Class: GraphQL::Language::Lexer

Inherits:
Object
  • Object
show all
Defined in:
lib/graphql/language/lexer.rb

Defined Under Namespace

Modules: ByteFor, Punctuation

Constant Summary collapse

ESCAPES =
/\\["\\\/bfnrt]/
ESCAPES_REPLACE =
{
  '\\"' => '"',
  "\\\\" => "\\",
  "\\/" => '/',
  "\\b" => "\b",
  "\\f" => "\f",
  "\\n" => "\n",
  "\\r" => "\r",
  "\\t" => "\t",
}
UTF_8 =
/\\u(?:([\dAa-f]{4})|\{([\da-f]{4,})\})(?:\\u([\dAa-f]{4}))?/i
VALID_STRING =
/\A(?:[^\\]|#{ESCAPES}|#{UTF_8})*\z/o
ESCAPED =
/(?:#{ESCAPES}|#{UTF_8})/o
IGNORE_REGEXP =
%r{
  (?:
    [, \c\r\n\t]+ |
    \#.*$
  )*
}x
IDENTIFIER_REGEXP =
/[_A-Za-z][_0-9A-Za-z]*/
INT_REGEXP =
/-?(?:[0]|[1-9][0-9]*)/
FLOAT_DECIMAL_REGEXP =
/[.][0-9]+/
FLOAT_EXP_REGEXP =
/[eE][+-]?[0-9]+/
NUMERIC_REGEXP =

TODO: FLOAT_EXP_REGEXP should not be allowed to follow INT_REGEXP, integers are not allowed to have exponent parts.

/#{INT_REGEXP}(#{FLOAT_DECIMAL_REGEXP}#{FLOAT_EXP_REGEXP}|#{FLOAT_DECIMAL_REGEXP}|#{FLOAT_EXP_REGEXP})?/
KEYWORDS =
[
  "on",
  "fragment",
  "true",
  "false",
  "null",
  "query",
  "mutation",
  "subscription",
  "schema",
  "scalar",
  "type",
  "extend",
  "implements",
  "interface",
  "union",
  "enum",
  "input",
  "directive",
  "repeatable"
].freeze
KEYWORD_REGEXP =
/#{Regexp.union(KEYWORDS.sort)}\b/
KEYWORD_BY_TWO_BYTES =
[
  :INTERFACE,
  :MUTATION,
  :EXTEND,
  :FALSE,
  :ENUM,
  :TRUE,
  :NULL,
  nil,
  nil,
  nil,
  nil,
  nil,
  nil,
  nil,
  :QUERY,
  nil,
  nil,
  :REPEATABLE,
  :IMPLEMENTS,
  :INPUT,
  :TYPE,
  :SCHEMA,
  nil,
  nil,
  nil,
  :DIRECTIVE,
  :UNION,
  nil,
  nil,
  :SCALAR,
  nil,
  :FRAGMENT
]
PUNCTUATION_NAME_FOR_BYTE =

A sparse array mapping the bytes for each punctuation to a symbol name for that punctuation

Punctuation.constants.each_with_object([]) { |name, arr|
  punct = Punctuation.const_get(name)
  arr[punct.ord] = name
}
QUOTE =
'"'
UNICODE_DIGIT =
/[0-9A-Za-z]/
FOUR_DIGIT_UNICODE =
/#{UNICODE_DIGIT}{4}/
N_DIGIT_UNICODE =
%r{#{Punctuation::LCURLY}#{UNICODE_DIGIT}{4,}#{Punctuation::RCURLY}}x
UNICODE_ESCAPE =
%r{\\u(?:#{FOUR_DIGIT_UNICODE}|#{N_DIGIT_UNICODE})}
STRING_ESCAPE =
%r{[\\][\\/bfnrt]}
BLOCK_QUOTE =
'"""'
ESCAPED_QUOTE =
/\\"/
STRING_CHAR =
/#{ESCAPED_QUOTE}|[^"\\\n\r]|#{UNICODE_ESCAPE}|#{STRING_ESCAPE}/
QUOTED_STRING_REGEXP =
%r{#{QUOTE} (?:#{STRING_CHAR})* #{QUOTE}}x
BLOCK_STRING_REGEXP =
%r{
  #{BLOCK_QUOTE}
  (?: [^"\\]               |  # Any characters that aren't a quote or slash
     (?<!") ["]{1,2} (?!") |  # Any quotes that don't have quotes next to them
     \\"{0,3}(?!")         |  # A slash followed by <= 3 quotes that aren't followed by a quote
     \\                    |  # A slash
     "{1,2}(?!")              # 1 or 2 " followed by something that isn't a quote
  )*
  (?:"")?
  #{BLOCK_QUOTE}
}xm
FIRST_BYTES =

Use this array to check, for a given byte that will start a token, what kind of token might it start?

Array.new(255)

Instance Attribute Summary collapse

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(graphql_str, filename: nil, max_tokens: nil) ⇒ Lexer

Returns a new instance of Lexer.



6
7
8
9
10
11
12
13
14
15
16
# File 'lib/graphql/language/lexer.rb', line 6

def initialize(graphql_str, filename: nil, max_tokens: nil)
  if !(graphql_str.encoding == Encoding::UTF_8 || graphql_str.ascii_only?)
    graphql_str = graphql_str.dup.force_encoding(Encoding::UTF_8)
  end
  @string = graphql_str
  @filename = filename
  @scanner = StringScanner.new(graphql_str)
  @pos = nil
  @max_tokens = max_tokens || Float::INFINITY
  @tokens_count = 0
end

Instance Attribute Details

#posObject (readonly)

Returns the value of attribute pos.



22
23
24
# File 'lib/graphql/language/lexer.rb', line 22

def pos
  @pos
end

Class Method Details

.replace_escaped_characters_in_place(raw_string) ⇒ Object

Replace any escaped unicode or whitespace with the actual characters To avoid allocating more strings, this modifies the string passed into it



318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
# File 'lib/graphql/language/lexer.rb', line 318

def self.replace_escaped_characters_in_place(raw_string)
  raw_string.gsub!(ESCAPED) do |matched_str|
    if (point_str_1 = $1 || $2)
      codepoint_1 = point_str_1.to_i(16)
      if (codepoint_2 = $3)
        codepoint_2 = codepoint_2.to_i(16)
        if (codepoint_1 >= 0xD800 && codepoint_1 <= 0xDBFF) && # leading surrogate
            (codepoint_2 >= 0xDC00 && codepoint_2 <= 0xDFFF) # trailing surrogate
          # A surrogate pair
          combined = ((codepoint_1 - 0xD800) * 0x400) + (codepoint_2 - 0xDC00) + 0x10000
          [combined].pack('U'.freeze)
        else
          # Two separate code points
          [codepoint_1].pack('U'.freeze) + [codepoint_2].pack('U'.freeze)
        end
      else
        [codepoint_1].pack('U'.freeze)
      end
    else
      ESCAPES_REPLACE[matched_str]
    end
  end
  nil
end

.tokenize(string) ⇒ Object

This is not used during parsing because the parser doesn’t actually need tokens.



345
346
347
348
349
350
351
352
353
354
355
356
357
358
# File 'lib/graphql/language/lexer.rb', line 345

def self.tokenize(string)
  lexer = GraphQL::Language::Lexer.new(string)
  tokens = []
  while (token_name = lexer.advance)
    new_token = [
      token_name,
      lexer.line_number,
      lexer.column_number,
      lexer.debug_token_value(token_name),
    ]
    tokens << new_token
  end
  tokens
end

Instance Method Details

#_hash(key) ⇒ Object

This produces a unique integer for bytes 2 and 3 of each keyword string See https://tenderlovemaking.com/2023/09/02/fast-tokenizers-with-stringscanner.html



239
240
241
# File 'lib/graphql/language/lexer.rb', line 239

def _hash key
  (key * 18592990) >> 27 & 0x1f
end

#advanceObject



24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
# File 'lib/graphql/language/lexer.rb', line 24

def advance
  @scanner.skip(IGNORE_REGEXP)
  return false if @scanner.eos?
  @tokens_count += 1
  if @tokens_count > @max_tokens
    raise_parse_error("This query is too large to execute.")
  end
  @pos = @scanner.pos
  next_byte = @string.getbyte(@pos)
  next_byte_is_for = FIRST_BYTES[next_byte]
  case next_byte_is_for
  when ByteFor::PUNCTUATION
    @scanner.pos += 1
    PUNCTUATION_NAME_FOR_BYTE[next_byte]
  when ByteFor::NAME
    if len = @scanner.skip(KEYWORD_REGEXP)
      case len
      when 2
        :ON
      when 12
        :SUBSCRIPTION
      else
        pos = @pos

        # Use bytes 2 and 3 as a unique identifier for this keyword
        bytes = (@string.getbyte(pos + 2) << 8) | @string.getbyte(pos + 1)
        KEYWORD_BY_TWO_BYTES[_hash(bytes)]
      end
    else
      @scanner.skip(IDENTIFIER_REGEXP)
      :IDENTIFIER
    end
  when ByteFor::IDENTIFIER
    @scanner.skip(IDENTIFIER_REGEXP)
    :IDENTIFIER
  when ByteFor::NUMBER
    @scanner.skip(NUMERIC_REGEXP)

    if GraphQL.reject_numbers_followed_by_names
      new_pos = @scanner.pos
      peek_byte = @string.getbyte(new_pos)
      next_first_byte = FIRST_BYTES[peek_byte]
      if next_first_byte == ByteFor::NAME || next_first_byte == ByteFor::IDENTIFIER
        number_part = token_value
        name_part = @scanner.scan(IDENTIFIER_REGEXP)
        raise_parse_error("Name after number is not allowed (in `#{number_part}#{name_part}`)")
      end
    end
    # Check for a matched decimal:
    @scanner[1] ? :FLOAT : :INT
  when ByteFor::ELLIPSIS
    if @string.getbyte(@pos + 1) != 46 || @string.getbyte(@pos + 2) != 46
      raise_parse_error("Expected `...`, actual: #{@string[@pos..@pos + 2].inspect}")
    end
    @scanner.pos += 3
    :ELLIPSIS
  when ByteFor::STRING
    if @scanner.skip(BLOCK_STRING_REGEXP) || @scanner.skip(QUOTED_STRING_REGEXP)
      :STRING
    else
      raise_parse_error("Expected string or block string, but it was malformed")
    end
  else
    @scanner.pos += 1
    :UNKNOWN_CHAR
  end
rescue ArgumentError => err
  if err.message == "invalid byte sequence in UTF-8"
    raise_parse_error("Parse error on bad Unicode escape sequence", nil, nil)
  end
end

#column_numberObject



158
159
160
# File 'lib/graphql/language/lexer.rb', line 158

def column_number
  @scanner.string[0..@pos].split("\n").last.length
end

#debug_token_value(token_name) ⇒ Object



102
103
104
105
106
107
108
109
110
111
112
113
114
# File 'lib/graphql/language/lexer.rb', line 102

def debug_token_value(token_name)
  if token_name && Lexer::Punctuation.const_defined?(token_name)
    Lexer::Punctuation.const_get(token_name)
  elsif token_name == :ELLIPSIS
    "..."
  elsif token_name == :STRING
    string_value
  elsif @scanner.matched_size.nil?
    @scanner.peek(1)
  else
    token_value
  end
end

#eos?Boolean

Returns:

  • (Boolean)


18
19
20
# File 'lib/graphql/language/lexer.rb', line 18

def eos?
  @scanner.eos?
end

#line_numberObject



154
155
156
# File 'lib/graphql/language/lexer.rb', line 154

def line_number
  @scanner.string[0..@pos].count("\n") + 1
end

#raise_parse_error(message, line = line_number, col = column_number) ⇒ Object



162
163
164
# File 'lib/graphql/language/lexer.rb', line 162

def raise_parse_error(message, line = line_number, col = column_number)
  raise GraphQL::ParseError.new(message, line, col, @string, filename: @filename)
end

#string_valueObject



131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
# File 'lib/graphql/language/lexer.rb', line 131

def string_value
  str = token_value
  is_block = str.start_with?('"""')
  if is_block
    str.gsub!(/\A"""|"""\z/, '')
    return Language::BlockString.trim_whitespace(str)
  else
    str.gsub!(/\A"|"\z/, '')

    if !str.valid_encoding? || !str.match?(VALID_STRING)
      raise_parse_error("Bad unicode escape in #{str.inspect}")
    else
      Lexer.replace_escaped_characters_in_place(str)

      if !str.valid_encoding?
        raise_parse_error("Bad unicode escape in #{str.inspect}")
      else
        str
      end
    end
  end
end

#token_valueObject



96
97
98
99
100
# File 'lib/graphql/language/lexer.rb', line 96

def token_value
  @string.byteslice(@scanner.pos - @scanner.matched_size, @scanner.matched_size)
rescue StandardError => err
  raise GraphQL::Error, "(token_value failed: #{err.class}: #{err.message})"
end