[ARVADOS] created: 64046fa6553461bed50d36f585ea1762e4ff6fdd

git at public.curoverse.com git at public.curoverse.com
Sun Mar 8 17:12:52 EDT 2015


        at  64046fa6553461bed50d36f585ea1762e4ff6fdd (commit)


commit 64046fa6553461bed50d36f585ea1762e4ff6fdd
Author: Brett Smith <brett at curoverse.com>
Date:   Sun Mar 8 17:12:40 2015 -0400

    5104: Add Collection class to the Ruby SDK.
    
    This provides an interface to do filesystem-like operations on a tree
    expressed in a manifest.  It currently supports copy, rename, and
    remove operations, and will give us common ground for further work in
    Ruby components.

diff --git a/sdk/ruby/lib/arvados/collection.rb b/sdk/ruby/lib/arvados/collection.rb
new file mode 100644
index 0000000..94f5138
--- /dev/null
+++ b/sdk/ruby/lib/arvados/collection.rb
@@ -0,0 +1,403 @@
+require "arvados/keep"
+
+module Arv
+  class Collection
+    def initialize(manifest_text="")
+      @tree = CollectionStream.new(".")
+      @manifest_text = ""
+      import_manifest!(manifest_text)
+    end
+
+    def manifest_text
+      @manifest_text ||= @tree.manifest_text
+    end
+
+    def import_manifest!(manifest_text)
+      manifest = Keep::Manifest.new(manifest_text)
+      manifest.each_line do |stream_root, locators, file_specs|
+        if stream_root.empty? or locators.empty? or file_specs.empty?
+          raise ArgumentError.new("manifest text includes malformed line")
+        end
+        file_specs.map { |s| manifest.split_file_token(s) }.
+            each do |file_start, file_len, file_path|
+          @tree.file_at(normalize_path(stream_root, file_path)).
+            add_range(locators, file_start, file_len)
+        end
+      end
+      if @manifest_text == ""
+        @manifest_text = manifest_text
+        self
+      else
+        modified!
+      end
+    end
+
+    def normalize!
+      # We generate normalized manifests, so all we have to do is force
+      # regeneration.
+      modified!
+    end
+
+    def copy!(source, target, source_collection=nil)
+      copy(:merge, source, target, source_collection)
+    end
+
+    def rename!(source, target)
+      copy(:add_copy, source, target) { remove!(source, recursive: true) }
+    end
+
+    def remove!(path, opts={})
+      stream, name = find(path)
+      if name.nil?
+        return self if @tree.leaf?
+        @tree = CollectionStream.new(".")
+      else
+        stream.delete(name, opts)
+      end
+      modified!
+    end
+
+    protected
+
+    def find(*parts)
+      normpath = normalize_path(*parts)
+      if normpath.empty?
+        [@tree, nil]
+      else
+        @tree.find(normpath)
+      end
+    end
+
+    private
+
+    def copy(copy_method, source, target, source_collection=nil)
+      # Find the item at path `source` in `source_collection`, find the
+      # destination stream at path `target`, and use `copy_method` to copy
+      # the found object there.  If a block is passed in, it will be called
+      # right before we do the actual copy, after we confirm that everything
+      # is found and can be copied.
+      source_collection = self if source_collection.nil?
+      src_stream, src_tail = source_collection.find(source)
+      dst_stream, dst_tail = find(target)
+      if (source_collection.equal?(self) and
+          (src_stream.path == dst_stream.path) and (src_tail == dst_tail))
+        return self
+      elsif src_tail.nil?
+        src_item = src_stream
+        src_tail = src_stream.name
+      else
+        src_item = src_stream[src_tail]
+      end
+      dst_tail ||= src_tail
+      check_method = "check_can_#{copy_method}".to_sym
+      begin
+        # Find out if `target` refers to a stream we should copy into.
+        tail_stream = dst_stream[dst_tail]
+        tail_stream.send(check_method, src_item, src_tail)
+      rescue Errno::ENOENT, Errno::ENOTDIR
+        # It does not.  Check that we can copy `source` to the full
+        # path specified by `target`.
+        dst_stream.send(check_method, src_item, dst_tail)
+        target_name = dst_tail
+      else
+        # Yes, `target` is a stream.  Copy the item at `source` into it with
+        # the same name.
+        dst_stream = tail_stream
+        target_name = src_tail
+      end
+      # At this point, we know the operation will work.  Call any block as
+      # a pre-copy hook.
+      if block_given?
+        yield
+        # Re-find the destination stream, in case the block removed
+        # the original (that's how rename is implemented).
+        dst_path = normalize_path(dst_stream.path)
+        if dst_path.empty?
+          dst_stream = @tree
+        else
+          dst_stream = @tree.stream_at(dst_path)
+        end
+      end
+      dst_stream.send(copy_method, src_item, target_name)
+      modified!
+    end
+
+    def modified!
+      @manifest_text = nil
+      self
+    end
+
+    def normalize_path(*parts)
+      path = File.join(*parts)
+      raise ArgumentError.new("empty path") if path.empty?
+      path.sub(/^\.(\/|$)/, "")
+    end
+
+    class CollectionItem
+      attr_reader :path, :name
+
+      def initialize(path)
+        @path = path
+        @name = File.basename(path)
+      end
+    end
+
+    LocatorRange = Struct.new(:locators, :start_pos, :length)
+
+    class CollectionFile < CollectionItem
+      def initialize(path)
+        super
+        @ranges = []
+      end
+
+      def self.human_name
+        "file"
+      end
+
+      def leaf?
+        true
+      end
+
+      def add_range(locators, start_pos, length)
+        # Given an array of locators, and this file's start position and
+        # length within them, store a LocatorRange with information about
+        # the locators actually used.
+        loc_sizes = locators.map { |s| Keep::Locator.parse(s).size.to_i }
+        start_index, start_pos = loc_size_index(loc_sizes, start_pos, 0, :>=)
+        end_index, _ = loc_size_index(loc_sizes, length, start_index, :>)
+        @ranges << LocatorRange.
+          new(locators[start_index..end_index], start_pos, length)
+      end
+
+      def each_range(&block)
+        @ranges.each(&block)
+      end
+
+      def check_can_add_copy(src_item, name)
+        raise Errno::ENOTDIR.new(path)
+      end
+
+      alias_method :check_can_merge, :check_can_add_copy
+
+      def copy_named(copy_path)
+        copy = self.class.new(copy_path)
+        each_range { |range| copy.add_range(*range) }
+        copy
+      end
+
+      private
+
+      def loc_size_index(loc_sizes, length, index, comp_op)
+        # Pass in an array of locator size hints (integers).  Starting from
+        # `index`, step through the size array until they provide a number
+        # of bytes that is `comp_op` (:>= or :>) to `length`.  Return the
+        # index of the end locator and the amount of data to read from it.
+        while length.send(comp_op, loc_sizes[index])
+          index += 1
+          length -= loc_sizes[index]
+        end
+        [index, length]
+      end
+    end
+
+    class CollectionStream < CollectionItem
+      def initialize(path)
+        super
+        @items = {}
+      end
+
+      def self.human_name
+        "stream"
+      end
+
+      def leaf?
+        items.empty?
+      end
+
+      def [](key)
+        items[key] or
+          raise Errno::ENOENT.new("%p not found in %p" % [key, path])
+      end
+
+      def delete(name, opts={})
+        item = self[name]
+        if item.leaf? or opts[:recursive]
+          items.delete(name)
+        else
+          raise Errno::ENOTEMPTY.new(path)
+        end
+      end
+
+      def find(find_path)
+        # Given a POSIX-style path, return the CollectionStream that
+        # contains the object at that path, and the name of the object
+        # inside it.
+        components = find_path.split("/")
+        tail = components.pop
+        [components.reduce(self, :[]), tail]
+      end
+
+      def stream_at(find_path)
+        key, rest = find_path.split("/", 2)
+        next_stream = get_or_new(key, CollectionStream)
+        if rest.nil?
+          next_stream
+        else
+          next_stream.stream_at(rest)
+        end
+      end
+
+      def file_at(find_path)
+        stream_path, _, file_name = find_path.rpartition("/")
+        if stream_path.empty?
+          get_or_new(file_name, CollectionFile)
+        else
+          stream_at(stream_path).file_at(file_name)
+        end
+      end
+
+      def manifest_text
+        # Return a string with the normalized manifest text for this stream,
+        # including all substreams.
+        file_keys, stream_keys = items.keys.sort.partition do |key|
+          items[key].is_a?(CollectionFile)
+        end
+        my_line = StreamManifest.new(path)
+        file_keys.each do |file_name|
+          my_line.add_file(items[file_name])
+        end
+        sub_lines = stream_keys.map do |sub_name|
+          items[sub_name].manifest_text
+        end
+        my_line.to_s + sub_lines.join("")
+      end
+
+      def check_can_add_copy(src_item, key)
+        if existing = check_can_merge(src_item, key) and not existing.leaf?
+          raise Errno::ENOTEMPTY.new(existing.path)
+        end
+      end
+
+      def check_can_merge(src_item, key)
+        if existing = items[key] and (existing.class != src_item.class)
+          raise Errno::ENOTDIR.new(existing.path)
+        end
+        existing
+      end
+
+      def add_copy(src_item, key)
+        items[key] = src_item.copy_named("#{path}/#{key}")
+      end
+
+      def merge(src_item, key)
+        # Do a recursive copy of the collection item `src_item` to destination
+        # `key`.  If a simple copy is safe, do that; otherwise, recursively
+        # merge the contents of the stream `src_item` into the stream at
+        # `key`.
+        begin
+          check_can_add_copy(src_item, key)
+          add_copy(src_item, key)
+        rescue Errno::ENOTEMPTY
+          dest = self[key]
+          error = nil
+          # Copy as much as possible, then raise any error encountered.
+          src_item.items.each_pair do |sub_key, sub_item|
+            begin
+              dest.merge(sub_item, sub_key)
+            rescue Errno::ENOTDIR => error
+            end
+          end
+          raise error unless error.nil?
+        end
+      end
+
+      def copy_named(copy_path)
+        copy = self.class.new(copy_path)
+        items.each_pair do |key, item|
+          copy.add_copy(item, key)
+        end
+        copy
+      end
+
+      protected
+
+      attr_reader :items
+
+      private
+
+      def get_or_new(key, klass)
+        # Return the collection item at `key` and ensure that it's a `klass`.
+        # If `key` does not exist, create a new `klass` there.
+        # If the value for `key` is not a `klass`, raise an ArgumentError.
+        item = items[key]
+        if item.nil?
+          items[key] = klass.new("#{path}/#{key}")
+        elsif not item.is_a?(klass)
+          raise ArgumentError.
+            new("in stream %p, %p is a %s, not a %s" %
+                [path, key, items[key].class.human_name, klass.human_name])
+        else
+          item
+        end
+      end
+    end
+
+    class StreamManifest
+      # Build a manifest text for a single stream, without substreams.
+
+      def initialize(name)
+        @name = name
+        @locators = []
+        @loc_sizes = []
+        @file_specs = []
+      end
+
+      def add_file(coll_file)
+        coll_file.each_range do |range|
+          add(coll_file.name, *range)
+        end
+      end
+
+      def to_s
+        if @file_specs.empty?
+          ""
+        else
+          "%s %s %s\n" % [escape_name(@name), @locators.join(" "),
+                          @file_specs.join(" ")]
+        end
+      end
+
+      private
+
+      def add(file_name, loc_a, file_start, file_len)
+        # Ensure that the locators in loc_a appear in this locator in sequence,
+        # adding as few as possible.  Save a new file spec based on those
+        # locators' position.
+        loc_size = @locators.size
+        add_size = loc_a.size
+        loc_ii = 0
+        add_ii = 0
+        while (loc_ii < loc_size) and (add_ii < add_size)
+          if @locators[loc_ii] == loc_a[add_ii]
+            add_ii += 1
+          else
+            add_ii = 0
+          end
+          loc_ii += 1
+        end
+        loc_ii -= add_ii
+        to_add = loc_a[add_ii, add_size] || []
+        @locators += to_add
+        @loc_sizes += to_add.map { |s| Keep::Locator.parse(s).size.to_i }
+        start = @loc_sizes[0, loc_ii].reduce(0, &:+) + file_start
+        @file_specs << "#{start}:#{file_len}:#{escape_name(file_name)}"
+      end
+
+      def escape_name(name)
+        name.gsub(/\\/, "\\\\\\\\").gsub(/\s/) do |s|
+          s.each_byte.map { |c| "\\%03o" % c }.join("")
+        end
+      end
+    end
+  end
+end
diff --git a/sdk/ruby/test/sdk_fixtures.rb b/sdk/ruby/test/sdk_fixtures.rb
index 8af81cd..52d7377 100644
--- a/sdk/ruby/test/sdk_fixtures.rb
+++ b/sdk/ruby/test/sdk_fixtures.rb
@@ -7,6 +7,10 @@ module SDKFixtures
     def random_block(size=nil)
       sprintf("%032x+%d", rand(16 ** 32), size || rand(64 * 1024 * 1024))
     end
+
+    def random_blocks(count, size=nil)
+      (0...count).map { |_| random_block(size) }
+    end
   end
 
   extend StaticMethods
@@ -50,4 +54,12 @@ module SDKFixtures
   NONNORMALIZED_MANIFEST =
     ["./dir2 #{random_block} 0:0:z 0:0:y 0:0:x",
      "./dir1 #{random_block} 0:0:p 0:0:o 0:0:n\n"].join("\n")
+
+  ### Non-tree manifests
+  # These manifests follow the spec, but they express a structure that can't
+  # can't be represented by a POSIX filesystem tree.  For example, there's a
+  # name conflict between a stream and a filename.
+  NAME_CONFLICT_MANIFEST =
+    [". #{random_block(9)} 0:9:conflict",
+     "./conflict #{random_block} 0:0:name\n"].join("\n")
 end
diff --git a/sdk/ruby/test/test_collection.rb b/sdk/ruby/test/test_collection.rb
new file mode 100644
index 0000000..5fba278
--- /dev/null
+++ b/sdk/ruby/test/test_collection.rb
@@ -0,0 +1,508 @@
+require "arvados/collection"
+require "minitest/autorun"
+require "sdk_fixtures"
+
+class CollectionTest < Minitest::Test
+  include SDKFixtures
+
+  TWO_BY_TWO_BLOCKS = SDKFixtures.random_blocks(2, 9)
+  TWO_BY_TWO_MANIFEST_A =
+    [". #{TWO_BY_TWO_BLOCKS.first} 0:5:f1 5:4:f2\n",
+     "./s1 #{TWO_BY_TWO_BLOCKS.last} 0:5:f1 5:4:f3\n"]
+  TWO_BY_TWO_MANIFEST_S = TWO_BY_TWO_MANIFEST_A.join("")
+
+  ### .new
+
+  def test_empty_construction
+    coll = Arv::Collection.new
+    assert_equal("", coll.manifest_text)
+  end
+
+  def test_successful_construction
+    [:SIMPLEST_MANIFEST, :MULTIBLOCK_FILE_MANIFEST, :MULTILEVEL_MANIFEST].
+        each do |manifest_name|
+      manifest_text = SDKFixtures.const_get(manifest_name)
+      coll = Arv::Collection.new(manifest_text)
+      assert_equal(manifest_text, coll.manifest_text,
+                   "did not get same manifest back out from #{manifest_name}")
+    end
+  end
+
+  def test_non_manifest_construction_error
+    ["word", ". abc def", ". #{random_block} 0:", ". / !"].each do |m_text|
+      assert_raises(ArgumentError,
+                    "built collection from manifest #{m_text.inspect}") do
+        Arv::Collection.new(m_text)
+      end
+    end
+  end
+
+  def test_file_directory_conflict_construction_error
+    assert_raises(ArgumentError) do
+      Arv::Collection.new(NAME_CONFLICT_MANIFEST)
+    end
+  end
+
+  def test_no_implicit_normalization
+    coll = Arv::Collection.new(NONNORMALIZED_MANIFEST)
+    assert_equal(NONNORMALIZED_MANIFEST, coll.manifest_text)
+  end
+
+  def test_no_implicit_normalization_from_first_import
+    coll = Arv::Collection.new
+    coll.import_manifest!(NONNORMALIZED_MANIFEST)
+    assert_equal(NONNORMALIZED_MANIFEST, coll.manifest_text)
+  end
+
+  ### .import_manifest!
+
+  def test_non_posix_path_handling
+    block = random_block(9)
+    coll = Arv::Collection.new("./.. #{block} 0:5:.\n")
+    coll.import_manifest!("./.. #{block} 5:4:..\n")
+    assert_equal("./.. #{block} 0:5:. 5:4:..\n", coll.manifest_text)
+  end
+
+  def test_escaping_through_normalization
+    coll = Arv::Collection.new(MANY_ESCAPES_MANIFEST)
+    coll.import_manifest!(MANY_ESCAPES_MANIFEST)
+    # The result should simply duplicate the file spec.
+    # The source file spec has an unescaped backslash in it.
+    # It's OK for the Collection class to properly escape that.
+    expect_text = MANY_ESCAPES_MANIFEST.sub(/ \d+:\d+:\S+/) do |file_spec|
+      file_spec.gsub(/([^\\])(\\[^\\\d])/, '\1\\\\\2') * 2
+    end
+    assert_equal(expect_text, coll.manifest_text)
+  end
+
+  def test_concatenation_from_multiple_imports(file_name="file.txt",
+                                               out_name=nil)
+    out_name ||= file_name
+    blocks = random_blocks(2, 9)
+    coll = Arv::Collection.new
+    blocks.each do |block|
+      coll.import_manifest!(". #{block} 1:8:#{file_name}\n")
+    end
+    assert_equal(". #{blocks.join(' ')} 1:8:#{out_name} 10:8:#{out_name}\n",
+                 coll.manifest_text)
+  end
+
+  def test_concatenation_from_multiple_escaped_imports
+    test_concatenation_from_multiple_imports('a\040\141.txt', 'a\040a.txt')
+  end
+
+  def test_concatenation_with_locator_overlap(over_index=0)
+    blocks = random_blocks(4, 2)
+    coll = Arv::Collection.new(". #{blocks.join(' ')} 0:8:file\n")
+    coll.import_manifest!(". #{blocks[over_index, 2].join(' ')} 0:4:file\n")
+    assert_equal(". #{blocks.join(' ')} 0:8:file #{over_index * 2}:4:file\n",
+                 coll.manifest_text)
+  end
+
+  def test_concatenation_with_middle_locator_overlap
+    test_concatenation_with_locator_overlap(1)
+  end
+
+  def test_concatenation_with_end_locator_overlap
+    test_concatenation_with_locator_overlap(2)
+  end
+
+  def test_concatenation_with_partial_locator_overlap
+    blocks = random_blocks(3, 3)
+    coll = Arv::Collection.new(". #{blocks[0, 2].join(' ')} 0:6:overlap\n")
+    coll.import_manifest!(". #{blocks[1, 2].join(' ')} 0:6:overlap\n")
+    assert_equal(". #{blocks.join(' ')} 0:6:overlap 3:6:overlap\n",
+                 coll.manifest_text)
+  end
+
+  ### .normalize!
+
+  def test_normalize
+    block = random_block
+    coll = Arv::Collection.new(". #{block} 0:0:f2 0:0:f1\n")
+    coll.normalize!
+    assert_equal(". #{block} 0:0:f1 0:0:f2\n", coll.manifest_text)
+  end
+
+  ### .copy!
+
+  def test_simple_file_copy
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    coll.copy!("./simple.txt", "./new")
+    assert_equal(SIMPLEST_MANIFEST.sub(" 0:9:", " 0:9:new 0:9:"),
+                 coll.manifest_text)
+  end
+
+  def test_copy_file_into_other_stream(target="./s1/f2", basename="f2")
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S)
+    coll.copy!("./f2", target)
+    expected = "%s./s1 %s 0:5:f1 14:4:%s 5:4:f3\n" %
+      [TWO_BY_TWO_MANIFEST_A.first,
+       TWO_BY_TWO_BLOCKS.reverse.join(" "), basename]
+    assert_equal(expected, coll.manifest_text)
+  end
+
+  def test_implicit_copy_file_into_other_stream
+    test_copy_file_into_other_stream("./s1")
+  end
+
+  def test_copy_file_into_other_stream_with_new_name
+    test_copy_file_into_other_stream("./s1/f2a", "f2a")
+  end
+
+  def test_copy_file_over_in_other_stream(target="./s1/f1")
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S)
+    coll.copy!("./f1", target)
+    expected = "%s./s1 %s 0:5:f1 14:4:f3\n" %
+      [TWO_BY_TWO_MANIFEST_A.first, TWO_BY_TWO_BLOCKS.join(" ")]
+    assert_equal(expected, coll.manifest_text)
+  end
+
+  def test_implicit_copy_file_over_in_other_stream
+    test_copy_file_over_in_other_stream("./s1")
+  end
+
+  def test_simple_stream_copy
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S)
+    coll.copy!("./s1", "./sNew")
+    new_line = TWO_BY_TWO_MANIFEST_A.last.sub("./s1 ", "./sNew ")
+    assert_equal(TWO_BY_TWO_MANIFEST_S + new_line, coll.manifest_text)
+  end
+
+  def test_copy_stream_into_other_stream(target="./dir2/subdir",
+                                         basename="subdir")
+    coll = Arv::Collection.new(MULTILEVEL_MANIFEST)
+    coll.copy!("./dir1/subdir", target)
+    new_line = MULTILEVEL_MANIFEST.lines[4].sub("./dir1/subdir ",
+                                                "./dir2/#{basename} ")
+    assert_equal(MULTILEVEL_MANIFEST + new_line, coll.manifest_text)
+  end
+
+  def test_implicit_copy_stream_into_other_stream
+    test_copy_stream_into_other_stream("./dir2")
+  end
+
+  def test_copy_stream_into_other_stream_with_new_name
+    test_copy_stream_into_other_stream("./dir2/newsub", "newsub")
+  end
+
+  def test_copy_stream_over_empty_stream
+    coll = Arv::Collection.new(MULTILEVEL_MANIFEST)
+    (1..3).each do |file_num|
+      coll.remove!("./dir0/subdir/file#{file_num}")
+    end
+    coll.copy!("./dir1/subdir", "./dir0")
+    expected = MULTILEVEL_MANIFEST.lines
+    expected[2] = expected[4].sub("./dir1/", "./dir0/")
+    assert_equal(expected.join(""), coll.manifest_text)
+  end
+
+  def test_copy_stream_over_file_raises_ENOTDIR
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S)
+    assert_raises(Errno::ENOTDIR) do
+      coll.copy!("./s1", "./f2")
+    end
+  end
+
+  def test_copy_stream_over_nonempty_stream_merges_and_overwrites
+    blocks = random_blocks(3, 9)
+    manifest_a =
+      ["./subdir #{blocks[0]} 0:1:s1 1:2:zero\n",
+       "./zdir #{blocks[1]} 0:9:zfile\n",
+       "./zdir/subdir #{blocks[2]} 0:1:s2 1:2:zero\n"]
+    coll = Arv::Collection.new(manifest_a.join(""))
+    coll.copy!("./subdir", "./zdir")
+    manifest_a[2] = "./zdir/subdir %s %s 0:1:s1 9:1:s2 1:2:zero\n" %
+      [blocks[0], blocks[2]]
+    assert_equal(manifest_a.join(""), coll.manifest_text)
+  end
+
+  def test_copy_stream_into_substream(source="./dir1",
+                                      target="./dir1/subdir/dir1")
+    coll = Arv::Collection.new(MULTILEVEL_MANIFEST)
+    coll.copy!(source, target)
+    expected = MULTILEVEL_MANIFEST.lines.flat_map do |line|
+      [line, line.gsub(/^#{Regexp.escape(source)}([\/ ])/, "#{target}\\1")].uniq
+    end
+    assert_equal(expected.sort.join(""), coll.manifest_text)
+  end
+
+  def test_copy_root
+    test_copy_stream_into_substream(".", "./root")
+  end
+
+  def test_adding_to_root_after_copy
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    coll.copy!(".", "./root")
+    coll.import_manifest!(COLON_FILENAME_MANIFEST)
+    got_lines = coll.manifest_text.lines
+    assert_equal(2, got_lines.size)
+    assert_match(/^\. \S{33,} \S{33,} 0:9:file:test\.txt 9:9:simple\.txt\n/,
+                 got_lines.first)
+    assert_equal(SIMPLEST_MANIFEST.sub(". ", "./root "), got_lines.last)
+  end
+
+  def test_copy_chaining
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    coll.copy!("./simple.txt", "./a").copy!("./a", "./b")
+    assert_equal(SIMPLEST_MANIFEST.sub(" 0:9:", " 0:9:a 0:9:b 0:9:"),
+                 coll.manifest_text)
+  end
+
+  def prep_two_collections_for_copy(src_stream, dst_stream)
+    blocks = random_blocks(2, 8)
+    src_text = "#{src_stream} #{blocks.first} 0:8:f1\n"
+    dst_text = "#{dst_stream} #{blocks.last} 0:8:f2\n"
+    return [blocks, src_text, dst_text,
+            Arv::Collection.new(src_text.dup),
+            Arv::Collection.new(dst_text.dup)]
+  end
+
+  def test_copy_file_from_other_collection(src_stream=".", dst_stream="./s1")
+    blocks, src_text, dst_text, src_coll, dst_coll =
+      prep_two_collections_for_copy(src_stream, dst_stream)
+    dst_coll.copy!("#{src_stream}/f1", dst_stream, src_coll)
+    assert_equal("#{dst_stream} #{blocks.join(' ')} 0:8:f1 8:8:f2\n",
+                 dst_coll.manifest_text)
+    assert_equal(src_text, src_coll.manifest_text)
+  end
+
+  def test_copy_file_from_other_collection_to_root
+    test_copy_file_from_other_collection("./s1", ".")
+  end
+
+  def test_copy_stream_from_other_collection
+    blocks, src_text, dst_text, src_coll, dst_coll =
+      prep_two_collections_for_copy("./s2", "./s1")
+    dst_coll.copy!("./s2", "./s1", src_coll)
+    assert_equal(dst_text + src_text.sub("./s2 ", "./s1/s2 "),
+                 dst_coll.manifest_text)
+    assert_equal(src_text, src_coll.manifest_text)
+  end
+
+  def test_copy_stream_from_other_collection_to_root
+    blocks, src_text, dst_text, src_coll, dst_coll =
+      prep_two_collections_for_copy("./s1", ".")
+    dst_coll.copy!("./s1", ".", src_coll)
+    assert_equal(dst_text + src_text, dst_coll.manifest_text)
+    assert_equal(src_text, src_coll.manifest_text)
+  end
+
+  def test_copy_empty_source_path_raises_ArgumentError(src="", dst="./s1")
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    assert_raises(ArgumentError) do
+      coll.copy!(src, dst)
+    end
+  end
+
+  def test_copy_empty_destination_path_raises_ArgumentError
+    test_copy_empty_source_path_raises_ArgumentError(".", "")
+  end
+
+  ### .rename!
+
+  def test_simple_file_rename
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    coll.rename!("./simple.txt", "./new")
+    assert_equal(SIMPLEST_MANIFEST.sub(":simple.txt", ":new"),
+                 coll.manifest_text)
+  end
+
+  def test_rename_file_into_other_stream(target="./s1/f2", basename="f2")
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S)
+    coll.rename!("./f2", target)
+    expected = ". %s 0:5:f1\n./s1 %s 0:5:f1 14:4:%s 5:4:f3\n" %
+      [TWO_BY_TWO_BLOCKS.first,
+       TWO_BY_TWO_BLOCKS.reverse.join(" "), basename]
+    assert_equal(expected, coll.manifest_text)
+  end
+
+  def test_implicit_rename_file_into_other_stream
+    test_rename_file_into_other_stream("./s1")
+  end
+
+  def test_rename_file_into_other_stream_with_new_name
+    test_rename_file_into_other_stream("./s1/f2a", "f2a")
+  end
+
+  def test_rename_file_over_in_other_stream(target="./s1/f1")
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S)
+    coll.rename!("./f1", target)
+    expected = ". %s 5:4:f2\n./s1 %s 0:5:f1 14:4:f3\n" %
+      [TWO_BY_TWO_BLOCKS.first, TWO_BY_TWO_BLOCKS.join(" ")]
+    assert_equal(expected, coll.manifest_text)
+  end
+
+  def test_implicit_rename_file_over_in_other_stream
+    test_rename_file_over_in_other_stream("./s1")
+  end
+
+  def test_simple_stream_rename
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S)
+    coll.rename!("./s1", "./newS")
+    assert_equal(TWO_BY_TWO_MANIFEST_S.sub("\n./s1 ", "\n./newS "),
+                 coll.manifest_text)
+  end
+
+  def test_rename_stream_into_other_stream(target="./dir2/subdir",
+                                           basename="subdir")
+    coll = Arv::Collection.new(MULTILEVEL_MANIFEST)
+    coll.rename!("./dir1/subdir", target)
+    expected = MULTILEVEL_MANIFEST.lines
+    replaced_line = expected.delete_at(4)
+    expected << replaced_line.sub("./dir1/subdir ", "./dir2/#{basename} ")
+    assert_equal(expected.join(""), coll.manifest_text)
+  end
+
+  def test_implicit_rename_stream_into_other_stream
+    test_rename_stream_into_other_stream("./dir2")
+  end
+
+  def test_rename_stream_into_other_stream_with_new_name
+    test_rename_stream_into_other_stream("./dir2/newsub", "newsub")
+  end
+
+  def test_rename_stream_over_empty_stream
+    coll = Arv::Collection.new(MULTILEVEL_MANIFEST)
+    (1..3).each do |file_num|
+      coll.remove!("./dir0/subdir/file#{file_num}")
+    end
+    coll.rename!("./dir1/subdir", "./dir0")
+    expected = MULTILEVEL_MANIFEST.lines
+    expected[2] = expected.delete_at(4).sub("./dir1/", "./dir0/")
+    assert_equal(expected.sort.join(""), coll.manifest_text)
+  end
+
+  def test_rename_stream_over_file_raises_ENOTDIR
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S)
+    assert_raises(Errno::ENOTDIR) do
+      coll.rename!("./s1", "./f2")
+    end
+  end
+
+  def test_rename_stream_over_nonempty_stream_raises_ENOTEMPTY
+    coll = Arv::Collection.new(MULTILEVEL_MANIFEST)
+    assert_raises(Errno::ENOTEMPTY) do
+      coll.rename!("./dir1/subdir", "./dir0")
+    end
+  end
+
+  def test_rename_stream_into_substream(source="./dir1",
+                                        target="./dir1/subdir/dir1")
+    coll = Arv::Collection.new(MULTILEVEL_MANIFEST)
+    coll.rename!(source, target)
+    assert_equal(MULTILEVEL_MANIFEST.gsub(/^#{Regexp.escape(source)}([\/ ])/m,
+                                          "#{target}\\1"),
+                 coll.manifest_text)
+  end
+
+  def test_rename_root
+    test_rename_stream_into_substream(".", "./root")
+  end
+
+  def test_adding_to_root_after_rename
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    coll.rename!(".", "./root")
+    coll.import_manifest!(SIMPLEST_MANIFEST)
+    assert_equal(SIMPLEST_MANIFEST + SIMPLEST_MANIFEST.sub(". ", "./root "),
+                 coll.manifest_text)
+  end
+
+  def test_rename_chaining
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    coll.rename!("./simple.txt", "./x").rename!("./x", "./simple.txt")
+    assert_equal(SIMPLEST_MANIFEST, coll.manifest_text)
+  end
+
+  ### .remove!
+
+  def test_simple_remove
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S.dup)
+    coll.remove!("./f2")
+    assert_equal(TWO_BY_TWO_MANIFEST_S.sub(" 5:4:f2", ""), coll.manifest_text)
+  end
+
+  def empty_stream_and_assert(expect_index=0)
+    coll = Arv::Collection.new(TWO_BY_TWO_MANIFEST_S)
+    yield coll
+    assert_equal(TWO_BY_TWO_MANIFEST_A[expect_index], coll.manifest_text)
+  end
+
+  def test_remove_all_files_in_substream
+    empty_stream_and_assert do |coll|
+      coll.remove!("./s1/f1")
+      coll.remove!("./s1/f3")
+    end
+  end
+
+  def test_remove_all_files_in_root_stream
+    empty_stream_and_assert(1) do |coll|
+      coll.remove!("./f1")
+      coll.remove!("./f2")
+    end
+  end
+
+  def test_remove_empty_stream
+    empty_stream_and_assert do |coll|
+      coll.remove!("./s1/f1")
+      coll.remove!("./s1/f3")
+      coll.remove!("./s1")
+    end
+  end
+
+  def test_recursive_remove
+    empty_stream_and_assert do |coll|
+      coll.remove!("./s1", recursive: true)
+    end
+  end
+
+  def test_recursive_remove_on_files
+    empty_stream_and_assert do |coll|
+      coll.remove!("./s1/f1", recursive: true)
+      coll.remove!("./s1/f3", recursive: true)
+    end
+  end
+
+  def test_chaining_removes
+    empty_stream_and_assert do |coll|
+      coll.remove!("./s1/f1").remove!("./s1/f3")
+    end
+  end
+
+  def test_remove_last_file
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    coll.remove!("./simple.txt")
+    assert_equal("", coll.manifest_text)
+  end
+
+  def test_remove_root_stream
+    coll = Arv::Collection.new(MULTILEVEL_MANIFEST)
+    coll.remove!(".", recursive: true)
+    assert_equal("", coll.manifest_text)
+  end
+
+  def test_remove_nonexistent_file_raises_ENOENT(path="./NoSuchFile")
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    assert_raises(Errno::ENOENT) do
+      coll.remove!(path)
+    end
+  end
+
+  def test_remove_from_nonexistent_stream_raises_ENOENT
+    test_remove_nonexistent_file_raises_ENOENT("./NoSuchStream/simple.txt")
+  end
+
+  def test_remove_nonempty_stream_raises_ENOTEMPTY
+    coll = Arv::Collection.new(MULTILEVEL_MANIFEST)
+    assert_raises(Errno::ENOTEMPTY) do
+      coll.remove!("./dir1/subdir")
+    end
+  end
+
+  def test_remove_empty_string_raises_ArgumentError
+    coll = Arv::Collection.new(SIMPLEST_MANIFEST)
+    assert_raises(ArgumentError) do
+      coll.remove!("")
+    end
+  end
+end

commit a9fdae7d5d78f2d7b89d338496f931288b3632ca
Author: Brett Smith <brett at curoverse.com>
Date:   Thu Mar 5 12:44:02 2015 -0500

    5104: Refactor Ruby SDK test infrastructure into sharable module.

diff --git a/sdk/ruby/test/sdk_fixtures.rb b/sdk/ruby/test/sdk_fixtures.rb
new file mode 100644
index 0000000..8af81cd
--- /dev/null
+++ b/sdk/ruby/test/sdk_fixtures.rb
@@ -0,0 +1,53 @@
+require "yaml"
+
+module SDKFixtures
+  module StaticMethods
+    # SDKFixtures will use these as class methods, and install them as
+    # instance methods on the test classes.
+    def random_block(size=nil)
+      sprintf("%032x+%d", rand(16 ** 32), size || rand(64 * 1024 * 1024))
+    end
+  end
+
+  extend StaticMethods
+
+  def self.included(base)
+    base.include(StaticMethods)
+  end
+
+  @@fixtures = {}
+  def fixtures name
+    @@fixtures[name] ||=
+      begin
+        path = File.
+          expand_path("../../../../services/api/test/fixtures/#{name}.yml",
+                      __FILE__)
+        file = IO.read(path)
+        trim_index = file.index('# Test Helper trims the rest of the file')
+        file = file[0, trim_index] if trim_index
+        YAML.load(file)
+      end
+  end
+
+  ### Valid manifests
+  SIMPLEST_MANIFEST = ". #{random_block(9)} 0:9:simple.txt\n"
+  MULTIBLOCK_FILE_MANIFEST =
+    [". #{random_block(8)} 0:4:repfile 4:4:uniqfile",
+     "./s1 #{random_block(6)} 0:3:repfile 3:3:uniqfile",
+     ". #{random_block(8)} 0:7:uniqfile2 7:1:repfile\n"].join("\n")
+  MULTILEVEL_MANIFEST =
+    [". #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n",
+     "./dir0 #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n",
+     "./dir0/subdir #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n",
+     "./dir1 #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n",
+     "./dir1/subdir #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n",
+     "./dir2 #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n"].join("")
+  COLON_FILENAME_MANIFEST = ". #{random_block(9)} 0:9:file:test.txt\n"
+  # Filename is `a a.txt`.
+  ESCAPED_FILENAME_MANIFEST = ". #{random_block(9)} 0:9:a\\040\\141.txt\n"
+  MANY_ESCAPES_MANIFEST =
+    "./dir\\040name #{random_block(9)} 0:9:file\\\\name\\011\\here.txt\n"
+  NONNORMALIZED_MANIFEST =
+    ["./dir2 #{random_block} 0:0:z 0:0:y 0:0:x",
+     "./dir1 #{random_block} 0:0:p 0:0:o 0:0:n\n"].join("\n")
+end
diff --git a/sdk/ruby/test/test_keep_manifest.rb b/sdk/ruby/test/test_keep_manifest.rb
index 8ad8134..2ad0941 100644
--- a/sdk/ruby/test/test_keep_manifest.rb
+++ b/sdk/ruby/test/test_keep_manifest.rb
@@ -1,22 +1,9 @@
-require "minitest/autorun"
 require "arvados/keep"
-require "yaml"
-
-def random_block(size=nil)
-  sprintf("%032x+%d", rand(16 ** 32), size || rand(64 * 1024 * 1024))
-end
+require "minitest/autorun"
+require "sdk_fixtures"
 
 class ManifestTest < Minitest::Test
-  SIMPLEST_MANIFEST = ". #{random_block(9)} 0:9:simple.txt\n"
-  MULTIBLOCK_FILE_MANIFEST =
-    [". #{random_block(8)} 0:4:repfile 4:4:uniqfile",
-     "./s1 #{random_block(6)} 0:3:repfile 3:3:uniqfile",
-     ". #{random_block(8)} 0:7:uniqfile2 7:1:repfile\n"].join("\n")
-  MULTILEVEL_MANIFEST =
-    [". #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n",
-     "./dir1 #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n",
-     "./dir1/subdir #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n",
-     "./dir2 #{random_block(9)} 0:3:file1 3:3:file2 6:3:file3\n"].join("")
+  include SDKFixtures
 
   def check_stream(stream, exp_name, exp_blocks, exp_files)
     assert_equal(exp_name, stream.first)
@@ -51,7 +38,8 @@ class ManifestTest < Minitest::Test
       seen << stream
       assert_equal(3, files.size, "wrong file count for stream #{stream}")
     end
-    assert_equal(4, seen.size, "wrong number of streams")
+    assert_equal(MULTILEVEL_MANIFEST.count("\n"), seen.size,
+                 "wrong number of streams")
   end
 
   def test_empty_each_line
@@ -87,13 +75,12 @@ class ManifestTest < Minitest::Test
   end
 
   def test_backslash_escape_parsing
-    m_text = "./dir\\040name #{random_block} 0:0:file\\\\name\\011\\here.txt\n"
-    manifest = Keep::Manifest.new(m_text)
+    manifest = Keep::Manifest.new(MANY_ESCAPES_MANIFEST)
     streams = manifest.each_line.to_a
     assert_equal(1, streams.size, "wrong number of streams with whitespace")
     assert_equal("./dir name", streams.first.first,
                  "wrong stream name with whitespace")
-    assert_equal(["0:0:file\\name\t\\here.txt"], streams.first.last,
+    assert_equal(["0:9:file\\name\t\\here.txt"], streams.first.last,
                  "wrong filename(s) with whitespace")
   end
 
@@ -118,12 +105,12 @@ class ManifestTest < Minitest::Test
   end
 
   def test_files_with_colons_in_names
-    manifest = Keep::Manifest.new(". #{random_block(9)} 0:9:file:test.txt\n")
+    manifest = Keep::Manifest.new(COLON_FILENAME_MANIFEST)
     assert_equal([[".", "file:test.txt", 9]], manifest.files)
   end
 
   def test_files_with_escape_sequence_in_filename
-    manifest = Keep::Manifest.new(". #{random_block(9)} 0:9:a\\040\\141.txt\n")
+    manifest = Keep::Manifest.new(ESCAPED_FILENAME_MANIFEST)
     assert_equal([[".", "a a.txt", 9]], manifest.files)
   end
 
@@ -177,11 +164,11 @@ class ManifestTest < Minitest::Test
   end
 
   def test_has_file_with_spaces
-    manifest = Keep::Manifest.new(". #{random_block(3)} 0:3:a\\040b\\040c\n")
-    assert(manifest.has_file?("./a b c"), "one-arg 'a b c' not found")
-    assert(manifest.has_file?(".", "a b c"), "two-arg 'a b c' not found")
-    refute(manifest.has_file?("a\\040b\\040c"), "one-arg unescaped found")
-    refute(manifest.has_file?(".", "a\\040b\\040c"), "two-arg unescaped found")
+    manifest = Keep::Manifest.new(ESCAPED_FILENAME_MANIFEST)
+    assert(manifest.has_file?("./a a.txt"), "one-arg path not found")
+    assert(manifest.has_file?(".", "a a.txt"), "two-arg path not found")
+    refute(manifest.has_file?("a\\040\\141"), "one-arg unescaped found")
+    refute(manifest.has_file?(".", "a\\040\\141"), "two-arg unescaped found")
   end
 
   def test_parse_all_fixtures
@@ -208,15 +195,4 @@ class ManifestTest < Minitest::Test
       assert !file_name.empty?, "empty file_name in #{name} fixture"
     end
   end
-
-  @@fixtures = nil
-  def fixtures name
-    return @@fixtures if @@fixtures
-    path = File.expand_path("../../../../services/api/test/fixtures/#{name}.yml",
-                            __FILE__)
-    file = IO.read(path)
-    trim_index = file.index('# Test Helper trims the rest of the file')
-    file = file[0, trim_index] if trim_index
-    @@fixtures = YAML.load(file)
-  end
 end

-----------------------------------------------------------------------


hooks/post-receive
-- 




More information about the arvados-commits mailing list