andromeda_cms 0.6.2 → 0.7.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 64670cf71e6544d5427eeb01b7fbd90eda3a717dbb28379dd836df5662d24563
-  data.tar.gz: 7966c38ad14347e3161f3a8f7018daa5438dddb00aacfee099e9d2c014e500d9
+  metadata.gz: df0dda6ff30ba8c1244c31c1af9951993028e7056ae943b4cca5513c14c499d2
+  data.tar.gz: 76f8e9c8de6a02d0ca9c8d22e40946c5e3bfb5961fa6888781add1d0a7e3195a
 SHA512:
-  metadata.gz: b6fe4b7a80ca8828ef6ecfeb77ec1ceb8ec5717df836bd6f68e328adc5e9fdb162e1607d26b92167d7420639874c5a203ea25bcf7325d5952a4acbb1f76b76c3
-  data.tar.gz: 5db3fa88aab475e40522a9024d8d22f8887ec106fad290fbfec7b18addb13ab68dba3db3e48e080d43603c3c50a6439e14c6b0fec912ba40a8dba270fae82799
+  metadata.gz: eac17b62179d28bb3ce45a772d0a9a2a630098c841d8468436d4a9b3239c9fb83d8f841252ffe42500a420cc11679e75d0bb3b0333a72ef535147bdd0021df60
+  data.tar.gz: 3dfae8e7efde1146d5b8881c8654acc4b527d1dc65b661259a658fa82ef3aef5cffc59be30e483f53fc7e3e892e37071e68cf46a95905e530aa1e6af85cb3a47

data/CHANGELOG.md

@@ -5,6 +5,43 @@ All notable changes to this project are documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.7.0] - 2026-10-16
+
+### Changed
+
+- YAML frontmatter is now read the way Astro 7.3.6 and later read it. Astro
+  replaced js-yaml with the `yaml` package (core schema, timestamps, merge
+  keys), and Andromeda follows. Plain scalars that only YAML 1.1 or Psych
+  treat as numbers are now strings, as they are in Astro: values with
+  underscores (`1_000`, `0x1_F`), the `0b` prefix, and signed hexadecimal
+  or octal (`-0x1F`). Numbers in the core schema are unchanged, including
+  `010` (10), `0o17`, `0x1F` and `1e3`.
+- A mapping that repeats a key now raises `Andromeda::SyntaxError` at the
+  second key, with its line and column, as Astro does. It used to keep the
+  last value silently. Keys are compared after resolution, so `1` and `01`
+  collide while `1` and `"1"` do not; repeated `<<` merge keys are allowed.
+  Content that repeats a key will fail to build until the duplicate is
+  removed.
+
+## [0.6.3] - 2026-10-16
+
+### Fixed
+
+- `andromeda:import_astro` no longer silently mishandles a `glob()` `pattern`
+  it cannot carry over. An array pattern is reduced to its first
+  non-exclusion element (it used to take the first string, even a `!`
+  exclusion), and a note in the report's "Needs a human" section plus a
+  `# TODO` above the generated `collection` line name what was lost:
+  exclusions that are not applied (those files are still copied and
+  loaded) and extra patterns that were dropped (those files are copied but
+  not loaded). An extglob pattern such as `!(drafts)/**/*.md`, which Ruby's
+  glob cannot match, is flagged the same way instead of leaving the
+  collection empty.
+- The gem now requires `tomlrb ~> 2.0`. tomlrb 3.0 removed
+  `Tomlrb::Scanner`, which TOML frontmatter parsing uses to report the line
+  of a syntax error, so on Ruby 3.3 and later a fresh install picked up 3.0
+  and every TOML frontmatter raised `NameError`.
+
 ## [0.6.2] - 2026-10-09
 
 ### Fixed
@@ -189,6 +226,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Generators: `andromeda:install`, `andromeda:collection`,
   `andromeda:component` and `andromeda:import_astro`.
 
+[0.7.0]: https://github.com/AndromedaCMS/andromeda_cms/releases/tag/v0.7.0
+[0.6.3]: https://github.com/AndromedaCMS/andromeda_cms/releases/tag/v0.6.3
 [0.6.2]: https://github.com/AndromedaCMS/andromeda_cms/releases/tag/v0.6.2
 [0.6.1]: https://github.com/AndromedaCMS/andromeda_cms/releases/tag/v0.6.1
 [0.6.0]: https://github.com/AndromedaCMS/andromeda_cms/releases/tag/v0.6.0

data/lib/andromeda/frontmatter.rb

@@ -25,45 +25,53 @@ module Andromeda
     # a non-blank line is seen, no `---`/`+++` after it is a fence.
     LEADING_RE = /\A\uFEFF?(?:[ \t]*\r?\n)*/
 
-    # Psych resolves plain scalars by YAML 1.1 rules; js-yaml, which Astro
-    # uses, follows YAML 1.2's core schema. The two disagree on values that
-    # really do appear in hand-written frontmatter, and a silent disagreement
-    # is worse than an error: `time: 12:30` becoming the integer 45000 passes
-    # every schema check and then displays nonsense. Each rule below mirrors
-    # what js-yaml 4 does with the same plain scalar.
+    # Psych resolves plain scalars by YAML 1.1 rules; Astro 7.3.6 and later
+    # read frontmatter with the `yaml` package (core schema, `timestamp` tag,
+    # merge keys), which follows YAML 1.2's core schema. The two disagree on
+    # values that really do appear in hand-written frontmatter, and a silent
+    # disagreement is worse than an error: `time: 12:30` becoming the integer
+    # 45000 passes every schema check and then displays nonsense. Each rule
+    # below mirrors what that parser does with the same plain scalar.
     class ScalarScanner < Psych::ScalarScanner
       # 1.1-only booleans; 1.2 keeps them as strings.
       YAML_1_1_ONLY_BOOL = /\A(?:[Yy]es|YES|[Nn]o|NO|[Oo]n|ON|[Oo]ff|OFF|[Yy]|[Nn])\z/
       # 1.1 base-60 (`1:30`) and comma-grouped (`1,000`) numbers; 1.2 has
       # neither, so these stay strings.
       YAML_1_1_ONLY_NUMBER = /\A[-+]?[0-9][0-9_]*(?:(?::[0-5]?[0-9])+|(?:,[0-9_]+)+)(?:\.[0-9_]*)?\z/
-      # 1.1 reads a leading zero as octal (`010` is 8); 1.2 reads it as decimal.
-      LEADING_ZERO_DECIMAL = /\A[-+]?0[0-9_]+\z/
-      # 1.2's explicit octal prefix, unknown to Psych.
-      OCTAL = /\A[-+]?0o[0-7_]+\z/
-      # 1.2 accepts an exponent without a decimal point (`1e3`); Psych does not.
-      EXPONENT_FLOAT = /\A[-+]?(?:[0-9][0-9_]*(?:\.[0-9_]*)?|\.[0-9_]+)[eE][-+]?[0-9]+\z/
+      # The core schema's integers. A leading zero is plain decimal (`010` is
+      # 10, where 1.1 reads octal); `0o` octal and `0x` hex take no sign.
+      INTEGER = /\A[-+]?[0-9]+\z/
+      OCTAL = /\A0o[0-7]+\z/
+      HEX = /\A0x[0-9a-fA-F]+\z/
+      # The core schema's floats, including an exponent without a decimal
+      # point (`1e3`), which Psych rejects. `.inf`/`.nan` are left to Psych.
+      FLOAT = /\A[-+]?(?:\.[0-9]+|[0-9]+(?:\.[0-9]*)?)(?:[eE][-+]?[0-9]+)?\z/
+      # Number-like spellings Psych accepts but the core schema does not
+      # (`1_000`, `0b1010`, `-0x1F`, `0o1_7`); they stay strings.
+      NON_CORE_NUMBER = /\A[-+]?(?:0[xXbBoO][0-9a-fA-F_]+|(?:[0-9][0-9_]*(?:\.[0-9_]*)?|\.[0-9_]+)(?:[eE][-+]?[0-9_]+)?)\z/
 
       def tokenize(string)
         case string
         when YAML_1_1_ONLY_BOOL, YAML_1_1_ONLY_NUMBER then string
-        when LEADING_ZERO_DECIMAL then Integer(string.delete("_"), 10)
-        when OCTAL then Integer(string.delete("_"))
-        when EXPONENT_FLOAT then Float(normalize_float(string.delete("_")))
+        when INTEGER then Integer(string, 10)
+        when OCTAL then Integer(string[2..], 8)
+        when HEX then Integer(string, 16)
+        when FLOAT then Float(normalize_float(string))
+        when NON_CORE_NUMBER then string
         else super
         end
       end
 
       private
 
-      # Ruby's Float() rejects the `1.e3` and `.5e3` spellings YAML allows.
+      # Ruby's Float() rejects the `1.`, `1.e3` and `.5e3` spellings YAML allows.
       def normalize_float(string)
-        string.sub(/\.(?=[eE])/, ".0").sub(/\A([-+]?)\./, "\\10.")
+        string.sub(/\.(?=[eE]|\z)/, ".0").sub(/\A([-+]?)\./, "\\10.")
       end
     end
 
-    # Restricts `!!` tags to Date/Time (matching js-yaml's default schema,
-    # which also has no notion of arbitrary Ruby classes) while still
+    # Restricts `!!` tags to Date/Time (matching the `yaml` package's core
+    # schema, which also has no notion of arbitrary Ruby classes) while still
     # permitting the two classes plain `YYYY-MM-DD`/timestamp scalars resolve
     # to, so unquoted dates round-trip instead of raising
     # Psych::DisallowedClass.
@@ -76,7 +84,7 @@ module Andromeda
     end
 
     # Plain (non-"safe") visitor so anchors/aliases resolve instead of
-    # raising -- js-yaml allows them by default, and copied Astro content may
+    # raising -- the `yaml` package resolves them by default, and copied Astro content may
     # rely on them.
     class ToRuby < Psych::Visitors::ToRuby
       def initialize
@@ -199,6 +207,8 @@ module Andromeda
       return {} if node.nil?
 
       reject_alias_explosion!(node, path: path, content_start_line: content_start_line)
+      reject_duplicate_keys!(node, ScalarScanner.new(ClassLoader.new), path: path,
+                                   content_start_line: content_start_line)
       ToRuby.new.accept(node) || {}
     rescue Psych::SyntaxError => e
       # Psych's own #line/#column are 0-based and relative to `text` (the
@@ -305,6 +315,36 @@ module Andromeda
     end
     private_class_method :expanded_size
 
+    # Astro's `yaml` parser reports a repeated mapping key as an error, where
+    # Psych silently keeps the last value. Keys are compared after resolution,
+    # so `1` and `01` collide while `1` and `"1"` do not. Plain `<<` merge
+    # keys, alias keys and collection keys are left out.
+    def reject_duplicate_keys!(node, scanner, path:, content_start_line:)
+      if node.is_a?(Psych::Nodes::Mapping)
+        seen = {}
+        node.children.each_slice(2) do |key, _value|
+          next unless key.is_a?(Psych::Nodes::Scalar)
+          next if key.plain && key.value == "<<"
+
+          resolved = key.plain ? scanner.tokenize(key.value) : key.value
+          resolved = resolved.to_f if resolved.is_a?(Numeric)
+          if seen.key?(resolved)
+            raise syntax_error("duplicate key #{key.value.inspect}", path: path,
+                                                                    content_start_line: content_start_line,
+                                                                    relative_line: key.start_line,
+                                                                    relative_column: key.start_column)
+          end
+
+          seen[resolved] = true
+        end
+      end
+
+      node.children.to_a.each do |child|
+        reject_duplicate_keys!(child, scanner, path: path, content_start_line: content_start_line)
+      end
+    end
+    private_class_method :reject_duplicate_keys!
+
     def syntax_error(message, path:, content_start_line:, relative_line:, relative_column:)
       Andromeda::SyntaxError.new(
         message,

data/lib/andromeda/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Andromeda
-  VERSION = "0.6.2"
+  VERSION = "0.7.0"
 end

data/lib/generators/andromeda/import_astro/content_config_converter.rb

@@ -18,7 +18,19 @@ module Andromeda
       # already makes for MDX import statements. Anything shaped differently
       # falls through to `Attribute#unsupported?` rather than guessing.
       module ContentConfigConverter
-        Collection = Struct.new(:name, :base, :pattern, :attributes, keyword_init: true)
+        # `notes` are things about the collection itself (not one field)
+        # that still need a human -- e.g. a `pattern` this importer cannot
+        # carry over as written. Defaults to empty so existing
+        # `Collection.new` calls stay valid.
+        Collection = Struct.new(:name, :base, :pattern, :attributes, :notes, keyword_init: true) do
+          def initialize(notes: [], **rest) = super(notes: notes, **rest)
+        end
+
+        DEFAULT_PATTERN = "**/*.{md,mdx}"
+
+        # Bash-style extglob openers (`!(drafts)`, `@(a|b)`, ...), which
+        # Astro's glob loader accepts but Ruby's `Dir.glob` does not.
+        EXTGLOB_RE = /[!@+*?]\(/.freeze
 
         # One frontmatter field. `ruby_type` is nil when the Zod expression
         # could not be recognized at all -- `source` (the original
@@ -65,16 +77,64 @@ module Andromeda
         def build_collection(name, body, astro_root:)
           body = BalancedScanner.strip_comments(body)
           base = extract_string(body, /base\s*:/) || "./src/content/#{name}"
-          pattern = extract_string(body, /pattern\s*:/) || "**/*.{md,mdx}"
+          pattern, notes = extract_pattern(body)
 
           Collection.new(
             name: name,
             base: File.expand_path(base, astro_root),
             pattern: pattern,
-            attributes: extract_attributes(body)
+            attributes: extract_attributes(body),
+            notes: notes
           )
         end
 
+        # Astro's `glob({ pattern: ... })` also takes an array (with `!`
+        # exclusions) and extglob syntax -- neither of which `Dir.glob`
+        # understands. Rather than silently dropping them, the closest
+        # usable pattern is kept and a note says what was lost.
+        #
+        # @return [Array(String, Array<String>)] the pattern and any notes.
+        def extract_pattern(body)
+          match = /pattern\s*:\s*\[/.match(body)
+          return extract_array_pattern(body, match.end(0) - 1) if match
+
+          pattern = extract_string(body, /pattern\s*:/) || DEFAULT_PATTERN
+          return [pattern, []] unless pattern.match?(EXTGLOB_RE)
+
+          [pattern, ["pattern `#{pattern}` uses extglob syntax, which Ruby's glob does not support -- " \
+                     "no files will match this pattern"]]
+        end
+
+        # Only what was actually lost gets a note: `!` exclusions (still
+        # copied and loaded) and any further non-exclusion patterns (copied
+        # but not loaded). A one-element or empty array loses nothing.
+        def extract_array_pattern(body, open_index)
+          close = BalancedScanner.matching_close(body, open_index)
+          # Squished onto one line: the note also becomes a `# TODO` comment.
+          source = body[open_index..close].gsub(/\s+/, " ")
+          elements = BalancedScanner.split_top_level(body[(open_index + 1)...close]).map do |literal|
+            literal.gsub(/\A(['"`])(.*)\1\z/, '\2')
+          end.reject(&:empty?)
+          exclusions, includes = elements.partition { |element| element.start_with?("!") }
+          pattern = includes.first || DEFAULT_PATTERN
+          dropped = includes.drop(1)
+
+          notes = []
+          unless exclusions.empty?
+            notes << "exclusions #{backticked(exclusions)} from pattern array #{source} are not applied, " \
+                     "so the files they exclude are still copied and loaded"
+          end
+          unless dropped.empty?
+            notes << "patterns #{backticked(dropped)} from pattern array #{source} were dropped " \
+                     "(only `#{pattern}` was kept), so files matching only them are copied but not loaded"
+          end
+          [pattern, notes]
+        end
+
+        def backticked(elements)
+          elements.map { |element| "`#{element}`" }.join(", ")
+        end
+
         # @return [String, nil] the quoted value right after `label_re`, e.g.
         #   `base: './src/content/blog'` -> `"./src/content/blog"`.
         def extract_string(body, label_re)

data/lib/generators/andromeda/import_astro/import_astro_generator.rb

@@ -80,7 +80,10 @@ module Andromeda
             fallback_collections
           end
 
-        @collections.each { |collection| refine_with_json_schema!(collection) }
+        @collections.each do |collection|
+          refine_with_json_schema!(collection)
+          collection.notes.each { |note| @report[:notes] << "#{collection.name}: #{note}" }
+        end
       end
 
       # The fallback clause: every top-level directory under
@@ -231,6 +234,9 @@ module Andromeda
         lines = ["# frozen_string_literal: true", ""]
         namespace_modules.each_with_index { |mod, i| lines << "#{"  " * i}module #{mod}" }
         lines << "#{indent}class #{collection.name.to_s.singularize.camelize} < Andromeda::Entry"
+        collection.notes.each do |note|
+          lines << "#{indent}  # TODO: #{note} (content.config.ts) -- fix `pattern` by hand."
+        end
         lines << "#{indent}  collection :#{collection.name}, base: \"app/content/#{collection.name}\", " \
                  "pattern: \"#{collection.pattern}\""
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: andromeda_cms
 version: !ruby/object:Gem::Version
-  version: 0.6.2
+  version: 0.7.0
 platform: ruby
 authors:
 - LIMHAUS Inc.
@@ -55,14 +55,14 @@ dependencies:
   name: tomlrb
   requirement: !ruby/object:Gem::Requirement
     requirements:
-    - - ">="
+    - - "~>"
       - !ruby/object:Gem::Version
         version: '2.0'
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
-    - - ">="
+    - - "~>"
       - !ruby/object:Gem::Version
         version: '2.0'
 - !ruby/object:Gem::Dependency