X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/1bcfe8651af341c6e7cd01a19443c7c288efa932..09cbdc3074b3f1e69c9c537875146f6da0a6ed8f:/sdk/ruby/lib/arvados/keep.rb

diff --git a/sdk/ruby/lib/arvados/keep.rb b/sdk/ruby/lib/arvados/keep.rb
index ede40c3fc2..e391b7a6ca 100644
--- a/sdk/ruby/lib/arvados/keep.rb
+++ b/sdk/ruby/lib/arvados/keep.rb
@@ -1,3 +1,7 @@
+# Copyright (C) The Arvados Authors. All rights reserved.
+#
+# SPDX-License-Identifier: Apache-2.0
+
 module Keep
   class Locator
     # A Locator is used to parse and manipulate Keep locator strings.
@@ -18,7 +22,7 @@ module Keep
     #   sign-timestamp ::= <8 lowercase hex digits>
     attr_reader :hash, :hints, :size
 
-    LOCATOR_REGEXP = /^([[:xdigit:]]{32})(\+([[:digit:]]+))?(\+([[:upper:]][[:alnum:]+@_-]*))?$/
+    LOCATOR_REGEXP = /^([[:xdigit:]]{32})(\+([[:digit:]]+))?((\+([[:upper:]][[:alnum:]@_-]*))+)?\z/
 
     def initialize(hasharg, sizearg, hintarg)
       @hash = hasharg
@@ -35,7 +39,7 @@ module Keep
     def self.parse(tok)
       begin
         Locator.parse!(tok)
-      rescue ArgumentError => e
+      rescue ArgumentError
         nil
       end
     end
@@ -47,19 +51,19 @@ module Keep
         raise ArgumentError.new "locator is nil or empty"
       end
 
-      m = LOCATOR_REGEXP.match(tok.strip)
+      m = LOCATOR_REGEXP.match(tok)
       unless m
         raise ArgumentError.new "not a valid locator #{tok}"
       end
 
-      tokhash, _, toksize, _, trailer = m[1..5]
+      tokhash, _, toksize, _, _, trailer = m[1..6]
       tokhints = []
       if trailer
         trailer.split('+').each do |hint|
-          if hint =~ /^[[:upper:]][[:alnum:]@_-]+$/
+          if hint =~ /^[[:upper:]][[:alnum:]@_-]*$/
             tokhints.push(hint)
           else
-            raise ArgumentError.new "unknown hint #{hint}"
+            raise ArgumentError.new "invalid hint #{hint}"
           end
         end
       end
@@ -97,6 +101,15 @@ module Keep
   end
 
   class Manifest
+    STREAM_TOKEN_REGEXP = /^([^\000-\040\\]|\\[0-3][0-7][0-7])+$/
+    STREAM_NAME_REGEXP = /^(\.)(\/[^\/]+)*$/
+
+    EMPTY_DIR_TOKEN_REGEXP = /^0:0:\.$/ # The exception when a file can have '.' as a name
+    FILE_TOKEN_REGEXP = /^[[:digit:]]+:[[:digit:]]+:([^\000-\040\\]|\\[0-3][0-7][0-7])+$/
+    FILE_NAME_REGEXP = /^[[:digit:]]+:[[:digit:]]+:([^\/]+(\/[^\/]+)*)$/
+
+    NON_8BIT_ENCODED_CHAR = /[^\\]\\[4-7][0-7][0-7]/
+
     # Class to parse a manifest text and provide common views of that data.
     def initialize(manifest_text)
       @text = manifest_text
@@ -109,7 +122,7 @@ module Keep
         stream_name = nil
         block_tokens = []
         file_tokens = []
-        line.scan /\S+/ do |token|
+        line.scan(/\S+/) do |token|
           if stream_name.nil?
             stream_name = unescape token
           elsif file_tokens.empty? and Locator.valid? token
@@ -124,7 +137,9 @@ module Keep
       end
     end
 
-    def unescape(s)
+    def self.unescape(s)
+      return nil if s.nil?
+
       # Parse backslash escapes in a Keep manifest stream or file name.
       s.gsub(/\\(\\|[0-7]{3})/) do |_|
         case $1
@@ -136,6 +151,10 @@ module Keep
       end
     end
 
+    def unescape(s)
+      self.class.unescape(s)
+    end
+
     def split_file_token token
       start_pos, filesize, filename = token.split(':', 3)
       if filename.nil?
@@ -149,12 +168,21 @@ module Keep
       @text.each_line do |line|
         stream_name = nil
         in_file_tokens = false
-        line.scan /\S+/ do |token|
+        line.scan(/\S+/) do |token|
           if stream_name.nil?
             stream_name = unescape token
           elsif in_file_tokens or not Locator.valid? token
             in_file_tokens = true
-            yield [stream_name] + split_file_token(token)
+
+            start_pos, file_size, file_name = split_file_token(token)
+            stream_name_adjuster = ''
+            if file_name.include?('/')                # '/' in filename
+              dirname, sep, basename = file_name.rpartition('/')
+              stream_name_adjuster = sep + dirname   # /dir_parts
+              file_name = basename
+            end
+
+            yield [stream_name + stream_name_adjuster, start_pos, file_size, file_name]
           end
         end
       end
@@ -181,16 +209,24 @@ module Keep
       # files.  This can help you avoid parsing the entire manifest if you
       # just want to check if a small number of files are specified.
       if stop_after.nil? or not @files.nil?
-        return files.size
+        # Avoid counting empty dir placeholders
+        return files.reject{|_, name, size| name == '.' and size == 0}.size
       end
       seen_files = {}
-      each_file_spec do |streamname, _, _, filename|
+      each_file_spec do |streamname, _, filesize, filename|
+        # Avoid counting empty dir placeholders
+        next if filename == "." and filesize == 0
         seen_files[[streamname, filename]] = true
         return stop_after if (seen_files.size >= stop_after)
       end
       seen_files.size
     end
 
+    def files_size
+      # Return the total size of all files in this manifest.
+      files.reduce(0) { |total, (_, _, size)| total + size }
+    end
+
     def exact_file_count?(want_count)
       files_count(want_count + 1) == want_count
     end
@@ -210,5 +246,67 @@ module Keep
       end
       false
     end
+
+    # Verify that a given manifest is valid according to
+    # https://dev.arvados.org/projects/arvados/wiki/Keep_manifest_format
+    def self.validate! manifest
+      raise ArgumentError.new "No manifest found" if !manifest
+
+      return true if manifest.empty?
+
+      raise ArgumentError.new "Invalid manifest: does not end with newline" if !manifest.end_with?("\n")
+      line_count = 0
+      manifest.each_line do |line|
+        line_count += 1
+
+        words = line[0..-2].split(/ /)
+        raise ArgumentError.new "Manifest invalid for stream #{line_count}: missing stream name" if words.empty?
+
+        count = 0
+
+        word = words.shift
+        raise ArgumentError.new "Manifest invalid for stream #{line_count}: >8-bit encoded chars not allowed on stream token #{word.inspect}" if word =~ NON_8BIT_ENCODED_CHAR
+        unescaped_word = unescape(word)
+        count += 1 if word =~ STREAM_TOKEN_REGEXP and unescaped_word =~ STREAM_NAME_REGEXP and unescaped_word !~ /\/\.\.?(\/|$)/
+        raise ArgumentError.new "Manifest invalid for stream #{line_count}: missing or invalid stream name #{word.inspect if word}" if count != 1
+
+        count = 0
+        word = words.shift
+        while word =~ Locator::LOCATOR_REGEXP
+          word = words.shift
+          count += 1
+        end
+        raise ArgumentError.new "Manifest invalid for stream #{line_count}: missing or invalid locator #{word.inspect if word}" if count == 0
+
+        count = 0
+        raise ArgumentError.new "Manifest invalid for stream #{line_count}: >8-bit encoded chars not allowed on file token #{word.inspect}" if word =~ NON_8BIT_ENCODED_CHAR
+        while unescape(word) =~ EMPTY_DIR_TOKEN_REGEXP or
+          (word =~ FILE_TOKEN_REGEXP and unescape(word) =~ FILE_NAME_REGEXP and ($~[1].split('/') & ['..', '.']).empty?)
+          word = words.shift
+          count += 1
+        end
+
+        if word
+          raise ArgumentError.new "Manifest invalid for stream #{line_count}: invalid file token #{word.inspect}"
+        elsif count == 0
+          raise ArgumentError.new "Manifest invalid for stream #{line_count}: no file tokens"
+        end
+
+        # Ruby's split() method silently drops trailing empty tokens
+        # (which are not allowed by the manifest format) so we have to
+        # check trailing spaces manually.
+        raise ArgumentError.new "Manifest invalid for stream #{line_count}: trailing space" if line.end_with? " \n"
+      end
+      true
+    end
+
+    def self.valid? manifest
+      begin
+        validate! manifest
+        true
+      rescue ArgumentError
+        false
+      end
+    end
   end
 end