Categorize bad backslashes, and make \" -> " because other shells do it.

core/id_kind.py

@@ -402,7 +402,8 @@ def _AddKinds(spec):
      'OneChar', 'Stop', 'Hex',
      # Two variants of Octal: \377, and \0377.
      'Octal3', 'Octal4',
      'Unicode4', 'Unicode8', 'Literals'
      'Unicode4', 'Unicode8', 'Literals', 
      'BadBackslash',  # \D or trailing \
  ])

core/word_compile.py

@@ -20,6 +20,8 @@
    't': '\t',
    'v': '\v',
    '\\': '\\',
    "'": "'",  # for $'' only, not echo -e
    '"': '"',  # not sure why this is escaped within $''
}

# TODO: Strict mode syntax errors:
@@ -34,8 +36,13 @@ def EvalCStringToken(id_, value):

  $'' could use it at compile time, much like brace expansion in braces.py.
  """
  if id_ == Id.Char_Literals:
    return value

  elif id_ == Id.Char_BadBackslash:  # TODO: error in strict mode
    return value

  if id_ == Id.Char_OneChar:
  elif id_ == Id.Char_OneChar:
    c = value[1]
    return _ONE_CHAR[c]

@@ -70,9 +77,6 @@ def EvalCStringToken(id_, value):
    i = int(s, 16)
    return unichr(i)

  elif id_ == Id.Char_Literals:
    return value

  else:
    raise AssertionError

osh/lex.py

@@ -406,9 +406,9 @@ def IsKeyword(name):
  # Backslash that ends a line.  Note '.' doesn't match a newline character.
  C('\\\n', Id.Char_Literals),

  # e.g. \A is not an escape, and \x doesn't match a hex escape.  This could be
  # an error.
  C('\\', Id.Char_Literals),
  # e.g. \A is not an escape, and \x doesn't match a hex escape.  We allow it,
  # but a lint tool could warn about it.
  C('\\', Id.Char_BadBackslash),
]

# Used by ECHO_LEXER in core/builtin.py.
@@ -419,8 +419,9 @@ def IsKeyword(name):

  C(r'\c', Id.Char_Stop),

  # Backslash that ends the string.
  R(r'\\$', Id.Char_Literals),
  # Bad Backslash should not end the string.  We allow it, but a lint tool
  # should warn about it.
  R(r'\\$', Id.Char_BadBackslash),

  # e.g. 'foo', anything that's not a backslash escape
  R(r'[^\\]+', Id.Char_Literals),
@@ -434,8 +435,9 @@ def IsKeyword(name):
  # with no leading 0.
  R(r'\\[0-7]{1,3}', Id.Char_Octal3),

  # \' is escaped in $'' mode, but not echo -e
  C(r'\'', Id.Char_OneChar),
  # ' is escaped in $'' mode, but not echo -e.  Ditto fr ", not sure why.
  C(r"\'", Id.Char_OneChar),
  C(r'\"', Id.Char_OneChar),

  # e.g. 'foo', anything that's not a backslash escape.  Need to exclude ' as
  # well.

spec/quote.test.sh

@@ -139,6 +139,12 @@ echo $'foo'
# stdout: foo
# N-I dash stdout: $foo

### $'' with quotes
echo $'single \' double \"'
# stdout: single ' double "
# N-I dash stdout-json: ""
# N-I dash status: 2

### $'' with newlines
echo $'col1\ncol2\ncol3'
# stdout-json: "col1\ncol2\ncol3\n"