+# 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.
def self.parse(tok)
begin
Locator.parse!(tok)
- rescue ArgumentError => e
+ rescue ArgumentError
nil
end
end
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
end
class Manifest
- STRICT_STREAM_TOKEN_REGEXP = /^(\.)(\/[^\/\s]+)*$/
- STRICT_FILE_TOKEN_REGEXP = /^[[:digit:]]+:[[:digit:]]+:([^\s\/]+(\/[^\s\/]+)*)$/
+ 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)
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
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
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?
@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
- file_tokens = split_file_token(token)
+ start_pos, file_size, file_name = split_file_token(token)
stream_name_adjuster = ''
- if file_tokens[2].include?('/') # '/' in filename
- parts = file_tokens[2].rpartition('/')
- stream_name_adjuster = parts[1] + parts[0] # /dir_parts
- file_tokens[2] = parts[2]
+ 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] + file_tokens
+ yield [stream_name + stream_name_adjuster, start_pos, file_size, file_name]
end
end
end
# 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
end
# Verify that a given manifest is valid according to
- # https://arvados.org/projects/arvados/wiki/Keep_manifest_format
+ # 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|
count = 0
word = words.shift
- count += 1 if word =~ STRICT_STREAM_TOKEN_REGEXP and word !~ /\/\.\.?(\/|$)/
+ 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 !~ /\/\.\.?(\/|$)/
raise ArgumentError.new "Manifest invalid for stream #{line_count}: missing or invalid stream name #{word.inspect if word}" if count != 1
count = 0
raise ArgumentError.new "Manifest invalid for stream #{line_count}: missing or invalid locator #{word.inspect if word}" if count == 0
count = 0
- while word =~ STRICT_FILE_TOKEN_REGEXP and ($~[1].split('/') & ['..','.']).empty?
+ 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
projects / arvados.git / blobdiff