yeptris 0.6.3.2 → 0.6.4.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 71f8626c8109cc1d84ee3593a0646a4005b22e00fad770426024fcd783ddf123
-  data.tar.gz: 34293da3ac105daa501ae1e7cc343294035341608696eb9f59bdea789e0cc4a8
+  metadata.gz: d45b2a3218406977adea5d7ae6d9184042864fce7bd2aeb9e69f3d87f0d11351
+  data.tar.gz: f2c61980fb7ca80b0d5686820718594c9cf8a0290975b555b2ce777d136bbfad
 SHA512:
-  metadata.gz: c040550f2f2cfa3b30b1dfca891c306dd2ce390471457575be118098bc16718f48b6031d4c332b18f063556bbaad422135de220e924314291fae6c7438361461
-  data.tar.gz: 8cfc1cf59607134a53ea27dae7a5af0de6801ac251de429863f45ca899566b631638c70e7c4d863e9eb997df562fd3c9a6e2ac7d386bb01193efcbffeee6213d
+  metadata.gz: ef6d04e084a9bf84458396a560176ccd56eacd375bbc8bbf215e6ba810f2e07d149cd95727842fd5864f7eafc90306a09395becc154f16e68e4ec43fcc183322
+  data.tar.gz: da3bc2434c554b1d656e4f2749a1ee98acaaef71d37f5fad4372da679e22ff1c15b5be55545f18cb4e62e3984fc409f8763f563bd1b21d682c765b5f4af8b59b

data/lib/yeptris/ffi.rb

@@ -75,6 +75,33 @@ module Yeptris
     attach_function :yeptris_tape_free, [:pointer], :void
     attach_function :yeptris_tape_convert, %i[pointer size_t size_t pointer pointer], :size_t
 
+    # the compiled plan walk (#293 / TODO.restructure/87): compile a
+    # strict-JSON spec once, apply it to a parsed tape in one C pass,
+    # read the typed COLUMNAR result
+    attach_function :yeptris_plan_compile, %i[pointer size_t pointer], :pointer
+    attach_function :yeptris_plan_free, [:pointer], :void
+    attach_function :yeptris_plan_column_count, [:pointer], :size_t
+    attach_function :yeptris_tape_plan_walk, %i[pointer pointer pointer], :pointer
+    attach_function :yeptris_plan_result_free, [:pointer], :void
+    attach_function :yeptris_plan_result_rows, [:pointer], :size_t
+    attach_function :yeptris_plan_result_kind, %i[pointer size_t], :int
+    attach_function :yeptris_plan_result_ints, %i[pointer size_t], :pointer
+    attach_function :yeptris_plan_result_floats, %i[pointer size_t], :pointer
+    attach_function :yeptris_plan_result_str_offs, %i[pointer size_t], :pointer
+    attach_function :yeptris_plan_result_str_lens, %i[pointer size_t], :pointer
+    attach_function :yeptris_plan_result_nulls, %i[pointer size_t], :pointer
+
+    # the DOM (YAML) leg of the plan walk (#293 slice three): same
+    # compiled plan over a parsed document; str columns expose
+    # (ptr,len) views into the document's regions
+    attach_function :yeptris_document_plan_walk, %i[yeptris_document pointer pointer], :pointer
+    attach_function :yeptris_plan_result_strs, %i[pointer size_t], :pointer
+
+    # yeptris_plan_str (plan.h): one string view
+    class PlanStr < ::FFI::Struct
+      layout :p, :pointer, :len, :size_t
+    end
+
     attach_function :yeptris_document_free, [:yeptris_document], :void
     attach_function :yeptris_document_count, [:yeptris_document], :size_t
     attach_function :yeptris_document_root, [:yeptris_document, :size_t], :yeptris_node

data/lib/yeptris/json/descriptor.rb

@@ -0,0 +1,266 @@
+# frozen_string_literal: true
+
+# The Descriptor plan walk (yeptris#293 slice two, over the C plan
+# ABI of libyeptris v0.6.3): compile a row shape once, then apply it
+# to a JSON document in ONE native pass producing typed COLUMNS —
+# the engine never materializes the intermediate Ruby Hash/Array
+# tree (lutaml-model's hydration reads columns, not trees).
+#
+#     descriptor = Yeptris::JSON::Descriptor.build(
+#       kind: :seq, # or :map with path: — the rows container
+#       children: [
+#         { name: "id", kind: :int },
+#         { name: "name", kind: :str },
+#       ])
+#     result = descriptor.walk(json)
+#     result.column("id") # => [1, 2, ...] (nil where null/missing)
+#     result.to_a        # => [{ "id" => 1, "name" => "x" }, ...]
+#
+# Vocabulary alignment with Leptris::XML::Descriptor (the shape
+# vocabulary lutaml-model shares across engines): :collection at the
+# root is the rows container (the seq form), :scalar is the string
+# leaf. Slice one covers rows of scalar leaves — nested plans, the
+# YAML leg, and partial descriptors ride the board item
+# (TODO.restructure/87).
+class Yeptris::JSON::Descriptor
+  class Error < ::Yeptris::JSON::Error; end
+
+  # spec leaf kinds → the C column kind codes (plan.h)
+  LEAF_KINDS = {
+    int: 0,
+    float: 1,
+    str: 2,
+    bool: 3,
+    scalar: 2, # leptris's string-value row kind
+  }.freeze
+  private_constant :LEAF_KINDS
+
+  # root forms: the rows container's shape (:collection is leptris's
+  # name for a repeated-rows container)
+  ROOT_KINDS = %i[seq map collection].freeze
+  private_constant :ROOT_KINDS
+
+  LEAF_SPEC_NAMES = %w[int float str bool].freeze
+  private_constant :LEAF_SPEC_NAMES
+
+  class Handle < ::FFI::AutoPointer
+    def self.release(ptr)
+      ::Yeptris::FFI.yeptris_plan_free(ptr)
+    end
+  end
+
+  # +kind+: :seq (rows are the root array) or :map (rows live under
+  # +path:+'s value). +path+: a String, or an Array of Strings for a
+  # segmented (nested) path. +children+: the leaf columns; each row
+  # is a mapping, every child names one typed column.
+  def self.build(kind:, path: nil, children:)
+    handle, names = compile_handle(kind, path, children)
+    new(handle, names)
+  end
+
+  # The compile half of .build (shared with the YAML subclass's
+  # build — ONE owner of the spec grammar). Returns the owned handle
+  # and the leaf names.
+  def self.compile_handle(kind, path, children)
+    unless ROOT_KINDS.include?(kind)
+      raise ArgumentError, "kind must be one of #{ROOT_KINDS.inspect}, got #{kind.inspect}"
+    end
+    unless children.is_a?(::Array) && !children.empty?
+      raise ArgumentError, "children must be a non-empty Array"
+    end
+
+    leaves = children.map do |child|
+      name = child[:name]
+      code = LEAF_KINDS[child[:kind]]
+      if code.nil?
+        raise ArgumentError,
+              "leaf kind must be one of #{LEAF_KINDS.keys.inspect}, got #{child[:kind].inspect}"
+      end
+      unless name.is_a?(::String) && !name.empty?
+        raise ArgumentError, "leaf name must be a non-empty String, got #{name.inspect}"
+      end
+
+      { "name" => name, "kind" => LEAF_SPEC_NAMES[code] }
+    end
+
+    case path
+    when nil then nil
+    when ::String then spec_path = path
+    when ::Array
+      unless path.all? { |seg| seg.is_a?(::String) && !seg.empty? }
+        raise ArgumentError, "path segments must be non-empty Strings"
+      end
+      spec_path = path
+    else
+      raise ArgumentError, "path must be a String or an Array of Strings"
+    end
+
+    spec = { "kind" => kind == :map ? "map" : "seq", "children" => leaves }
+    spec["path"] = spec_path unless spec_path.nil?
+
+    # the spec is strict JSON (the C compiler parses it through the
+    # strict JSON DOM); compile once, the engine owns the copy
+    spec_json = ::JSON.generate(spec)
+    st = ::FFI::MemoryPointer.new(:int)
+    raw = ::Yeptris::FFI.yeptris_plan_compile(spec_json, spec_json.bytesize, st)
+    if raw.null?
+      raise Error, "plan compile failed (status=#{st.read_int}): #{spec_json}"
+    end
+
+    [Handle.new(raw), leaves.map { |leaf| leaf["name"] }]
+  end
+  private_class_method :compile_handle
+
+  attr_reader :names # the planned leaf names, in spec order
+
+  # @api private — handles come from .build
+  def initialize(handle, names)
+    @handle = handle
+    @names = names.dup.freeze
+  end
+
+  # Applies the plan to +json+ (String or IO): parse the tape, one C
+  # plan pass, bulk-read the typed columns. The returned PlanResult
+  # is standalone Ruby (nothing borrowed from the document).
+  def walk(json)
+    src = ::Yeptris.read_input(json).to_s
+    tape = ::Yeptris::FFI::JsonTape.new
+    rc = ::Yeptris::FFI.yeptris_parse_json_tape(src, src.bytesize, tape)
+    raise ::Yeptris::JSON::ParseError, ::Yeptris::FFI.last_error_message if rc != ::Yeptris::FFI::OK
+
+    begin
+      st = ::FFI::MemoryPointer.new(:int)
+      raw = ::Yeptris::FFI.yeptris_tape_plan_walk(tape, @handle, st)
+      if raw.null?
+        raise Error,
+              "plan walk failed (status=#{st.read_int}) — the document shape " \
+              "disagrees with the plan"
+      end
+
+      begin
+        PlanResult.read(raw, src, @names)
+      ensure
+        ::Yeptris::FFI.yeptris_plan_result_free(raw)
+      end
+    ensure
+      ::Yeptris::FFI.yeptris_tape_free(tape)
+    end
+  end
+
+  def column_count
+    ::Yeptris::FFI.yeptris_plan_column_count(@handle)
+  end
+
+  # The eager columnar result of Descriptor#walk: typed Ruby columns
+  # keyed by leaf name, plus the row-hash conveniences. Nothing here
+  # borrows C memory.
+  class PlanResult
+    # @api private — one bulk read per column; the C result is freed
+    # on return (PlanResult owns plain Ruby data)
+    def self.read(raw, src, names)
+      rows = ::Yeptris::FFI.yeptris_plan_result_rows(raw)
+      columns = {}
+      names.each_with_index do |name, c|
+        kind = ::Yeptris::FFI.yeptris_plan_result_kind(raw, c)
+        columns[name] = read_column(raw, c, kind, rows, src)
+      end
+      new(rows, names, columns)
+    end
+
+    def self.read_column(raw, c, kind, rows, src)
+      return [] if rows.zero?
+
+      nulls = ::Yeptris::FFI.yeptris_plan_result_nulls(raw, c).read_bytes(rows).unpack("C*")
+      has_nulls = nulls.include?(1)
+      case kind
+      when 0 # int: the unpacked lane IS the column when no nulls
+        ints = ::Yeptris::FFI.yeptris_plan_result_ints(raw, c).read_bytes(rows * 8).unpack("q<*")
+        return ints unless has_nulls
+
+        Array.new(rows) { |i| nulls[i] == 1 ? nil : ints[i] }
+      when 1
+        floats = ::Yeptris::FFI.yeptris_plan_result_floats(raw, c).read_bytes(rows * 8).unpack("E*")
+        return floats unless has_nulls
+
+        Array.new(rows) { |i| nulls[i] == 1 ? nil : floats[i] }
+      when 3 # bool: 0/1 → false/true (always maps)
+        ints = ::Yeptris::FFI.yeptris_plan_result_ints(raw, c).read_bytes(rows * 8).unpack("q<*")
+        Array.new(rows) { |i| nulls[i] == 1 ? nil : ints[i] == 1 }
+      else # str: spans into the source, escapes decoded on the Ruby side
+        offs = ::Yeptris::FFI.yeptris_plan_result_str_offs(raw, c).read_bytes(rows * 4).unpack("V*")
+        lens = ::Yeptris::FFI.yeptris_plan_result_str_lens(raw, c).read_bytes(rows * 4).unpack("V*")
+        decode = ::Yeptris::JSON.method(:decode_span)
+        Array.new(rows) { |i| nulls[i] == 1 ? nil : decode.call(src, offs[i], lens[i]) }
+      end
+    end
+    private_class_method :read_column
+
+    # @api private — the DOM leg's read: typed lanes as above; str
+    # columns materialize from the (ptr,len) views into the document
+    def self.read_dom(raw, names)
+      rows = ::Yeptris::FFI.yeptris_plan_result_rows(raw)
+      columns = {}
+      names.each_with_index do |name, c|
+        kind = ::Yeptris::FFI.yeptris_plan_result_kind(raw, c)
+        if kind == 2
+          nulls = ::Yeptris::FFI.yeptris_plan_result_nulls(raw, c).read_bytes(rows).unpack("C*")
+          base = ::FFI::Pointer.new(::Yeptris::FFI.yeptris_plan_result_strs(raw, c))
+          stride = ::Yeptris::FFI::PlanStr.size
+          columns[name] = Array.new(rows) do |i|
+            next nil if nulls[i] == 1
+
+            v = ::Yeptris::FFI::PlanStr.new(base + i * stride)
+            v[:p].read_bytes(v[:len]).force_encoding(Encoding::UTF_8)
+          end
+        else
+          columns[name] = read_column(raw, c, kind, rows, nil)
+        end
+      end
+      new(rows, names, columns)
+    end
+
+    def initialize(rows, names, columns)
+      @rows = rows
+      @names = names
+      @columns = columns
+    end
+
+    def rows
+      @rows
+    end
+    alias count rows
+
+    # The planned leaf names, in spec order.
+    def names
+      @names.dup
+    end
+
+    # One typed column: an Array of rows values (nil where the leaf
+    # was null, missing, or shape-mismatched).
+    def column(name)
+      @columns.fetch(name) { raise ArgumentError, "no planned column #{name.inspect}" }
+    end
+
+    # Row +i+ as a Hash of the planned leaves (nil out of range).
+    def at(i)
+      return nil if i >= @rows || i.negative?
+
+      row = {}
+      @names.each { |n| row[n] = @columns[n][i] }
+      row
+    end
+
+    # Rows as Hashes — the tree-shaped convenience; column readers
+    # are the allocation-lean path.
+    def to_a
+      Array.new(@rows) { |i| at(i) }
+    end
+    alias to_ruby to_a # leptris's eager-tree vocabulary
+
+    def each_row(&block)
+      return enum_for(:each_row) unless block
+
+      @rows.times { |i| block.call(at(i)) }
+    end
+  end
+end

data/lib/yeptris/json.rb

@@ -17,6 +17,8 @@ module Yeptris
     class Error < ::Yeptris::Error; end
     class ParseError < Error; end
 
+    autoload :Descriptor, "yeptris/json/descriptor"
+
     # json gem 3.0 made duplicate keys an error by DEFAULT; 2.x is
     # last-wins. The parity target is the RESOLVED json gem's own
     # behavior — strictness follows it (issue #37, found by canon's

data/lib/yeptris/yaml/descriptor.rb

@@ -0,0 +1,54 @@
+# frozen_string_literal: true
+
+# The YAML leg of the Descriptor plan walk (yeptris#293 slice three,
+# over the C DOM plan walk of libyeptris v0.6.4): the same compiled
+# row shape as Yeptris::JSON::Descriptor, applied to a parsed YAML
+# document — block YAML hydrates to typed columns with no
+# intermediate Ruby tree.
+#
+#     descriptor = Yeptris::YAML::Descriptor.build(
+#       kind: :map, path: ["data", "items"], # segmented paths work
+#       children: [{ name: "id", kind: :int }, { name: "name", kind: :str }])
+#     descriptor.walk(yaml).column("id") # => [1, 2, ...]
+#
+# Typed extraction rides the parse-time tag ids (schema: selects the
+# resolver — :compat_11 keeps the Psych/libyaml implicit typing
+# YAML.load uses); string columns are materialized from the
+# document's regions before the result returns (nothing borrowed).
+class Yeptris::YAML::Descriptor < ::Yeptris::JSON::Descriptor
+  # schema: :core_12 (YAML 1.2) or :compat_11 (the Psych-compatible
+  # implicit typing Yeptris::YAML.load uses).
+  def initialize(handle, names, schema)
+    super(handle, names)
+    @schema = schema
+  end
+
+  # @api private — the compile rides the JSON Descriptor's grammar
+  # (inherited private class method); schema selects the resolver
+  def self.build(kind:, path: nil, children:, schema: :compat_11)
+    handle, names = compile_handle(kind, path, children)
+    new(handle, names, schema)
+  end
+
+  def walk(yaml)
+    src = ::Yeptris.read_input(yaml).to_s
+    doc = ::Yeptris::Document.parse(src, schema: @schema)
+    begin
+      st = ::FFI::MemoryPointer.new(:int)
+      raw = ::Yeptris::FFI.yeptris_document_plan_walk(doc.c_ptr, @handle, st)
+      if raw.null?
+        raise ::Yeptris::JSON::Descriptor::Error,
+              "plan walk failed (status=#{st.read_int}) — the document shape " \
+              "disagrees with the plan"
+      end
+
+      begin
+        ::Yeptris::JSON::Descriptor::PlanResult.read_dom(raw, @names)
+      ensure
+        ::Yeptris::FFI.yeptris_plan_result_free(raw)
+      end
+    ensure
+      doc.free
+    end
+  end
+end

data/lib/yeptris/yaml.rb

@@ -9,6 +9,8 @@ module Yeptris
   # the recorder-driven Visitors in phase B; this is the yeptris-native
   # face users target first).
   module YAML
+    autoload :Descriptor, "yeptris/yaml/descriptor"
+
     module_function
 
     # Loads the FIRST document of a YAML stream as native Ruby objects.

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 = "0.6.3.2".freeze
+  VERSION = "0.6.4.1".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: 0.6.3.2
+  version: 0.6.4.1
 platform: ruby
 authors:
 - Ribose Inc.
@@ -42,6 +42,7 @@ files:
 - lib/yeptris/document.rb
 - lib/yeptris/ffi.rb
 - lib/yeptris/json.rb
+- lib/yeptris/json/descriptor.rb
 - lib/yeptris/materializer.rb
 - lib/yeptris/node.rb
 - lib/yeptris/psych.rb
@@ -54,6 +55,7 @@ files:
 - lib/yeptris/schema.rb
 - lib/yeptris/valueml.rb
 - lib/yeptris/yaml.rb
+- lib/yeptris/yaml/descriptor.rb
 homepage: https://github.com/leptris/yeptris
 licenses:
 - MIT