yeptris 0.6.51.0 → 0.6.52.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: a140e44f7568dcc17cebcacaacd71aa08762fbd95316514f7095550326ed7b83
-  data.tar.gz: eb554c000a449f1fb573708c0b7f256984aa64cd33666ec4da97cf1905c1fc64
+  metadata.gz: 39841fd37a06d428edeecabd2ec9b89ebbf3b0576e3a113a43b36eca0a0c67d0
+  data.tar.gz: 3fae8c03db5dbb3ddc31f0711a36cf6851b3abb7883e334dc161105070c6f6df
 SHA512:
-  metadata.gz: 57e467a5138b4e7e98fafa311625a45af0187a9635331a5c70494ba4f2ead7630bc12ae29e641eb52d9565f87f9488dea205612404b9d20683a52ff94a7a8807
-  data.tar.gz: 626fb65660cd98c68270e582d692987a9eb270d068d94e2b0a97b34d4b15e3ba3a916e8a6937cfd4a1fbb5100b43278f47972c871f38d99f74193e49246ccd3a
+  metadata.gz: 288cf0885ea87923b8974793617a8420ac6115df05b03615170b6d4abb7d779227412904d324951dcd81009e99a3240b9cc0bb77be2087fd6d4b03d43f0c572c
+  data.tar.gz: 1556db96baf38b002a65a0b79e0dc112de10a5bddd808e66600ada366c33c6fd408590da118f7d2d47d2f605c427bd2ae886b0d0246b2785bea7eaab2882e836

data/lib/yeptris/document.rb

@@ -16,7 +16,7 @@ class Yeptris::Document
   # The schema this document was parsed with (a parse property).
   attr_reader :parse_schema
 
-  def initialize(c_ptr = nil, freed = Freed.new(:alive), parse_schema = :core_12)
+  def initialize(c_ptr = nil, freed = Freed.new(:alive), parse_schema = :core_12, source = nil)
     # The pointer is produced by the FFI parse/create calls; a nil (or
     # non-pointer) argument is an API misuse — Document.parse and
     # Document.create are the entry points. Fail with that guidance
@@ -30,6 +30,13 @@ class Yeptris::Document
     @c_ptr = c_ptr
     @freed = freed
     @parse_schema = parse_schema
+    # the C document BORROWS the input: string values are zero-copy
+    # views over the caller's bytes (parse.h/doc.h) — the FFI call only
+    # pins the String for the call's duration, so the document must
+    # retain it or the next GC frees memory every later node.value/
+    # tag read still points into (audit 2026-10-31: parse(io.read) was
+    # collectable immediately)
+    @source = source
     @readonly = false
     # Strong wrapper cache keyed on the C node address: the same C node
     # always yields the same Ruby object (identity for aliases/eql?),
@@ -61,7 +68,7 @@ class Yeptris::Document
       raise Yeptris::ParseError,
             "parse failed: #{Yeptris::FFI.last_error_message}"
     end
-    wrap_schema(doc, schema)
+    wrap_schema(doc, schema, yaml)
   end
 
   def self.parse_json(json)
@@ -71,7 +78,7 @@ class Yeptris::Document
     raise Yeptris::ParseError,
           "json parse failed: #{Yeptris::FFI.last_error_message}" if doc.null?
 
-    wrap_schema(doc, :core_12) # strict JSON is core by construction
+    wrap_schema(doc, :core_12, json) # strict JSON is core by construction
   end
 
   # An empty document for from-scratch construction (TODO.impl/11 p3).
@@ -99,8 +106,8 @@ class Yeptris::Document
   # host scalar policies (Psych's dot-required float) are conditioned
   # on it. The parse constructors set the real one; plain wrap keeps
   # the core_12 default (yeptris_parse's default).
-  def self.wrap_schema(c_ptr, schema)
-    new(c_ptr, Freed.new(:alive), schema)
+  def self.wrap_schema(c_ptr, schema, source = nil)
+    new(c_ptr, Freed.new(:alive), schema, source)
   end
 
   # #182: the stream-children wrapper owns NOTHING (the parent

data/lib/yeptris/psych/parser.rb

@@ -25,6 +25,10 @@ module Yeptris
         yaml = yaml.to_s
         @filename = filename
         flat, arena = Materializer.drain(yaml, schema: :compat_11)
+        # the engine's arena is UTF-8 bytes; without the tag every
+        # scalar reached handlers as ASCII-8BIT (stdlib Psych hands
+        # UTF-8; TreeBuilder then propagated binary into node values)
+        arena.force_encoding(Encoding::UTF_8)
         dispatch(flat, arena)
         self
       rescue ParseError => e

data/lib/yeptris/psych.rb

@@ -143,8 +143,21 @@ module Yeptris
       # Psych 5: load is safe — plain data structures only. Tagged
       # nodes raise DisallowedClass unless their class is permitted
       # (Date/Time/Symbol are built in; they are plain data here).
-      def load(yaml, permitted_classes: [::Date, ::Time], aliases: false, **)
-        safe_load(yaml, permitted_classes: permitted_classes, aliases: aliases)
+      def load(yaml, permitted_classes: [::Date, ::Time], aliases: false,
+               symbolize_names: false, fallback: nil, filename: nil, **)
+        res = safe_load(yaml, permitted_classes: permitted_classes, aliases: aliases,
+                        symbolize_names: symbolize_names)
+        if res.nil? && !fallback.nil?
+          # stdlib's fallback: answers when the stream is empty or has
+          # NO documents — a document whose content IS nil still loads
+          # as nil. The safe path cannot tell those two apart, so the
+          # corner re-checks the count (only when a fallback rides).
+          doc = ::Yeptris::Document.parse(yaml, schema: :compat_11)
+          empty = doc.nil? || doc.document_count.zero?
+          doc&.free
+          return fallback if empty
+        end
+        res
       end
 
       # Full revival over the Nodes tree: !ruby/object, !ruby/struct,

data/lib/yeptris/valueml.rb

@@ -20,6 +20,25 @@ module Yeptris
   # int64 the C resolver leaves the scalar a string; Psych
   # materializes Integer (issue #31) — every walk rebuilds it.
   PSYCH_INT_SHAPE = /\A[-+]?(?:0|[1-9](?:[0-9]|,[0-9]|_[0-9])*)\z/.freeze
+
+  # The typed-string conversion law, ONE place: a big-int-shaped
+  # string rebuilds to Integer, an implicit-plain ':name' scans to a
+  # Symbol (Psych's ScalarScanner; quoted ':x' stays a String).
+  # Every str-column conversion rides this — the pair fast path
+  # drifted once and handed bignums/symbol-keys through as Strings
+  # (audit 2026-10-31).
+  def self.str_to_value(text, typed)
+    return text unless typed
+
+    if text.length > 18 && PSYCH_INT_SHAPE.match?(text)
+      text.delete(",_").to_i
+    elsif text.length > 1 && text.start_with?(":") && !text.start_with?("::")
+      text[1..].to_sym
+    else
+      text
+    end
+  end
+
     V_TS = 6
     SEQ_OPEN = 7
     MAP_OPEN = 8
@@ -155,15 +174,7 @@ module Yeptris
           text = arena.byteslice(flat[i + OFF], flat[i + LEN])
           # implicit-plain ':name' scans to a Symbol (Psych's
           # ScalarScanner); quoted ':x' stays a String
-          v = if flat[i + B] == 1 && text.length > 18 &&
-                PSYCH_INT_SHAPE.match?(text)
-                text.delete(",_").to_i
-              elsif flat[i + B] == 1 && text.length > 1 &&
-                text.start_with?(":") && !text.start_with?("::")
-                text[1..].to_sym
-              else
-                text
-              end
+          v = str_to_value(text, flat[i + B] == 1)
           if pending_anchor
             anchors[pending_anchor] = v
             pending_anchor = nil
@@ -284,20 +295,9 @@ module Yeptris
            pending_anchor.nil? && !stack.empty? && stack.last.is_a?(Hash)
           kt = arena.byteslice(offs[i], lens[i])
           if kt != "<<"
-            key = if bools[i] == 1 && kt.length > 1 && kt.start_with?(":") &&
-                     !kt.start_with?("::")
-              kt[1..].to_sym
-            else
-              kt
-            end
+            key = str_to_value(kt, bools[i] == 1)
             vt = arena.byteslice(offs[i + 1], lens[i + 1])
-            stack.last[key] =
-              if bools[i + 1] == 1 && vt.length > 1 && vt.start_with?(":") &&
-                 !vt.start_with?("::")
-                vt[1..].to_sym
-              else
-                vt
-              end
+            stack.last[key] = str_to_value(vt, bools[i + 1] == 1)
             i += 2
             next
           end
@@ -305,15 +305,7 @@ module Yeptris
         case kinds[i]
         when V_STR
           text = arena.byteslice(offs[i], lens[i])
-          v = if bools[i] == 1 && text.length > 18 &&
-                PSYCH_INT_SHAPE.match?(text)
-                text.delete(",_").to_i
-              elsif bools[i] == 1 && text.length > 1 &&
-                text.start_with?(":") && !text.start_with?("::")
-                text[1..].to_sym
-              else
-                text
-              end
+          v = str_to_value(text, bools[i] == 1)
           if pending_anchor
             anchors[pending_anchor] = v
             pending_anchor = nil

data/lib/yeptris.rb

@@ -3,7 +3,7 @@
 module Yeptris
   # The gem's version lives in the parent namespace's file — the last
   # internal require (yeptris/version) retired with it.
-  VERSION = "0.6.51.0".freeze
+  VERSION = "0.6.52.0".freeze
 
   # Informational notices are OPT-IN (yeptris-ruby#217): consumers
   # embed yeptris inside CLIs that assert clean stderr, and a

data/vendor/libyeptris/src/yeptris/parse/engine.c

@@ -379,7 +379,8 @@ static int e_comment_at(const yep_engine* e) {
 }
 
 static int e_at_eol(yep_engine* e) {
-    return e->pos >= e->len || e->p[e->pos] == '\n' || e->p[e->pos] == '\r' || e_comment_at(e);
+    return e->pos >= e->len || e->p[e->pos] == '\n' || e->p[e->pos] == '\r' || e_comment_at(e) ||
+           yep_scan_mb_break_len(e->p, e->len, e->pos) > 0;
 }
 
 static int e_skip_to_eol(yep_engine* e) {
@@ -3890,6 +3891,16 @@ static int e_parse_value(yep_engine* e, yep_ctx ctx, uint16_t floor_col) {
         if (e->pos >= e->len) {
             goto empty_value;
         }
+        /* a raw NEL/LS/PS here is this line's break: consume it exactly
+         * as the BLANK branch below consumes an LF-ended empty line —
+         * yep_scan_line's line model does not treat a break byte at the
+         * loop head as a line end, so without this the props-only line
+         * re-entered e_node on the break byte (rows 1-2) and the marker
+         * became the scalar's value (row 5's consume-first form) */
+        if (yep_scan_mb_break_len(e->p, e->len, e->pos) > 0) {
+            e_line_done(e, e->pos);
+            continue;
+        }
         yep_line_info li;
         if (e->pos == e->line_start) {
             /* seed the memo: the fast arms re-read this line's facts —

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: yeptris
 version: !ruby/object:Gem::Version
-  version: 0.6.51.0
+  version: 0.6.52.0
 platform: ruby
 authors:
 - Ribose Inc.
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-10-30 00:00:00.000000000 Z
+date: 2026-10-31 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ffi