yeptris 0.4.0.1-aarch64-linux → 0.5.1.1-aarch64-linux

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b0564531041910fb120994e2f4cac8c9f3ddf0b176bf5ae9ca3289cb58f6ac87
-  data.tar.gz: 4b6da94d1b0706fdfdd597bc173ca9556e26482a517b391670ef2bb1471f17df
+  metadata.gz: c4a44a9db7ff23a982534d2f3f18bc5e829c3e824acd80c774f5b0c73674ffc3
+  data.tar.gz: '00944fc916a5ddb777810ddcb57113038778443d2b975cb2689b4d92bbb64e13'
 SHA512:
-  metadata.gz: 47c6f2bfe6920039b2e2e0a69f84688cac61bd11e61df76d1f9ec39ccfad227ff5a831b0b7bdbc644829c8dd364bd4b112b7b54183758dee182f66f325371350
-  data.tar.gz: 1e7e5b9bbe7b38b3db23b7c16bf03f3981178f6e52f2db99e091a5560c21c2a98a50823b9598bbeb4d93bb14302c34a3f1e5e2fe97eb0de5c3b8c6ac873cc36f
+  metadata.gz: faa61207b3f1c19a78e5fe61cab47acef7d59c160298520f963ccbfb426187fd3a6b7a2cb7312d4fc5af51f8724e06491a3e76d78f1140b44b99dc0733a4a7e0
+  data.tar.gz: da1187d0c569cb4193a8b0edd5b33356dcdec32b4772f77669a1513824412a17f6a95f40fcca49470ea74ab89ee8cf2c5b5f904b3619a49905872ad83e57c65c

data/lib/yeptris/document.rb

@@ -137,15 +137,16 @@ class Yeptris::Document
     Yeptris::FFI.yeptris_document_build(@c_ptr, entries, count, blob, blob_len)
   end
 
-  def serialize(canonical: false, best_width: 0)
+  def serialize(canonical: false, best_width: 0, explicit_doc_start: false)
     ensure_alive!
     len = ::FFI::MemoryPointer.new(:uint64)
     ptr =
-      if canonical || best_width.positive?
+      if canonical || best_width.positive? || explicit_doc_start
         opts = Yeptris::FFI::EmitOptions.new
         opts[:size] = Yeptris::FFI::EmitOptions.size
         opts[:canonical] = canonical ? 1 : 0
         opts[:best_width] = best_width
+        opts[:explicit_doc_start] = explicit_doc_start ? 1 : 0
         Yeptris::FFI.yeptris_serialize_ex(@c_ptr, opts, len)
       else
         Yeptris::FFI.yeptris_serialize(@c_ptr, len)

data/lib/yeptris/ffi.rb

@@ -47,7 +47,8 @@ module Yeptris
     class EmitOptions < ::FFI::Struct
       layout :size, :uint32,
              :canonical, :int,
-             :best_width, :int
+             :best_width, :int,
+             :explicit_doc_start, :int
     end
 
     attach_function :yeptris_version, [], :string

data/lib/yeptris/psych/visitors.rb

@@ -28,13 +28,13 @@ module Yeptris
         end
 
         def finish
-          @tree.serialize
+          @tree.serialize(explicit_doc_start: true)
         end
 
         def visit(obj)
           case obj
           when nil, true, false, ::Integer, ::Float, ::String then scalar(obj)
-          when ::Symbol then scalar(":#{obj}")
+          when ::Symbol then @tree.new_scalar(":#{obj}", :plain) # psych emits symbols bare, never through visit_String's quoting rules
           when ::Date, ::Time then scalar(obj) # timestamp text, never ivars
           when ::Hash then visit_hash(obj)
           when ::Array then visit_array(obj)

data/lib/yeptris/psych.rb

@@ -138,7 +138,7 @@ module Yeptris
         out =
           case obj
           when nil, true, false, ::String, ::Integer, ::Float, ::Symbol, ::Date, ::Time
-            Yeptris::YAML.dump(obj)
+            Yeptris::YAML.dump(obj, header: true)
           else
             Visitors::YAMLTree.new.push(obj).finish
           end
@@ -366,3 +366,12 @@ end
 # yeptris/psych/drop_in — process-exclusive by nature, since the
 # stdlib cannot be prevented from re-opening whatever ::Psych points
 # at once IT loads.
+# 0.4-contract migration signal (issue #95): this require used to
+# rebind ::Psych. It does not anymore, and re-binding it here would
+# regress #69 (any stdlib psych loaded afterwards would explode with
+# a superclass mismatch), so the old path warns instead of acting.
+if defined?(::Psych) && !::Psych.equal?(Yeptris::Psych)
+  warn "yeptris: \"yeptris/psych\" defines the namespace only — it no " \
+       "longer rebinds ::Psych. Require \"yeptris/psych/drop_in\" for the " \
+       "drop-in rebind, or call Yeptris::Psych explicitly."
+end

data/lib/yeptris/yaml.rb

@@ -1,5 +1,9 @@
 # frozen_string_literal: true
 
+# Self-sufficient (issue #95): the neutral surface is a valid entry
+# point on its own — lutaml-model requires exactly this file.
+require "yeptris"
+
 module Yeptris
   # The neutral Ruby surface (the Psych-compat namespace arrives with
   # the recorder-driven Visitors in phase B; this is the yeptris-native
@@ -76,8 +80,8 @@ module Yeptris
     # (TODO.impl/11 phase 3). Strings are emitted plain only when the
     # resolver round-trips them as strings — everything else takes a
     # quoted style, so dump(load(x)) == x for the scalar types.
-    def dump(obj, canonical: false)
-      BulkBuilder.dump(obj, canonical: canonical)
+      def dump(obj, canonical: false, header: false)
+      BulkBuilder.dump(obj, canonical: canonical, header: header)
     end
 
 
@@ -95,6 +99,7 @@ module Yeptris
       MAP = Yeptris::FFI::BUILD_MAP
       STOP = Yeptris::FFI::BUILD_END
       STYLE_PLAIN = 1
+      STYLE_SQ = 2
       STYLE_DQ = 3
 
       module_function
@@ -107,7 +112,7 @@ module Yeptris
                      Yeptris::FFI::BUILD_SEQ => SEQ_ENTRY,
                      Yeptris::FFI::BUILD_END => END_ENTRY }.freeze
 
-      def dump(obj, canonical: false)
+      def dump(obj, canonical: false, header: false)
         parts = []
         blob = String.new(encoding: Encoding::BINARY)
         off = [0]
@@ -122,7 +127,7 @@ module Yeptris
         bblob = ::FFI::MemoryPointer.from_string(blob)
         rc = doc.build_entries(buf, parts.length, bblob, blob.bytesize)
         raise DumpError, "document_build failed: #{rc}" unless rc == FFI::OK
-        doc.serialize(canonical: canonical)
+        doc.serialize(canonical: canonical, explicit_doc_start: header)
       ensure
         doc&.free
       end
@@ -133,7 +138,14 @@ module Yeptris
           cycle_guard(obj, seen) do
             emit.call(MAP, 0, 0, 0)
             obj.each do |k, v|
-              place(key_text(k), emit, blob, off, seen)
+              # Symbol keys emit as bare ":k" plain — the compat
+              # reader resolves them back to Symbols; string keys
+              # ride the visit_String rules like any other scalar
+              if k.is_a?(Symbol)
+                scalar(":#{k}", STYLE_PLAIN, emit, blob, off)
+              else
+                place(k.to_s, emit, blob, off, seen)
+              end
               place(v, emit, blob, off, seen)
             end
             emit.call(STOP, 0, 0, 0)
@@ -144,12 +156,12 @@ module Yeptris
             obj.each { |e| place(e, emit, blob, off, seen) }
             emit.call(STOP, 0, 0, 0)
           end
-        when String then scalar(obj, plain_string?(obj), emit, blob, off)
-        when Symbol then scalar(":#{obj}", true, emit, blob, off)
-        when Integer, Float then scalar(obj.to_s, true, emit, blob, off)
-        when true, false then scalar(obj.to_s, true, emit, blob, off)
-        when nil then scalar("null", true, emit, blob, off)
-        when Date, Time then scalar(obj.iso8601, true, emit, blob, off)
+        when String then scalar(obj, STYLE_BY_NAME[string_style(obj)], emit, blob, off)
+        when Symbol then scalar(":#{obj}", STYLE_PLAIN, emit, blob, off)
+        when Integer, Float then scalar(obj.to_s, STYLE_PLAIN, emit, blob, off)
+        when true, false then scalar(obj.to_s, STYLE_PLAIN, emit, blob, off)
+        when nil then scalar("null", STYLE_PLAIN, emit, blob, off)
+        when Date, Time then scalar(obj.iso8601, STYLE_PLAIN, emit, blob, off)
         else
           raise DumpError,
                 "cannot dump #{obj.class}: unsupported object " \
@@ -157,12 +169,12 @@ module Yeptris
         end
       end
 
-      def scalar(text, plain, emit, blob, off)
+      def scalar(text, style, emit, blob, off)
         # a BINARY blob absorbs any String bytewise (String#<< on
         # ASCII-8BIT is compatible with every encoding) — the old
         # text.b made a throwaway copy of every scalar before the
         # blob's own copy
-        emit.call(SCALAR, plain ? STYLE_PLAIN : STYLE_DQ, off[0], text.bytesize)
+        emit.call(SCALAR, style, off[0], text.bytesize)
         blob << text
         off[0] += text.bytesize
       end
@@ -200,15 +212,40 @@ module Yeptris
         return false if s.include?(": ") || s.end_with?(":") || s.include?(" #")
         return false if RESHAPES_SET[s]
         # compat's float grammar REQUIRES the dot ("1e3" re-reads as a
-        # String and may dump plain); ints/sexagesimals still reshape
+        # String and may dump plain); ints/sexagesimals still reshape.
+        # psych's FLOAT has no trailing [.:] group — "1.2.3" is a
+        # String and dumps plain (pinned by spec/yaml_spec.rb)
         return false if s.match?(/\A[-+]?(0|[1-9][0-9_]*)(:[0-5]?[0-9])+\z/)
         return false if s.match?(/\A[-+]?(0|[1-9][0-9_]*)\z/)
-        return false if s.match?(/\A[-+]?[0-9][0-9_]*\.[0-9_]*([eE][-+]?[0-9]+)?([.:][0-9_:.]*)?\z/)
+        return false if s.match?(/\A[-+]?[0-9][0-9_]*\.[0-9_]*([eE][-+]?[0-9]+)?\z/)
+        return false if s.match?(/\A[-+]?[0-9][0-9_]*(:[0-5]?[0-9])+\.[0-9_]*\z/)
         return false if s.match?(/\A[-+]?(0x[0-9a-fA-F_]+|0b[01_]+|0o?[0-7_]+)\z/)
         return false if s.match?(/\A[-+]?\.(inf|Inf|INF)\z|\A\.(nan|NaN|NAN)\z/)
         !s.match?(/\A\d{4}-\d\d?-\d\d?([Tt ]|$)/)
       end
 
+      # Psych parity (psych yaml_tree#visit_String) — the quoting
+      # decision for a String, in psych's own order:
+      #   y/Y/n/N (the 1.1 bool words psych double-quotes) → double
+      #   a leading non-word character, no " anywhere          → double
+      #   a 1.1 bad-octal shape (0[0-7]*[89])                 → single
+      #   re-loads as something other than String (plain_string?
+      #   rejects the 1.1 number/bool/null/timestamp shapes)   → single
+      #   needs escapes (control bytes, multiline)             → double
+      #   otherwise                                            → plain
+      # psych's literal/folded/binary forms and its !!str tag for
+      # "<<" are follow-ups; those strings double-quote today.
+      STYLE_BY_NAME = { plain: STYLE_PLAIN, single: STYLE_SQ, double: STYLE_DQ }.freeze
+
+      def string_style(s)
+        return :double if s == "y" || s == "Y" || s == "n" || s == "N"
+        return :double if !s.empty? && !s.include?('"') && s.match?(/\A[^[:word:]]/)
+        return :single if s.match?(/\A0[0-7]*[89]/)
+        return :plain if plain_string?(s)
+        return :double if s.each_byte.any? { |b| b < 0x20 || b == 0x7f }
+        :single
+      end
+
       def cycle_guard(obj, seen)
         id = obj.object_id
         raise DumpError, "cycle detected: cannot dump recursive #{obj.class}" if seen[id]
@@ -257,12 +294,15 @@ module Yeptris
         end
       end
 
-      # A plain scalar that re-resolves to STR stays plain (nice
-      # round-trips); anything ambiguous is double-quoted so the
-      # reparse yields String again.
+      # One quoting table (BulkBuilder.string_style — psych's
+      # visit_String matrix) serves both builders; this side carries
+      # the style onto the per-node DOM.
       def build_string(doc, s)
-        n = new_scalar(doc, s)
-        n.tag_id == :str ? n : new_scalar(doc, s, :force_str)
+        case BulkBuilder.string_style(s)
+        when :single then new_scalar(doc, s, :force_str_sq)
+        when :double then new_scalar(doc, s, :force_str)
+        else new_scalar(doc, s)
+        end
       end
 
       def key_text(k)
@@ -272,7 +312,8 @@ module Yeptris
 
       def new_scalar(doc, text, mode = nil)
         doc.new_scalar(text.to_s,
-                       mode == :force_str ? :double_quoted : :plain)
+                       mode == :force_str ? :double_quoted :
+                       mode == :force_str_sq ? :single_quoted : :plain)
       end
 
       def cycle_guard(obj, seen)

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 = "0.4.0.1".freeze
+  VERSION = "0.5.1.1".freeze
   # The error hierarchy lives in THIS file (the parent namespace's
   # own file): nested constants do not trigger a parent-constant
   # autoload, and the law forbids internal requires — defining the

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: yeptris
 version: !ruby/object:Gem::Version
-  version: 0.4.0.1
+  version: 0.5.1.1
 platform: aarch64-linux
 authors:
 - Ribose Inc.