def _TwoArgs(w_parser):
"""Returns an expression tree to be evaluated."""
w0 = w_parser.Read()
w1 = w_parser.Read()
if w0.s == '!':
return bool_expr.LogicalNot(bool_expr.WordTest(w1))
unary_id = TEST_UNARY_LOOKUP.get(w0.s)
if unary_id is None:
# TODO:
# - separate lookup by unary
p_die('Expected unary operator, got %r (2 args)', w0.s, word=w0)
return bool_expr.Unary(unary_id, w1)
def _ParseFormatStr(self):
# type: () -> printf_part_t
self._Next(lex_mode_e.PrintfPercent) # move past %
part = printf_part.Percent()
while self.token_type in (Id.Format_Flag, Id.Format_Zero):
# space and + could be implemented
flag = self.cur_token.val
if flag in '# +':
p_die("osh printf doesn't support the %r flag",
flag,
token=self.cur_token)
part.flags.append(self.cur_token)
self._Next(lex_mode_e.PrintfPercent)
if self.token_type in (Id.Format_Num, Id.Format_Star):
part.width = self.cur_token
self._Next(lex_mode_e.PrintfPercent)
if self.token_type == Id.Format_Dot:
part.precision = self.cur_token
self._Next(lex_mode_e.PrintfPercent) # past dot
if self.token_type in (Id.Format_Num, Id.Format_Star,
Id.Format_Zero):
part.precision = self.cur_token
self._Next(lex_mode_e.PrintfPercent)
if self.token_type in (Id.Format_Type, Id.Format_Time):
part.type = self.cur_token
# ADDITIONAL VALIDATION outside the "grammar".
if part.type.val in 'eEfFgG':
p_die("osh printf doesn't support floating point",
token=part.type)
# These two could be implemented. %c needs utf-8 decoding.
if part.type.val == 'c':
p_die("osh printf doesn't support single characters (bytes)",
token=part.type)
else:
if self.cur_token.val:
msg = 'Invalid printf format character'
else: # for printf '%'
msg = 'Expected a printf format character'
p_die(msg, token=self.cur_token)
# Do this check AFTER the floating point checks
if part.precision and part.type.val[-1] not in 'fsT':
p_die("precision can't be specified when here",
token=part.precision)
return part
def ParseForBuiltin(self):
# type: () -> bool_expr_t
"""For test builtin."""
self._Next()
node = self.ParseExpr()
if self.op_id != Id.Eof_Real:
p_die('Unexpected trailing word %s',
word.Pretty(self.cur_word),
word=self.cur_word)
return node
def ParseOilArgList(self, lexer, out):
# type: (Lexer, arg_list) -> token
if self.parsing_expr:
p_die("TODO: can't be nested")
pnode, last_token = self._ParseOil(lexer, grammar_nt.oil_arglist)
if 0:
self.p_printer.Print(pnode)
self.tr.ArgList(pnode, out)
return last_token
def _ReadExtGlobPart(self):
# type: () -> word_part__ExtGlobPart
"""
Grammar:
Item = CompoundWord | EPSILON # important: @(foo|) is allowed
LEFT = '@(' | '*(' | '+(' | '?(' | '!('
RIGHT = ')'
ExtGlob = LEFT (Item '|')* Item RIGHT # ITEM may be empty
CompoundWord includes ExtGlobPart
"""
left_token = self.cur_token
arms = [] # type: List[word_t]
spids = []
spids.append(left_token.span_id)
self.lexer.PushHint(Id.Op_RParen, Id.Right_ExtGlob)
self._Next(lex_mode_e.ExtGlob) # advance past LEFT
read_word = False # did we just a read a word? To handle @(||).
while True:
self._Peek()
if self.token_type == Id.Right_ExtGlob:
if not read_word:
arms.append(osh_word.CompoundWord())
spids.append(self.cur_token.span_id)
break
elif self.token_type == Id.Op_Pipe:
if not read_word:
arms.append(osh_word.CompoundWord())
read_word = False
self._Next(lex_mode_e.ExtGlob)
# lex mode EXTGLOB should only produce these 4 kinds of tokens
elif self.token_kind in (Kind.Lit, Kind.Left, Kind.VSub,
Kind.ExtGlob):
w = self._ReadCompoundWord(lex_mode=lex_mode_e.ExtGlob)
arms.append(w)
read_word = True
elif self.token_kind == Kind.Eof:
p_die('Unexpected EOF reading extended glob that began here',
token=left_token)
else:
raise AssertionError('Unexpected token %r' % self.cur_token)
part = word_part.ExtGlobPart(left_token, arms)
part.spids.extend(spids)
return part
def _TwoArgs(argv):
"""Returns an expression tree to be evaluated."""
a0, a1 = argv
if a0 == '!':
return ast.LogicalNot(_StringWordTest(a1))
unary_id = _UNARY_LOOKUP.get(a0)
if unary_id is None:
# TODO:
# - syntax error
# - separate lookup by unary
util.p_die('Expected unary operator, got %r (2 args)', a0)
child = ast.StringWord(Id.Word_Compound, a1)
return ast.BoolUnary(unary_id, child)
def _TwoArgs(w_parser):
# type: (_StringWordEmitter) -> bool_expr_t
"""Returns an expression tree to be evaluated."""
w0 = w_parser.Read()
w1 = w_parser.Read()
if w0.s == '!':
return bool_expr.LogicalNot(bool_expr.WordTest(w1))
unary_id = match.BracketUnary(w0.s)
if unary_id == -1:
# TODO:
# - separate lookup by unary
p_die('Expected unary operator, got %r (2 args)', w0.s, word=w0)
return bool_expr.Unary(unary_id, w1)
def ParseProc(self, node):
# type: (command__Proc) -> None
# proc name-with-hyphens() must be accepted
self._Next(lex_mode_e.ShCommand)
self._Peek()
# example: 'proc f[' gets you Lit_ArrayLhsOpen
if self.token_type != Id.Lit_Chars:
p_die('Invalid proc name %r', self.cur_token.val, token=self.cur_token)
node.name = self.cur_token
last_token = self.parse_ctx.ParseProc(self.lexer, node)
if last_token.id == Id.Op_LBrace: # Translate to what CommandParser wants
last_token.id = Id.Lit_LBrace
self.buffered_word = word.Token(last_token)
self._Next(lex_mode_e.ShCommand) # required
def _ReadLikeDQ(self, left_dq_token, out_parts):
# type: (Optional[Token], List[word_part_t]) -> None
"""
Args:
left_dq_token: A token if we are reading a double quoted part, or None if
we're reading a here doc.
out_parts: list of word_part to append to
"""
done = False
while not done:
self._Next(lex_mode_e.DQ)
self._Peek()
if self.token_kind == Kind.Lit:
if self.token_type == Id.Lit_EscapedChar:
part = word_part.EscapedLiteral(self.cur_token) # type: word_part_t
else:
part = self.cur_token
out_parts.append(part)
elif self.token_kind == Kind.Left:
part = self._ReadDoubleQuotedLeftParts()
out_parts.append(part)
elif self.token_kind == Kind.VSub:
part = simple_var_sub(self.cur_token)
out_parts.append(part)
# NOTE: parsing "$f(x)" would BREAK CODE. Could add a more for it
# later.
elif self.token_kind == Kind.Right:
assert self.token_type == Id.Right_DoubleQuote, self.token_type
if left_dq_token:
done = True
else:
# In a here doc, the right quote is literal!
out_parts.append(self.cur_token)
elif self.token_kind == Kind.Eof:
if left_dq_token:
p_die('Unexpected EOF reading double-quoted string that began here',
token=left_dq_token)
else: # here docs will have an EOF in their token stream
done = True
else:
raise AssertionError(self.cur_token)
def _NameInClass(self, negated_tok, tok):
# type: (token, token) -> class_literal_term_t
"""
Like the above, but 'dot' doesn't mean anything.
"""
if negated_tok: # For error messages
negated_speck = speck(negated_tok.id, negated_tok.span_id)
else:
negated_speck = None
val = tok.val
if val in self.POSIX_CLASSES:
return posix_class(negated_speck, val)
perl = self.PERL_CLASSES.get(val)
if perl:
return perl_class(negated_speck, perl)
p_die("%r isn't a character class", val, token=tok)
def _ReadArithSub2Part(self):
# type: () -> word_part__ArithSubPart
"""Non-standard arith sub $[a + 1]."""
left_span_id = self.cur_token.span_id
self._Next(lex_mode_e.Arith)
anode = self._ReadArithExpr()
if self.token_type != Id.Arith_RBracket:
p_die('Expected ], got %r',
self.cur_token.val,
token=self.cur_token)
right_span_id = self.cur_token.span_id
node = word_part.ArithSubPart(anode)
node.spids.append(left_span_id)
node.spids.append(right_span_id)
return node
def _ReadArithSubPart(self):
# type: () -> word_part__ArithSubPart
"""
Read an arith substitution, which contains an arith expression, e.g.
$((a + 1)).
"""
left_span_id = self.cur_token.span_id
# The second one needs to be disambiguated in stuff like stuff like:
# $(echo $(( 1+2 )) )
self.lexer.PushHint(Id.Op_RParen, Id.Right_ArithSub)
# NOTE: To disambiguate $(( as arith sub vs. command sub and subshell, we
# could save the lexer/reader state here, and retry if the arithmetic parse
# fails. But we can almost always catch this at parse time. There could
# be some exceptions like:
# $((echo * foo)) # looks like multiplication
# $((echo / foo)) # looks like division
self._Next(lex_mode_e.Arith)
anode = self._ReadArithExpr()
if self.token_type != Id.Arith_RParen:
p_die('Expected first ) to end arith sub, got %r',
self.cur_token.val,
token=self.cur_token)
self._Next(lex_mode_e.Outer) # TODO: This could be DQ or ARITH too
# PROBLEM: $(echo $(( 1 + 2 )) )
# Two right parens break the Id.Eof_RParen scheme
self._Peek()
if self.token_type != Id.Right_ArithSub:
p_die('Expected second ) to end arith sub, got %r',
self.cur_token.val,
token=self.cur_token)
right_span_id = self.cur_token.span_id
node = word_part.ArithSubPart(anode)
node.spids.append(left_span_id)
node.spids.append(right_span_id)
return node
def ParseVarDecl(self, kw_token, lexer):
# type: (Token, Lexer) -> Tuple[command__VarDecl, Token]
""" var mylist = [1, 2, 3] """
# TODO: We do need re-entrancy for var x = @[ (1+2) ] and such
if self.parsing_expr:
p_die("ShAssignment expression can't be nested like this", token=kw_token)
self.parsing_expr = True
try:
pnode, last_token = self.e_parser.Parse(lexer, grammar_nt.oil_var_decl)
finally:
self.parsing_expr = False
if 0:
self.p_printer.Print(pnode)
ast_node = self.tr.MakeVarDecl(pnode)
ast_node.keyword = kw_token # VarDecl didn't fill this in
return ast_node, last_token
def _Classify(gr, tok):
# type: (Grammar, token) -> int
# We have to match up what ParserGenerator.make_grammar() did when
# calling make_label() and make_first(). See classify() in
# opy/pgen2/driver.py.
# 'x' and 'for' are both tokenized as Expr_Name. This handles the 'for'
# case.
if tok.id == Id.Expr_Name:
if tok.val in gr.keywords:
return gr.keywords[tok.val]
# This handles 'x'.
typ = tok.id.enum_id
if typ in gr.tokens:
return gr.tokens[typ]
type_str = '' if tok.id == Id.Unknown_Tok else (' (%s)' % tok.id.name)
p_die('Unexpected token in expression mode%s', type_str, token=tok)
def _Tuple(self, children):
# type: (List[PNode]) -> expr_t
n = len(children)
# (x) -- not a tuple
if n == 1:
return self.Expr(children[0])
# x, and (x,) aren't allowed
if n == 2:
p_die('Write singleton tuples with tup(), not a trailing comma',
token=children[1].tok)
elts = [] # type: List[expr_t]
for i in xrange(0, n, 2): # skip commas
p_node = children[i]
elts.append(self.Expr(p_node))
return expr.Tuple(elts, expr_context_e.Store) # unused expr_context_e
def _ReadArithWord(self):
# type: () -> Tuple[word_t, bool]
"""Helper function for ReadArithWord."""
self._Peek()
if self.token_kind == Kind.Unknown:
p_die('Unexpected token in arithmetic context',
token=self.cur_token)
elif self.token_kind == Kind.Eof:
# Just return EOF token
w = osh_word.TokenWord(self.cur_token) # type: word_t
return w, False
elif self.token_kind == Kind.Ignored:
# Space should be ignored. TODO: change this to SPACE_SPACE and
# SPACE_NEWLINE? or SPACE_TOK.
self._Next(lex_mode_e.Arith)
return None, True # Tell wrapper to try again
elif self.token_kind in (Kind.Arith, Kind.Right):
# Id.Right_ArithSub IS just a normal token, handled by ArithParser
self._Next(lex_mode_e.Arith)
w = osh_word.TokenWord(self.cur_token)
return w, False
elif self.token_kind in (Kind.Lit, Kind.Left):
w = self._ReadCompoundWord(lex_mode=lex_mode_e.Arith)
return w, False
elif self.token_kind == Kind.VSub:
part = word_part.SimpleVarSub(self.cur_token)
self._Next(lex_mode_e.Arith)
w = osh_word.CompoundWord([part])
return w, False
else:
assert False, ("Unexpected token parsing arith sub: %s" %
self.cur_token)
raise AssertionError("Shouldn't get here")
def _ReadSliceVarOp(self):
# type: () -> suffix_op__Slice
""" VarOf ':' ArithExpr (':' ArithExpr )? """
self._Next(lex_mode_e.Arith)
self._Peek()
if self.token_type == Id.Arith_Colon: # A pun for Id.VOp2_Colon
begin = None # no beginning specified
else:
begin = self._ReadArithExpr()
if self.token_type == Id.Arith_RBrace:
return suffix_op.Slice(begin, None) # No length specified
# Id.Arith_Colon is a pun for Id.VOp2_Colon
if self.token_type == Id.Arith_Colon:
self._Next(lex_mode_e.Arith)
length = self._ReadArithExpr()
return suffix_op.Slice(begin, length)
p_die("Unexpected token in slice: %r", self.cur_token.val,
token=self.cur_token)
def _Classify(gr, tok):
# type: (Grammar, token) -> int
# We have to match up what ParserGenerator.make_grammar() did when
# calling make_label() and make_first(). See classify() in
# opy/pgen2/driver.py.
# 'x' and 'for' are both tokenized as Expr_Name. This handles the 'for'
# case.
if tok.id == Id.Expr_Name:
ilabel = gr.keywords.get(tok.val)
if ilabel is not None:
return ilabel
# This handles 'x'.
typ = tok.id.enum_id
ilabel = gr.tokens.get(typ)
if ilabel is not None:
return ilabel
p_die('Invalid token %s', tok, token=tok)
def _PlaceList(self, p_node):
# type: (PNode) -> List[place_expr_t]
"""
place_list: expr (',' expr)*
"""
assert p_node.typ == grammar_nt.place_list
places = [] # type: List[place_expr_t]
for p in p_node.children[::2]:
e = self.Expr(p)
if e.tag == expr_e.Var: # COMPATIBILITY hack
assert isinstance(e, expr__Var)
places.append(place_expr.Var(e.name))
elif e.tag in (place_expr_e.Var, place_expr_e.Subscript,
place_expr_e.Attribute):
places.append(cast(place_expr_t, e))
else:
# This blame mechanism seems to work. Otherwise we don't have a method
# to blame an arbitrary expr_t.
p_die("Can't assign to this expression",
token=p.tok if p.tok else None)
return places
def _NameInRegex(self, negated_tok, tok):
# type: (token, token) -> re_t
if negated_tok: # For error messages
negated_speck = speck(negated_tok.id, negated_tok.span_id)
else:
negated_speck = None
val = tok.val
if val == 'dot':
if negated_tok:
p_die("Can't negate this symbol", token=tok)
return tok
if val in self.POSIX_CLASSES:
return posix_class(negated_speck, val)
perl = self.PERL_CLASSES.get(val)
if perl:
return perl_class(negated_speck, perl)
p_die("%r isn't a character class", val, token=tok)
def _ReadSliceVarOp(self):
# type: () -> suffix_op__Slice
""" VarOf ':' ArithExpr (':' ArithExpr )? """
self._Next(lex_mode_e.Arith)
self._Peek()
if self.token_type == Id.Arith_Colon: # A pun for Id.VOp2_Colon
# no beginning specified
begin = None # type: Optional[arith_expr_t]
else:
begin = self._ReadArithExpr()
if self.token_type == Id.Arith_RBrace:
no_length = None # type: Optional[arith_expr_t] # No length specified
return suffix_op.Slice(begin, no_length)
# Id.Arith_Colon is a pun for Id.VOp2_Colon
if self.token_type == Id.Arith_Colon:
self._Next(lex_mode_e.Arith)
length = self._ReadArithExpr()
return suffix_op.Slice(begin, length)
p_die("Expected : or } in slice", token=self.cur_token)
def _ReadPatSubVarOp(self, lex_mode):
# type: (lex_mode_t) -> suffix_op__PatSub
"""
Match = ('/' | '#' | '%') WORD
VarSub = ...
| VarOf '/' Match '/' WORD
"""
pat = self._ReadVarOpArg(lex_mode,
eof_type=Id.Lit_Slash,
empty_ok=False)
assert isinstance(pat, word__CompoundWord) # Because empty_ok=False
if len(pat.parts) == 1:
ok, s, quoted = word.StaticEval(pat)
if ok and s == '/' and not quoted: # Looks like ${a////c}, read again
self._Next(lex_mode)
self._Peek()
p = word_part.LiteralPart(self.cur_token)
pat.parts.append(p)
if len(pat.parts) == 0:
p_die('Pattern in ${x/pat/replace} must not be empty',
token=self.cur_token)
replace_mode = Id.Undefined_Tok
# Check for / # % modifier on pattern.
first_part = pat.parts[0]
if isinstance(first_part, word_part__LiteralPart):
lit_id = first_part.token.id
if lit_id in (Id.Lit_Slash, Id.Lit_Pound, Id.Lit_Percent):
pat.parts.pop(0)
replace_mode = lit_id
# NOTE: If there is a modifier, the pattern can be empty, e.g.
# ${s/#/foo} and ${a/%/foo}.
if self.token_type == Id.Right_VarSub:
# e.g. ${v/a} is the same as ${v/a/} -- empty replacement string
return suffix_op.PatSub(pat, None, replace_mode)
if self.token_type == Id.Lit_Slash:
replace = self._ReadVarOpArg(lex_mode) # do not stop at /
self._Peek()
if self.token_type != Id.Right_VarSub:
# NOTE: I think this never happens.
# We're either in the VS_ARG_UNQ or VS_ARG_DQ lex state, and everything
# there is Lit_ or Left_, except for }.
p_die("Expected } after replacement string, got %s",
self.cur_token,
token=self.cur_token)
return suffix_op.PatSub(pat, replace, replace_mode)
# Happens with ${x//} and ${x///foo}, see test/parse-errors.sh
p_die("Expected } after pat sub, got %r",
self.cur_token.val,
token=self.cur_token)
def ReadForExpression(self):
# type: () -> command__ForExpr
"""Read ((i=0; i<5; ++i)) -- part of command context."""
self._NextNonSpace() # skip over ((
self._Peek()
if self.token_type == Id.Arith_Semi: # for (( ; i < 10; i++ ))
init_node = None # type: Optional[arith_expr_t]
else:
init_node = self._ReadArithExpr()
self._NextNonSpace()
self._Peek()
if self.token_type == Id.Arith_Semi: # for (( ; ; i++ ))
cond_node = None # type: Optional[arith_expr_t]
else:
cond_node = self._ReadArithExpr()
self._NextNonSpace()
self._Peek()
if self.token_type == Id.Arith_RParen: # for (( ; ; ))
update_node = None # type: Optional[arith_expr_t]
else:
update_node = self._ReadArithExpr()
self._NextNonSpace()
self._Peek()
if self.token_type != Id.Arith_RParen:
p_die('Expected ) to end for loop expression', token=self.cur_token)
self._Next(lex_mode_e.ShCommand)
node = command.ForExpr()
node.init = init_node
node.cond = cond_node
node.update = update_node
return node
def _PlaceList(self, p_node):
# type: (PNode) -> List[place_expr_t]
"""
place_list: expr (',' expr)*
"""
assert p_node.typ == grammar_nt.place_list
places = [] # type: List[place_expr_t]
n = len(p_node.children)
for i in xrange(0, n, 2): # was children[::2]
p = p_node.children[i]
e = self.Expr(p)
UP_e = e
tag = e.tag_()
if tag == expr_e.Var: # COMPATIBILITY hack
e = cast(expr__Var, UP_e)
places.append(place_expr.Var(e.name))
elif tag in (
place_expr_e.Var, place_expr_e.Subscript, place_expr_e.Attribute):
places.append(cast(place_expr_t, UP_e))
else:
# This blame mechanism seems to work. Otherwise we don't have a method
# to blame an arbitrary expr_t.
p_die("Can't assign to this expression", token=p.tok if p.tok else None)
return places
def Parse(self, lexer, start_symbol):
# type: (Lexer, int) -> Tuple[PNode, Token]
# Reuse the parser
self.push_parser.setup(start_symbol)
try:
last_token = _PushOilTokens(self.parse_ctx, self.gr,
self.push_parser, lexer)
except parse.ParseError as e:
#log('ERROR %s', e)
# TODO:
# - Describe what lexer mode we're in (Invalid syntax in regex)
# - Maybe say where the mode started
# - Id.Unknown_Tok could say "This character is invalid"
# ParseError has a "too much input" case but I haven't been able to
# tickle it. Mabye it's because of the Eof tokens?
p_die('Syntax error in expression (near %s)',
ui.PrettyId(e.tok.id),
token=e.tok)
#raise error.Parse('Syntax error in expression', token=e.tok)
return self.push_parser.rootnode, last_token
def Parse(self):
self._Next(lex_mode_e.PrintfOuter)
parts = []
while True:
if (self.token_kind == Kind.Char
or self.token_type == Id.Format_EscapedPercent):
# TODO: Could handle Char_BadBackslash.
# Maybe make that a different kind?
parts.append(printf_part.Literal(self.cur_token))
elif self.token_type == Id.Format_Percent:
parts.append(self._ParseFormatStr())
elif self.token_type == Id.Eof_Real:
break
else:
p_die('Invalid token %r', token=self.cur_token)
self._Next(lex_mode_e.PrintfOuter)
return parts
def _ReadPatSubVarOp(self):
# type: () -> suffix_op__PatSub
"""
Match = ('/' | '#' | '%') WORD
VarSub = ...
| VarOf '/' Match '/' WORD
"""
# Exception: VSub_ArgUnquoted even if it's quoted
# stop at eof_type=Lit_Slash, empty_ok=False
UP_pat = self._ReadVarOpArg3(lex_mode_e.VSub_ArgUnquoted, Id.Lit_Slash, False)
assert UP_pat.tag_() == word_e.Compound, UP_pat # Because empty_ok=False
pat = cast(compound_word, UP_pat)
if len(pat.parts) == 1:
ok, s, quoted = word_.StaticEval(pat)
if ok and s == '/' and not quoted: # Looks like ${a////c}, read again
self._Next(lex_mode_e.VSub_ArgUnquoted)
self._Peek()
pat.parts.append(self.cur_token)
if len(pat.parts) == 0:
p_die('Pattern in ${x/pat/replace} must not be empty',
token=self.cur_token)
replace_mode = Id.Undefined_Tok
# Check for / # % modifier on pattern.
UP_first_part = pat.parts[0]
if UP_first_part.tag_() == word_part_e.Literal:
lit_id = cast(Token, UP_first_part).id
if lit_id in (Id.Lit_Slash, Id.Lit_Pound, Id.Lit_Percent):
pat.parts.pop(0)
replace_mode = lit_id
# NOTE: If there is a modifier, the pattern can be empty, e.g.
# ${s/#/foo} and ${a/%/foo}.
if self.token_type == Id.Right_DollarBrace:
# e.g. ${v/a} is the same as ${v/a/} -- empty replacement string
return suffix_op.PatSub(pat, None, replace_mode)
if self.token_type == Id.Lit_Slash:
replace = self._ReadVarOpArg(lex_mode_e.VSub_ArgUnquoted) # do not stop at /
self._Peek()
if self.token_type != Id.Right_DollarBrace:
# NOTE: I think this never happens.
# We're either in the VS_ARG_UNQ or VS_ARG_DQ lex state, and everything
# there is Lit_ or Left_, except for }.
p_die("Expected } after replacement string, got %s",
ui.PrettyId(self.token_type), token=self.cur_token)
return suffix_op.PatSub(pat, replace, replace_mode)
# Happens with ${x//} and ${x///foo}, see test/parse-errors.sh
p_die('Expected } or / to close pattern', token=self.cur_token)
def _RangeChar(self, p_node):
# type: (PNode) -> str
"""Evaluate a range endpoints.
- the 'a' in 'a'-'z'
- the \x00 in \x00-\x01
etc.
TODO: This function doesn't respect the LST invariant.
"""
assert p_node.typ == grammar_nt.range_char, p_node
children = p_node.children
typ = children[0].typ
if ISNONTERMINAL(typ):
# 'a' in 'a'-'b'
if typ == grammar_nt.sq_string:
sq_part = cast(single_quoted, children[0].children[1].tok)
tokens = sq_part.tokens
if len(
tokens
) > 1: # Can happen with multiline single-quoted strings
p_die(RANGE_POINT_TOO_LONG, part=sq_part)
if len(tokens[0].val) > 1:
p_die(RANGE_POINT_TOO_LONG, part=sq_part)
s = tokens[0].val[0]
return s
if typ == grammar_nt.char_literal:
raise AssertionError('TODO')
# TODO: This brings in a lot of dependencies, and this type checking
# errors. We want to respect the LST invariant anyway.
#from osh import word_compile
#tok = children[0].children[0].tok
#s = word_compile.EvalCStringToken(tok.id, tok.val)
#return s
raise NotImplementedError()
else:
# Expr_Name or Expr_DecInt
tok = p_node.tok
if tok.id in (Id.Expr_Name, Id.Expr_DecInt):
# For the a in a-z, 0 in 0-9
if len(tok.val) != 1:
p_die(RANGE_POINT_TOO_LONG, token=tok)
return tok.val[0]
raise NotImplementedError()
def NullError(p, t, bp):
# type: (TdopParser, word_t, int) -> arith_expr_t
# TODO: I need position information
p_die("Token can't be used in prefix position", word=t)
return None # never reached
def LeftError(p, t, left, rbp):
# type: (TdopParser, word_t, arith_expr_t, int) -> arith_expr_t
# Hm is this not called because of binding power?
p_die("Token can't be used in infix position", word=t)
return None # never reached