leptris 1.9.27 → 1.9.29

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 42dec4b3313ea805ec507734e16907a059f894b85057eef2bd1f51833440ba8a
-  data.tar.gz: 7163fb0b7ae9d1e5116f1d637bbb8905dfa94dd7a383c4c358174e82caa70cd0
+  metadata.gz: f91b02217a813c5c7c8eb267b6b82b3eb7689d857fbd76b7a8b108962d56d18c
+  data.tar.gz: c15ad7755d849144c45176d4e28f819ac090927c8a1e1d6ad58c89fe8a3006fe
 SHA512:
-  metadata.gz: c9b449e54b8eecfea8ba76f07b8f7169e37c2ee1aee601e84b836052b05b1811857e7674b29fd93980754b82a53508146c26e3d7054003e82129cb536ec426aa
-  data.tar.gz: b1dba812e5317c7f9c806183277816e06e4847ece3432f231991d3435c55ae03b43debf23d16d3199a71ce79e7ee66031a92fa33528faa5ce32473e37fa84f2e
+  metadata.gz: db0800af7aafe8cf8efb263a0fde99d8876b3d8f487a9cbe738bf2d086a9be8a9e6805e9033b3b695b9064e58d4593f0aee4361dcdc55fa2f920e5a71484acb7
+  data.tar.gz: ae1282d974199e33a98c9361aee2afb36b7dacfb782259380688cd13e480065d3ff59108be44f868b3283f0f11fb958c37d16f28c9eb5fcfc1a6ac6331020c26

data/CHANGELOG.md

@@ -5,6 +5,61 @@ All notable changes to Leptris will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.9.29] - 2026-08-28
+
+### Fixed
+
+- **Element-batch truncation (1.9.28 regression)**:
+  `fetch_element_children`'s opportunistic growth read its sizing
+  from `leptris_element_children(ptr, NULL, 0)` — but that entry
+  point has NO count-only mode (NULL returns 0, unlike
+  `leptris_node_children`), so the truncation check always judged
+  "fits" and wide element families were cut to the scratch
+  capacity (32 on a fresh thread/process). `#element_children`
+  and `#last_element_child` both affected; elements with 32 or
+  fewer element children — and any process whose scratch had
+  already grown — were unaffected, which is why the round-XXIII
+  suite passed. Sizing now comes from the dedicated
+  `leptris_element_child_count`. Two fresh-Thread regression
+  specs pin the fresh-scratch state (the scratch is thread-local:
+  same-process test ordering had masked it); they fail on 1.9.28.
+  The count query fires only when the buffer fills exactly —
+  steady-state cost unchanged, the batch's -27% element_children
+  win holds.
+
+### Meta
+
+- Compiled-expression cache for `Searchable#xpath` prototyped and
+  retired by measurement: mixed deltas across expression shapes,
+  one stable regression (parent-axis `..` +58% through
+  `leptris_xpath_compiled_eval` vs string eval, three interleaved
+  passes), suggesting the engine's string path already amortizes
+  compilation. Reverted in full; the parent-axis asymmetry is
+  worth an upstream look.
+
+## [1.9.28] - 2026-08-28
+
+### Changed
+
+- **Element-child reads pay only for elements** — the
+  interest-proportional principle applied to tree navigation:
+  - `#element_children` on element receivers rides the element-only
+    batch (`leptris_element_children`, opportunistic single
+    dispatch on the shared scratch): text/comment children are
+    never wrapped, and every kept element carries the ELEMENT hint
+    so the per-child `get_type` disappears. Cold pass over
+    big.xml's 25,000 items: 160 -> **117 ms per doc (-27%)**.
+    Non-element receivers (the document node) keep the filter.
+  - `#first_element_child` types candidate siblings with one raw
+    `get_type` each — nothing wrapped or cached until the element
+    is found (it previously full-wrapped leading text nodes).
+  - `#last_element_child` on element receivers wraps only the
+    batch's final pointer (it previously materialized every child,
+    text wraps included, to scan backwards). The first/last pair
+    across fresh docs: **-46%**.
+  - The all-kind `#children` path is untouched — the cold full-tree
+    walk battery stays at parity.
+
 ## [1.9.27] - 2026-08-28
 
 ### Fixed

data/lib/leptris/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Leptris
-  VERSION = "1.9.27"
+  VERSION = "1.9.29"
 end

data/lib/leptris/xml/ffi.rb

@@ -533,7 +533,6 @@ module Leptris
       attach_function :leptris_xpath_compiled_free,
 
         [:leptris_xpath_compiled], :void
-
       attach_function :leptris_sax_parse,
         [:string, :size_t, :pointer, :pointer], :int
       attach_function :leptris_sax_parser_create,
@@ -863,6 +862,35 @@ module Leptris
         buffer.get_array_of_pointer(0, copied)
       end
 
+      # Element-only child batch (the mirror of fetch_children over
+      # leptris_element_children): element receivers skip wrapping —
+      # and paying for — the text/comment children they will filter
+      # out anyway; every wrap carries the ELEMENT hint so the
+      # per-child get_type dispatch disappears too.
+      def self.fetch_element_children(el_ptr)
+        scratch = (Thread.current[:leptris_scratch] ||= {})
+        buffer = scratch[:pointers]
+        if buffer.nil?
+          buffer = scratch[:pointers] =
+            ::FFI::MemoryPointer.new(:pointer, FETCH_INITIAL)
+        end
+        cap = buffer.size / PTR_BYTES
+        copied = leptris_element_children(el_ptr, buffer, cap)
+        if copied == cap
+          # leptris_element_children has NO count-only mode (NULL
+          # returns 0, unlike leptris_node_children) — size the
+          # regrow from the dedicated counter.
+          total = leptris_element_child_count(el_ptr)
+          if total > cap
+            buffer.free
+            buffer = scratch[:pointers] =
+              ::FFI::MemoryPointer.new(:pointer, total)
+            copied = leptris_element_children(el_ptr, buffer, total)
+          end
+        end
+        buffer.get_array_of_pointer(0, copied)
+      end
+
       # XPath result-set batch (leptris_xpath_result_get_nodes_ex):
       # the call fills out_kinds in the 4-value XPATH_NODE space
       # (ELEMENT/ATTRIBUTE/TEXT/OTHER). ELEMENT is the only value

data/lib/leptris/xml/node.rb

@@ -190,12 +190,17 @@ class Leptris::XML::Node
   def first_element_child
     return @first_element_child if memo_hit?(@first_element_child_version)
     ensure_alive!
+    # Raw pointer scan: non-element siblings are typed with one C
+    # call each — never wrapped, never cached — and the found
+    # element carries the ELEMENT hint into the wrap.
     ptr = Leptris::XML::FFI.leptris_node_first_child(@c_ptr)
     result = nil
     until ptr.nil? || ptr.null?
-      node = Leptris::XML::Node.wrap(ptr, @document, parent: as_element_or_self)
-      if node.element?
-        result = node
+      if Leptris::XML::FFI.leptris_node_get_type(ptr) ==
+          Leptris::XML::FFI::NODE_ELEMENT
+        result = Leptris::XML::Node.wrap(
+          ptr, @document, parent: as_element_or_self,
+          node_type: Leptris::XML::FFI::NODE_ELEMENT)
         break
       end
       ptr = Leptris::XML::FFI.leptris_node_next_sibling(ptr)
@@ -208,12 +213,34 @@ class Leptris::XML::Node
   end
 
   def last_element_child
+    # Element receivers: the element-only batch fetches pointers
+    # without wrapping any text child; only the last is wrapped.
+    if is_a?(Leptris::XML::Element)
+      kids = Leptris::XML::FFI.fetch_element_children(@c_ptr)
+      return nil if kids.empty?
+      return Leptris::XML::Node.wrap(
+        kids.last, @document, parent: self,
+        node_type: Leptris::XML::FFI::NODE_ELEMENT)
+    end
     children.reverse_each.find(&:element?)
   end
 
   def element_children
     return @element_children if memo_hit?(@element_children_version)
-    result = children.select(&:element?)
+    ensure_alive!
+    # Element receivers ride the element-only batch: text/comment
+    # children are never wrapped (nor their get_type paid — the
+    # ELEMENT hint rides along). Other nodes keep the filter.
+    result =
+      if is_a?(Leptris::XML::Element)
+        parent = as_element_or_self
+        Leptris::XML::FFI.fetch_element_children(@c_ptr).map do |ptr|
+          Leptris::XML::Node.wrap(ptr, @document, parent: parent,
+                                  node_type: Leptris::XML::FFI::NODE_ELEMENT)
+        end
+      else
+        children.select(&:element?)
+      end
     if @document
       @element_children = result
       @element_children_version = @document.version

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: leptris
 version: !ruby/object:Gem::Version
-  version: 1.9.27
+  version: 1.9.29
 platform: ruby
 authors:
 - Ribose Inc.