| 1 | """
|
| 2 | tdop.py - Library for expression parsing.
|
| 3 | """
|
| 4 |
|
| 5 | from _devbuild.gen.id_kind_asdl import Id, Id_t
|
| 6 | from _devbuild.gen.syntax_asdl import (loc, arith_expr, arith_expr_e,
|
| 7 | arith_expr_t, word_t, CompoundWord,
|
| 8 | NameTok)
|
| 9 | from core.error import p_die
|
| 10 | from core import ui
|
| 11 | from frontend import lexer
|
| 12 | from mycpp import mylib
|
| 13 | from mycpp.mylib import tagswitch
|
| 14 | from osh import word_
|
| 15 |
|
| 16 | from typing import (Callable, List, Dict, Tuple, Any, cast, TYPE_CHECKING)
|
| 17 |
|
| 18 | if TYPE_CHECKING: # break circular dep
|
| 19 | from osh.word_parse import WordParser
|
| 20 | from core import optview
|
| 21 | LeftFunc = Callable[['TdopParser', word_t, arith_expr_t, int],
|
| 22 | arith_expr_t]
|
| 23 | NullFunc = Callable[['TdopParser', word_t, int], arith_expr_t]
|
| 24 |
|
| 25 |
|
| 26 | def IsIndexable(node):
|
| 27 | # type: (arith_expr_t) -> bool
|
| 28 | """In POSIX shell arith, a[1] is allowed but a[1][1] isn't.
|
| 29 |
|
| 30 | We also allow $a[i] and foo$x[i] (formerly parse_dynamic_arith)
|
| 31 | """
|
| 32 | with tagswitch(node) as case:
|
| 33 | if case(arith_expr_e.VarSub, arith_expr_e.Word):
|
| 34 | return True
|
| 35 | return False
|
| 36 |
|
| 37 |
|
| 38 | def CheckLhsExpr(node, blame_word):
|
| 39 | # type: (arith_expr_t, word_t) -> None
|
| 40 | """Determine if a node is a valid L-value by whitelisting tags.
|
| 41 |
|
| 42 | Valid:
|
| 43 | x = y
|
| 44 | a[1] = y
|
| 45 | Invalid:
|
| 46 | a[0][0] = y
|
| 47 | """
|
| 48 | UP_node = node
|
| 49 | if node.tag() == arith_expr_e.Binary:
|
| 50 | node = cast(arith_expr.Binary, UP_node)
|
| 51 | if node.op_id == Id.Arith_LBracket and IsIndexable(node.left):
|
| 52 | return
|
| 53 | # But a[0][0] = 1 is NOT valid.
|
| 54 |
|
| 55 | if IsIndexable(node):
|
| 56 | return
|
| 57 |
|
| 58 | p_die("Left-hand side of this assignment is invalid", loc.Word(blame_word))
|
| 59 |
|
| 60 |
|
| 61 | #
|
| 62 | # Null Denotation
|
| 63 | #
|
| 64 |
|
| 65 |
|
| 66 | def NullError(p, t, bp):
|
| 67 | # type: (TdopParser, word_t, int) -> arith_expr_t
|
| 68 | # TODO: I need position information
|
| 69 | p_die("Token can't be used in prefix position", loc.Word(t))
|
| 70 | return None # never reached
|
| 71 |
|
| 72 |
|
| 73 | def NullConstant(p, w, bp):
|
| 74 | # type: (TdopParser, word_t, int) -> arith_expr_t
|
| 75 | name_tok = word_.LooksLikeArithVar(w)
|
| 76 | if name_tok:
|
| 77 | return NameTok(name_tok, lexer.TokenVal(name_tok))
|
| 78 |
|
| 79 | # Id.Word_Compound in the spec ensures this cast is valid
|
| 80 | return cast(CompoundWord, w)
|
| 81 |
|
| 82 |
|
| 83 | def NullParen(p, t, bp):
|
| 84 | # type: (TdopParser, word_t, int) -> arith_expr_t
|
| 85 | """Arithmetic grouping."""
|
| 86 | r = p.ParseUntil(bp)
|
| 87 | p.Eat(Id.Arith_RParen)
|
| 88 | return r
|
| 89 |
|
| 90 |
|
| 91 | def NullPrefixOp(p, w, bp):
|
| 92 | # type: (TdopParser, word_t, int) -> arith_expr_t
|
| 93 | """Prefix operator.
|
| 94 |
|
| 95 | Low precedence: return, raise, etc.
|
| 96 | return x+y is return (x+y), not (return x) + y
|
| 97 |
|
| 98 | High precedence: logical negation, bitwise complement, etc.
|
| 99 | !x && y is (!x) && y, not !(x && y)
|
| 100 | """
|
| 101 | right = p.ParseUntil(bp)
|
| 102 | return arith_expr.Unary(word_.ArithId(w), right)
|
| 103 |
|
| 104 |
|
| 105 | #
|
| 106 | # Left Denotation
|
| 107 | #
|
| 108 |
|
| 109 |
|
| 110 | def LeftError(p, t, left, rbp):
|
| 111 | # type: (TdopParser, word_t, arith_expr_t, int) -> arith_expr_t
|
| 112 | # Hm is this not called because of binding power?
|
| 113 | p_die("Token can't be used in infix position", loc.Word(t))
|
| 114 | return None # never reached
|
| 115 |
|
| 116 |
|
| 117 | def LeftBinaryOp(p, w, left, rbp):
|
| 118 | # type: (TdopParser, word_t, arith_expr_t, int) -> arith_expr_t
|
| 119 | """Normal binary operator like 1+2 or 2*3, etc."""
|
| 120 | # TODO: w should be a Token, and we should extract the token from it.
|
| 121 | return arith_expr.Binary(word_.ArithId(w), left, p.ParseUntil(rbp))
|
| 122 |
|
| 123 |
|
| 124 | def LeftAssign(p, w, left, rbp):
|
| 125 | # type: (TdopParser, word_t, arith_expr_t, int) -> arith_expr_t
|
| 126 | """Normal binary operator like 1+2 or 2*3, etc."""
|
| 127 | # x += 1, or a[i] += 1
|
| 128 |
|
| 129 | CheckLhsExpr(left, w)
|
| 130 | return arith_expr.BinaryAssign(word_.ArithId(w), left, p.ParseUntil(rbp))
|
| 131 |
|
| 132 |
|
| 133 | #
|
| 134 | # Parser definition
|
| 135 | # TODO: To be consistent, move this to osh/tdop_def.py.
|
| 136 | #
|
| 137 |
|
| 138 | if mylib.PYTHON:
|
| 139 |
|
| 140 | def _ModuleAndFuncName(f):
|
| 141 | # type: (Any) -> Tuple[str, str]
|
| 142 | namespace = f.__module__.split('.')[-1]
|
| 143 | return namespace, f.__name__
|
| 144 |
|
| 145 | def _CppFuncName(f):
|
| 146 | # type: (Any) -> str
|
| 147 | return '%s::%s' % _ModuleAndFuncName(f)
|
| 148 |
|
| 149 | class LeftInfo(object):
|
| 150 | """Row for operator.
|
| 151 |
|
| 152 | In C++ this should be a big array.
|
| 153 | """
|
| 154 |
|
| 155 | def __init__(self, led=None, lbp=0, rbp=0):
|
| 156 | # type: (LeftFunc, int, int) -> None
|
| 157 | self.led = led or LeftError
|
| 158 | self.lbp = lbp
|
| 159 | self.rbp = rbp
|
| 160 |
|
| 161 | def __str__(self):
|
| 162 | # type: () -> str
|
| 163 | """Used by C++ code generation."""
|
| 164 | return '{ %s, %d, %d },' % (_CppFuncName(
|
| 165 | self.led), self.lbp, self.rbp)
|
| 166 |
|
| 167 | def ModuleAndFuncName(self):
|
| 168 | # type: () -> Tuple[str, str]
|
| 169 | """Used by C++ code generation."""
|
| 170 | return _ModuleAndFuncName(self.led)
|
| 171 |
|
| 172 | class NullInfo(object):
|
| 173 | """Row for operator.
|
| 174 |
|
| 175 | In C++ this should be a big array.
|
| 176 | """
|
| 177 |
|
| 178 | def __init__(self, nud=None, bp=0):
|
| 179 | # type: (NullFunc, int) -> None
|
| 180 | self.nud = nud or LeftError
|
| 181 | self.bp = bp
|
| 182 |
|
| 183 | def __str__(self):
|
| 184 | # type: () -> str
|
| 185 | """Used by C++ code generation."""
|
| 186 | return '{ %s, %d },' % (_CppFuncName(self.nud), self.bp)
|
| 187 |
|
| 188 | def ModuleAndFuncName(self):
|
| 189 | # type: () -> Tuple[str, str]
|
| 190 | """Used by C++ code generation."""
|
| 191 | return _ModuleAndFuncName(self.nud)
|
| 192 |
|
| 193 | class ParserSpec(object):
|
| 194 | """Specification for a TDOP parser.
|
| 195 |
|
| 196 | This can be compiled to a table in C++.
|
| 197 | """
|
| 198 |
|
| 199 | def __init__(self):
|
| 200 | # type: () -> None
|
| 201 | self.nud_lookup = {} # type: Dict[Id_t, NullInfo]
|
| 202 | self.led_lookup = {} # type: Dict[Id_t, LeftInfo]
|
| 203 |
|
| 204 | def Null(self, bp, nud, tokens):
|
| 205 | # type: (int, NullFunc, List[Id_t]) -> None
|
| 206 | """Register a token that doesn't take anything on the left.
|
| 207 |
|
| 208 | Examples: constant, prefix operator, error.
|
| 209 | """
|
| 210 | for token in tokens:
|
| 211 | self.nud_lookup[token] = NullInfo(nud=nud, bp=bp)
|
| 212 | if token not in self.led_lookup:
|
| 213 | self.led_lookup[token] = LeftInfo() # error
|
| 214 |
|
| 215 | def _RegisterLed(self, lbp, rbp, led, tokens):
|
| 216 | # type: (int, int, LeftFunc, List[Id_t]) -> None
|
| 217 | for token in tokens:
|
| 218 | if token not in self.nud_lookup:
|
| 219 | self.nud_lookup[token] = NullInfo(NullError)
|
| 220 | self.led_lookup[token] = LeftInfo(lbp=lbp, rbp=rbp, led=led)
|
| 221 |
|
| 222 | def Left(self, bp, led, tokens):
|
| 223 | # type: (int, LeftFunc, List[Id_t]) -> None
|
| 224 | """Register a token that takes an expression on the left."""
|
| 225 | self._RegisterLed(bp, bp, led, tokens)
|
| 226 |
|
| 227 | def LeftRightAssoc(self, bp, led, tokens):
|
| 228 | # type: (int, LeftFunc, List[Id_t]) -> None
|
| 229 | """Register a right associative operator."""
|
| 230 | self._RegisterLed(bp, bp - 1, led, tokens)
|
| 231 |
|
| 232 | def LookupNud(self, token):
|
| 233 | # type: (Id_t) -> NullInfo
|
| 234 |
|
| 235 | # As long as the table is complete, this shouldn't fail
|
| 236 | return self.nud_lookup[token]
|
| 237 |
|
| 238 | def LookupLed(self, token):
|
| 239 | # type: (Id_t) -> LeftInfo
|
| 240 | """Get a left_info for the token."""
|
| 241 |
|
| 242 | # As long as the table is complete, this shouldn't fail
|
| 243 | return self.led_lookup[token]
|
| 244 |
|
| 245 |
|
| 246 | class TdopParser(object):
|
| 247 |
|
| 248 | def __init__(self, spec, w_parser, parse_opts):
|
| 249 | # type: (ParserSpec, WordParser, optview.Parse) -> None
|
| 250 | self.spec = spec
|
| 251 | self.w_parser = w_parser
|
| 252 | self.parse_opts = parse_opts
|
| 253 |
|
| 254 | # NOTE: Next() overwrites this state, so we don't need a Reset() method in
|
| 255 | # between reuses of this TdopParser instance.
|
| 256 | self.cur_word = None # type: word_t # current token
|
| 257 | self.op_id = Id.Undefined_Tok
|
| 258 |
|
| 259 | def CurrentId(self):
|
| 260 | # type: () -> Id_t
|
| 261 | """Glue used by the WordParser to check for extra tokens."""
|
| 262 | return word_.CommandId(self.cur_word)
|
| 263 |
|
| 264 | def AtToken(self, token_type):
|
| 265 | # type: (Id_t) -> bool
|
| 266 | return self.op_id == token_type
|
| 267 |
|
| 268 | def Eat(self, token_type):
|
| 269 | # type: (Id_t) -> None
|
| 270 | """Assert that we're at the current token and advance."""
|
| 271 | if not self.AtToken(token_type):
|
| 272 | p_die(
|
| 273 | 'Parser expected %s, got %s' %
|
| 274 | (ui.PrettyId(token_type), ui.PrettyId(self.op_id)),
|
| 275 | loc.Word(self.cur_word))
|
| 276 | self.Next()
|
| 277 |
|
| 278 | def Next(self):
|
| 279 | # type: () -> bool
|
| 280 | self.cur_word = self.w_parser.ReadArithWord()
|
| 281 | self.op_id = word_.ArithId(self.cur_word)
|
| 282 | return True
|
| 283 |
|
| 284 | def ParseUntil(self, rbp):
|
| 285 | # type: (int) -> arith_expr_t
|
| 286 | """Parse to the right, eating tokens until we encounter a token with
|
| 287 | binding power LESS THAN OR EQUAL TO rbp."""
|
| 288 | # TODO: use Kind.Eof
|
| 289 | if self.op_id in (Id.Eof_Real, Id.Eof_RParen, Id.Eof_Backtick):
|
| 290 | p_die('Unexpected end of input', loc.Word(self.cur_word))
|
| 291 |
|
| 292 | t = self.cur_word
|
| 293 | null_info = self.spec.LookupNud(self.op_id)
|
| 294 |
|
| 295 | self.Next() # skip over the token, e.g. ! ~ + -
|
| 296 | node = null_info.nud(self, t, null_info.bp)
|
| 297 |
|
| 298 | while True:
|
| 299 | t = self.cur_word
|
| 300 | left_info = self.spec.LookupLed(self.op_id)
|
| 301 |
|
| 302 | # Examples:
|
| 303 | # If we see 1*2+ , rbp = 27 and lbp = 25, so stop.
|
| 304 | # If we see 1+2+ , rbp = 25 and lbp = 25, so stop.
|
| 305 | # If we see 1**2**, rbp = 26 and lbp = 27, so keep going.
|
| 306 | if rbp >= left_info.lbp:
|
| 307 | break
|
| 308 | self.Next() # skip over the token, e.g. / *
|
| 309 |
|
| 310 | node = left_info.led(self, t, node, left_info.rbp)
|
| 311 |
|
| 312 | return node
|
| 313 |
|
| 314 | def Parse(self):
|
| 315 | # type: () -> arith_expr_t
|
| 316 |
|
| 317 | self.Next() # may raise ParseError
|
| 318 |
|
| 319 | if not self.parse_opts.parse_sh_arith():
|
| 320 | # Affects:
|
| 321 | # echo $(( x ))
|
| 322 | # ${a[i]} which should be $[a[i]] -- could have better error
|
| 323 | #
|
| 324 | # Note: sh_expr_eval.UnsafeArith has a dynamic e_die() check
|
| 325 | #
|
| 326 | # Doesn't affect:
|
| 327 | # printf -v x # unsafe_arith.ParseLValue
|
| 328 | # unset x # unsafe_arith.ParseLValue
|
| 329 | # ${!ref} # unsafe_arith.ParseVarRef
|
| 330 | # declare -n # not fully implemented yet
|
| 331 | #
|
| 332 | # a[i+1]= # parse_sh_assign
|
| 333 | #
|
| 334 | # (( a = 1 )) # parse_dparen
|
| 335 | # for (( i = 0; ... # parse_dparen
|
| 336 |
|
| 337 | p_die("POSIX shell arithmetic isn't allowed (parse_sh_arith)",
|
| 338 | loc.Word(self.cur_word))
|
| 339 |
|
| 340 | return self.ParseUntil(0)
|