[gravitypdf-bot]

Fixes every open issue: #132, #133, #134, #135, #136, #137, #138, #139, #140, #142, #143, #144, #145, #146.

Where an issue offered a choice between matching jQuery and documenting the current behaviour (#140, #143, #144, #145, #146), this PR matches jQuery, consistent with the behaviour changes already listed under # Unreleased changes. Each one has a Behaviour change CHANGELOG entry.

Index and position methods

• eq(), get() and slice() ignore negative indexes; slice(-2) returns the whole set #142 eq(), get() and slice() count a negative index back from the end (DOM::getNthMatch()).
• slice() takes a length as its second argument, where jQuery takes an end index #144 slice($start, $end) takes an exclusive end index, as jQuery does. This is a breaking change; slice($start, $start + $length) keeps the old result.
• even() and odd() are swapped relative to jQuery and the :even/:odd pseudo-classes #145 even()/odd() count from zero, as jQuery's methods and :even/:odd do. This is a breaking change.
• even() and odd() count text nodes but drop them, so contents()->odd() can be empty #143, first() and last() skip text and other non-element nodes, unlike jQuery and eq(0) #146 even(), odd(), first() and last() count and keep every node. first()/last() are now eq(0)/eq(-1).
• siblings() raises warnings on an element with no parent #139 siblings() skips a match that has no parent, so it no longer raises warnings.

Positional selectors

• filter() and is() match nothing for a positional pseudo-class left of a combinator (ul:first > li) #132 In filter()/is(), a positional compound left of a combinator (ul:first > li) now counts across the document, as it does in find().
• is() counts positional pseudo-classes within the match set, where jQuery's .is() counts across the document #140 is() counts positional selectors across the document, as jQuery's .is() does, via the new NodeMatcher::matchesAnyInDocument(). Selectors without a positional pseudo-class still take a single set pass.
• Positional selectors in prevAll(), parents() and the *Until() methods count in document order, not jQuery's walk order #134 In prevAll(), parents() and nextAll(), a positional selector counts in walk order (prevAll(':first') is the nearest sibling). This adds DOMTraverser::filterList(), which counts along a list in the order given.

CSS engine

• :scope always refers to the document element, so find(':scope > li') matches nothing #133 In find(), findInPlace(), remove($selector) and the qp($node, $selector) constructor, :scope is now the element the search starts from, including inside :not()/:matches(). When the set has several elements and the selector uses :scope, the search runs from each element, as jQuery's does. PseudoClass::is()/isNot() take a new optional $scope argument.
• Support :is() and :where() as aliases of :matches() #136 :is() and :where() are aliases of :matches().
• SimpleSelector::notEmpty() can never return true; deprecate it in favour of isBlank() #138 SimpleSelector::notEmpty() is deprecated; it now returns ! isBlank().

Cloning

• replaceAll() inserts empty copies: it clones the replacement without its children #135 replaceAll() copies each replacement with its children.
• Every inst() deep-copies the match set, so each find() from the root copies the whole document #137 inst() no longer deep-copies the match set it is about to replace. On a 50,000-paragraph document, children() from the root drops from about 27 ms to about 12 ms. A side effect: end() now returns the original nodes instead of detached copies. clone $qp still deep-copies.

Tests

• A regression test file for each issue in tests/Issues/ (Issue132Test.php … Issue146Test.php), covering the issue tables. CI runs them on PHP 7.1–8.5.
• Existing tests updated where they pinned the old behaviour: DOMQueryTest::testSlice, testEven/testOdd, testFirstAndLastWithNonElementNodes, and Issue118Test's positional-order test, which now uses jQuery's walk-order values.

Notes

• The multi-element walk-order results in Positional selectors in prevAll(), parents() and the *Until() methods count in document order, not jQuery's walk order #134 are worked out from jQuery 3.7.1's source. They were not run against jQuery.
• Not changed here, possible follow-ups:
  • siblings(':first') on a multi-element set still counts in document order.
  • filterNodes() counts in document order rather than in the order its input is given.
  • A clone keeps the parent's extension instances, which were built with the parent object. This was already the case before this PR.
• Locally (PHP 8.5): 840 tests pass, and composer run lint and lint:min-php pass.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 99.27007% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 91.66%. Comparing base (6ae6acf) to head (e7ca0d4).
⚠️ Report is 19 commits behind head on main.

Files with missing lines	Patch %	Lines
src/CSS/DOMTraverser/Util.php	95.65%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##               main     #147      +/-   ##
============================================
+ Coverage     90.69%   91.66%   +0.96%     
- Complexity     1424     1469      +45     
============================================
  Files            25       26       +1     
  Lines          3170     3238      +68     
============================================
+ Hits           2875     2968      +93     
+ Misses          295      270      -25     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[gravitypdf-bot]

Split into one PR per issue: #148 (#139), #149 (#142, #144), #150 (#145), #158 (#143, #146; stacked on #150), #151 (#132), #152 (#140), #153 (#134), #159 (#133), #154 (#136), #155 (#138), #156 (#135), #157 (#137).