Implement echo -e '\c'.

This is an unusual feature which not only truncates the current string,
but stops later arguments from being printed.

dash, bash, and zsh all implement it.

core/builtin.py

@@ -229,6 +229,7 @@ def Resolve(argv0):

ECHO_LEXER = lexer.SimpleLexer([
  R(r'\\[0abeEfrtnv]', Id.Char_OneChar),
  C(r'\c', Id.Char_Stop),

  # Note: tokens above \0377 can either be truncated or be flagged a syntax
  # error in strict mode.
@@ -276,6 +277,9 @@ def _EvalStringPart(id_, value):
    c = value[1]
    return _ONE_CHAR[c]

  elif id_ == Id.Char_Stop:  # \c returns a special sentinel
    return None

  elif id_ == Id.Char_Octal:
    # TODO: Error checking for \0777
    s = value[2:]
@@ -315,34 +319,49 @@ def _EvalStringPart(id_, value):

def Echo(argv):
  """
  echo builtin.  Doesn't depend on executor state.
  echo builtin.

  set -o sane-echo could do the following:
  - only one arg, no implicit joining.
  - no -e: should be echo c'one\ttwo\t'
  - no -n: should be write 'one'

  TODO: Where to put help?  docstring?
  multiple args on a line:
  echo-lines one two three
  """
  # NOTE: both getopt and optparse are unsuitable for 'echo' because:
  # - 'echo -c' should print '-c', not fail
  # - echo '---' should print ---, not fail

  arg, i = echo_spec.ParseLikeEcho(argv)
  arg, arg_index = echo_spec.ParseLikeEcho(argv)
  argv = argv[arg_index:]
  if arg.e:
    new_argv = []
    for a in argv:
      parts = []
      for id_, value in ECHO_LEXER.Tokens(a):
        p = _EvalStringPart(id_, value)

        # Unusual behavior: '\c' prints what is there and aborts processing!
        if p is None:
          new_argv.append(''.join(parts))
          for i, a in enumerate(new_argv):
            if i != 0:
              sys.stdout.write(' ')  # arg separator
            sys.stdout.write(a)
          return 0  # EARLY RETURN

        parts.append(p)
      new_argv.append(''.join(parts))

    # Replace it
    argv = new_argv

  #log('echo argv %s', argv)
  n = len(argv)
  for i in xrange(i, n-1):
    sys.stdout.write(argv[i])
    sys.stdout.write(' ')  # arg separator
  if argv:
    sys.stdout.write(argv[-1])
  for i, a in enumerate(argv):
    if i != 0:
      sys.stdout.write(' ')  # arg separator
    sys.stdout.write(a)
  if not arg.n:
    sys.stdout.write('\n')

core/id_kind.py

@@ -399,7 +399,7 @@ def _AddKinds(spec):

  # For C-escaped strings.
  spec.AddKind('Char', [
      'OneChar', 'Hex', 'Octal', 'Unicode4', 'Unicode8', 'Literals'
      'OneChar', 'Stop', 'Hex', 'Octal', 'Unicode4', 'Unicode8', 'Literals'
  ])

spec/builtin-io.test.sh

@@ -43,6 +43,12 @@ echo -e 'ab\0cd'
# dash truncates it
# BUG dash stdout-json: "-e ab\n"

### \c stops processing input
echo -e xy  'ab\cde'  'ab\cde'
# stdout-json: "xy ab"
# OK dash stdout-json: "-e xy ab"
# N-I mksh stdout-json: "xy abde abde"

### echo -e with hex escape
echo -e 'abcd\x65f'
# stdout-json: "abcdef\n"