateException
as the ListIterator contract requires.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Exempt explicit and parenthesized TreapableHashCode.plus arguments.

The Treapability hashCode check exempted a property only when its
direct parent resolved to TreapableHashCode.plus. That covered
`it + e` but not `it.plus(e)` (the parent is a value argument) or
`it + (e)` (the parent is a parenthesized expression), so both were
reported. Look through parentheses and from a value argument to its
call before resolving.

Also fix a stale 'utils.HashCode.plus' reference in the rule's comment.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Describe the code, not the change, in recent comments.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Move ProductBugTest cases next to the behavior they cover.

The map cases now live in TreapMapTest, so they run against both the
hashed and sorted implementations (the null-value collision case stays
in HashTreapMapTest). The parallel zero-threshold case uses the same
larger maps as the existing parallel test. The list builder cases move
to TreapListTest.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>