| 1 | #!/usr/bin/env python2
|
| 2 | # Copyright 2016 Andy Chu. All rights reserved.
|
| 3 | # Licensed under the Apache License, Version 2.0 (the "License");
|
| 4 | # you may not use this file except in compliance with the License.
|
| 5 | # You may obtain a copy of the License at
|
| 6 | #
|
| 7 | # http://www.apache.org/licenses/LICENSE-2.0
|
| 8 | """
|
| 9 | sh_expr_eval.py -- Shell boolean and arithmetic expressions.
|
| 10 | """
|
| 11 | from __future__ import print_function
|
| 12 |
|
| 13 | from _devbuild.gen.id_kind_asdl import Id
|
| 14 | from _devbuild.gen.runtime_asdl import error_code_e, scope_t
|
| 15 | from _devbuild.gen.syntax_asdl import (
|
| 16 | word_t,
|
| 17 | CompoundWord,
|
| 18 | Token,
|
| 19 | loc,
|
| 20 | loc_t,
|
| 21 | source,
|
| 22 | arith_expr,
|
| 23 | arith_expr_e,
|
| 24 | arith_expr_t,
|
| 25 | bool_expr,
|
| 26 | bool_expr_e,
|
| 27 | bool_expr_t,
|
| 28 | sh_lhs,
|
| 29 | sh_lhs_e,
|
| 30 | sh_lhs_t,
|
| 31 | BracedVarSub,
|
| 32 | )
|
| 33 | from _devbuild.gen.option_asdl import option_i
|
| 34 | from _devbuild.gen.types_asdl import bool_arg_type_e
|
| 35 | from _devbuild.gen.value_asdl import (
|
| 36 | value,
|
| 37 | value_e,
|
| 38 | value_t,
|
| 39 | sh_lvalue,
|
| 40 | sh_lvalue_e,
|
| 41 | sh_lvalue_t,
|
| 42 | LeftName,
|
| 43 | eggex_ops,
|
| 44 | regex_match,
|
| 45 | RegexMatch,
|
| 46 | )
|
| 47 | from core import alloc
|
| 48 | from core import bash_impl
|
| 49 | from core import error
|
| 50 | from core.error import e_die, e_die_status, e_strict, e_usage
|
| 51 | from core import num
|
| 52 | from core import state
|
| 53 | from display import ui
|
| 54 | from core import util
|
| 55 | from frontend import consts
|
| 56 | from frontend import lexer
|
| 57 | from frontend import location
|
| 58 | from frontend import match
|
| 59 | from frontend import reader
|
| 60 | from mycpp import mops
|
| 61 | from mycpp import mylib
|
| 62 | from mycpp.mylib import log, tagswitch, switch, str_cmp
|
| 63 | from osh import bool_stat
|
| 64 | from osh import word_eval
|
| 65 |
|
| 66 | import libc # for fnmatch
|
| 67 | # Import these names directly because the C++ translation uses macros literally.
|
| 68 | from libc import FNM_CASEFOLD, REG_ICASE
|
| 69 |
|
| 70 | from typing import Tuple, Optional, cast, TYPE_CHECKING
|
| 71 | if TYPE_CHECKING:
|
| 72 | from core import optview
|
| 73 | from frontend import parse_lib
|
| 74 |
|
| 75 | _ = log
|
| 76 |
|
| 77 | #
|
| 78 | # Arith and Command/Word variants of assignment
|
| 79 | #
|
| 80 | # Calls EvalShellLhs()
|
| 81 | # a[$key]=$val # osh/cmd_eval.py:814 (command_e.ShAssignment)
|
| 82 | # Calls EvalArithLhs()
|
| 83 | # (( a[key] = val )) # osh/sh_expr_eval.py:326 (_EvalLhsArith)
|
| 84 | #
|
| 85 | # Calls OldValue()
|
| 86 | # a[$key]+=$val # osh/cmd_eval.py:795 (assign_op_e.PlusEqual)
|
| 87 | # (( a[key] += val )) # osh/sh_expr_eval.py:308 (_EvalLhsAndLookupArith)
|
| 88 | #
|
| 89 | # RHS Indexing
|
| 90 | # val=${a[$key]} # osh/word_eval.py:639 (bracket_op_e.ArrayIndex)
|
| 91 | # (( val = a[key] )) # osh/sh_expr_eval.py:509 (Id.Arith_LBracket)
|
| 92 | #
|
| 93 |
|
| 94 |
|
| 95 | def OldValue(lval, mem, exec_opts, blame_loc):
|
| 96 | # type: (sh_lvalue_t, state.Mem, Optional[optview.Exec], loc_t) -> value_t
|
| 97 | """Look up for augmented assignment.
|
| 98 |
|
| 99 | For s+=val and (( i += 1 ))
|
| 100 |
|
| 101 | Args:
|
| 102 | lval: value we need to
|
| 103 | exec_opts: can be None if we don't want to check set -u!
|
| 104 | Because s+=val doesn't check it.
|
| 105 |
|
| 106 | TODO: A stricter and less ambiguous version for YSH.
|
| 107 | - Problem: why does sh_lvalue have Indexed and Keyed, while sh_lhs only has
|
| 108 | IndexedName?
|
| 109 | - should I have location.LName and sh_lvalue.Indexed only?
|
| 110 | - and Indexed uses the index_t type?
|
| 111 | - well that might be Str or Int
|
| 112 | """
|
| 113 | assert isinstance(lval, sh_lvalue_t), lval
|
| 114 |
|
| 115 | # TODO: refactor sh_lvalue_t to make this simpler
|
| 116 | UP_lval = lval
|
| 117 | with tagswitch(lval) as case:
|
| 118 | if case(sh_lvalue_e.Var): # (( i++ ))
|
| 119 | lval = cast(LeftName, UP_lval)
|
| 120 | var_name = lval.name
|
| 121 | elif case(sh_lvalue_e.Indexed): # (( a[i]++ ))
|
| 122 | lval = cast(sh_lvalue.Indexed, UP_lval)
|
| 123 | var_name = lval.name
|
| 124 | elif case(sh_lvalue_e.Keyed): # (( A['K']++ )) ? I think this works
|
| 125 | lval = cast(sh_lvalue.Keyed, UP_lval)
|
| 126 | var_name = lval.name
|
| 127 | else:
|
| 128 | raise AssertionError()
|
| 129 |
|
| 130 | cell = mem.GetCellDeref(var_name)
|
| 131 | if cell is not None:
|
| 132 | if cell.readonly:
|
| 133 | e_die("Can't assign to readonly variable %r" % var_name, blame_loc)
|
| 134 | val = cell.val
|
| 135 | else:
|
| 136 | val = value.Undef
|
| 137 |
|
| 138 | if exec_opts and exec_opts.nounset() and val.tag() == value_e.Undef:
|
| 139 | e_die('Undefined variable %r' % var_name, blame_loc)
|
| 140 |
|
| 141 | UP_val = val
|
| 142 | with tagswitch(lval) as case:
|
| 143 | if case(sh_lvalue_e.Var):
|
| 144 | return val
|
| 145 |
|
| 146 | elif case(sh_lvalue_e.Indexed):
|
| 147 | lval = cast(sh_lvalue.Indexed, UP_lval)
|
| 148 |
|
| 149 | with tagswitch(val) as case2:
|
| 150 | if case2(value_e.Undef):
|
| 151 | s = None # type: Optional[str]
|
| 152 | elif case2(value_e.InternalStringArray):
|
| 153 | array_val = cast(value.InternalStringArray, UP_val)
|
| 154 | s, _ = bash_impl.InternalStringArray_GetElement(
|
| 155 | array_val, lval.index)
|
| 156 | # Note: We ignore error_code in the return value of
|
| 157 | # InternalStringArray_GetElement because an invalid index will be
|
| 158 | # reported on the assignment stage anyway.
|
| 159 | elif case2(value_e.BashArray):
|
| 160 | sparse_val = cast(value.BashArray, UP_val)
|
| 161 | s, _ = bash_impl.BashArray_GetElement(
|
| 162 | sparse_val, mops.IntWiden(lval.index))
|
| 163 | else:
|
| 164 | e_die("Can't use [] on value of type %s" % ui.ValType(val))
|
| 165 |
|
| 166 | if s is None:
|
| 167 | val = value.Str('') # NOTE: Other logic is value.Undef? 0?
|
| 168 | else:
|
| 169 | assert isinstance(s, str), s
|
| 170 | val = value.Str(s)
|
| 171 |
|
| 172 | elif case(sh_lvalue_e.Keyed):
|
| 173 | lval = cast(sh_lvalue.Keyed, UP_lval)
|
| 174 |
|
| 175 | assoc_val = None # type: value.BashAssoc
|
| 176 | with tagswitch(val) as case2:
|
| 177 | if case2(value_e.Undef):
|
| 178 | # This never happens, because undef[x]+= is assumed to
|
| 179 | raise AssertionError()
|
| 180 | elif case2(value_e.BashAssoc):
|
| 181 | tmp2 = cast(value.BashAssoc, UP_val)
|
| 182 | # mycpp rewrite: add tmp. cast() creates a new var in inner scope
|
| 183 | assoc_val = tmp2
|
| 184 | s = bash_impl.BashAssoc_GetElement(assoc_val, lval.key)
|
| 185 | else:
|
| 186 | e_die("Can't use [] on value of type %s" % ui.ValType(val))
|
| 187 |
|
| 188 | if s is None:
|
| 189 | val = value.Str('')
|
| 190 | else:
|
| 191 | val = value.Str(s)
|
| 192 |
|
| 193 | else:
|
| 194 | raise AssertionError()
|
| 195 |
|
| 196 | return val
|
| 197 |
|
| 198 |
|
| 199 | # TODO: Should refactor for int/char-based processing
|
| 200 | if mylib.PYTHON:
|
| 201 |
|
| 202 | def IsLower(ch):
|
| 203 | # type: (str) -> bool
|
| 204 | return 'a' <= ch and ch <= 'z'
|
| 205 |
|
| 206 | def IsUpper(ch):
|
| 207 | # type: (str) -> bool
|
| 208 | return 'A' <= ch and ch <= 'Z'
|
| 209 |
|
| 210 |
|
| 211 | class UnsafeArith(object):
|
| 212 | """For parsing a[i] at RUNTIME."""
|
| 213 |
|
| 214 | def __init__(
|
| 215 | self,
|
| 216 | mem, # type: state.Mem
|
| 217 | exec_opts, # type: optview.Exec
|
| 218 | mutable_opts, # type: state.MutableOpts
|
| 219 | parse_ctx, # type: parse_lib.ParseContext
|
| 220 | arith_ev, # type: ArithEvaluator
|
| 221 | errfmt, # type: ui.ErrorFormatter
|
| 222 | ):
|
| 223 | # type: (...) -> None
|
| 224 | self.mem = mem
|
| 225 | self.exec_opts = exec_opts
|
| 226 | self.mutable_opts = mutable_opts
|
| 227 | self.parse_ctx = parse_ctx
|
| 228 | self.arith_ev = arith_ev
|
| 229 | self.errfmt = errfmt
|
| 230 |
|
| 231 | self.arena = self.parse_ctx.arena
|
| 232 |
|
| 233 | def ParseLValue(self, s, location):
|
| 234 | # type: (str, loc_t) -> sh_lvalue_t
|
| 235 | """Parse sh_lvalue for 'unset' and 'printf -v'.
|
| 236 |
|
| 237 | It uses the arith parser, so it behaves like the LHS of (( a[i] = x ))
|
| 238 | """
|
| 239 | if not self.parse_ctx.parse_opts.parse_sh_arith():
|
| 240 | # Do something simpler for YSH
|
| 241 | if not match.IsValidVarName(s):
|
| 242 | e_die('Invalid variable name %r (parse_sh_arith is off)' % s,
|
| 243 | location)
|
| 244 | return LeftName(s, location)
|
| 245 |
|
| 246 | a_parser = self.parse_ctx.MakeArithParser(s)
|
| 247 |
|
| 248 | with alloc.ctx_SourceCode(self.arena,
|
| 249 | source.Dynamic('dynamic LHS', location)):
|
| 250 | try:
|
| 251 | anode = a_parser.Parse()
|
| 252 | except error.Parse as e:
|
| 253 | self.errfmt.PrettyPrintError(e)
|
| 254 | # Exception for builtins 'unset' and 'printf'
|
| 255 | e_usage('got invalid LHS expression', location)
|
| 256 |
|
| 257 | # Note: we parse '1+2', and then it becomes a runtime error because
|
| 258 | # it's not a valid LHS. Could be a parse error.
|
| 259 |
|
| 260 | if self.exec_opts.eval_unsafe_arith():
|
| 261 | lval = self.arith_ev.EvalArithLhs(anode)
|
| 262 | else:
|
| 263 | # Prevent attacks like these by default:
|
| 264 | #
|
| 265 | # unset -v 'A["$(echo K; rm *)"]'
|
| 266 | with state.ctx_Option(self.mutable_opts,
|
| 267 | [option_i._allow_command_sub], False):
|
| 268 | lval = self.arith_ev.EvalArithLhs(anode)
|
| 269 |
|
| 270 | return lval
|
| 271 |
|
| 272 | def ParseVarRef(self, ref_str, blame_tok):
|
| 273 | # type: (str, Token) -> BracedVarSub
|
| 274 | """Parse and evaluate value for ${!ref}
|
| 275 |
|
| 276 | This supports:
|
| 277 | - 0 to 9 for $0 to $9
|
| 278 | - @ for "$@" etc.
|
| 279 |
|
| 280 | See grammar in osh/word_parse.py, which is related to grammar in
|
| 281 | osh/word_parse.py _ReadBracedVarSub
|
| 282 |
|
| 283 | Note: declare -n allows 'varname' and 'varname[i]' and 'varname[@]', but it
|
| 284 | does NOT allow 0 to 9, @, *
|
| 285 |
|
| 286 | NamerefExpr = NAME Subscript? # this allows @ and * too
|
| 287 |
|
| 288 | _ResolveNameOrRef currently gives you a 'cell'. So it might not support
|
| 289 | sh_lvalue.Indexed?
|
| 290 | """
|
| 291 | line_reader = reader.StringLineReader(ref_str, self.arena)
|
| 292 | lexer = self.parse_ctx.MakeLexer(line_reader)
|
| 293 | w_parser = self.parse_ctx.MakeWordParser(lexer, line_reader)
|
| 294 |
|
| 295 | src = source.VarRef(blame_tok)
|
| 296 | with alloc.ctx_SourceCode(self.arena, src):
|
| 297 | try:
|
| 298 | bvs_part = w_parser.ParseVarRef()
|
| 299 | except error.Parse as e:
|
| 300 | # This prints the inner location
|
| 301 | self.errfmt.PrettyPrintError(e)
|
| 302 |
|
| 303 | # this affects builtins 'unset' and 'printf'
|
| 304 | e_die("Invalid var ref expression", blame_tok)
|
| 305 |
|
| 306 | return bvs_part
|
| 307 |
|
| 308 |
|
| 309 | def _ParseOshInteger(s, blame_loc):
|
| 310 | # type: (str, loc_t) -> Tuple[bool, mops.BigInt]
|
| 311 | """
|
| 312 | Returns:
|
| 313 | (True, value) when the string looks like an integer
|
| 314 | (False, ...) when it doesn't
|
| 315 |
|
| 316 | Integer formats that are recognized:
|
| 317 | 0xAB hex
|
| 318 | 042 octal
|
| 319 | 42 decimal
|
| 320 | 64#z arbitrary base
|
| 321 | """
|
| 322 | id_, pos = match.MatchShNumberToken(s, 0) # use re2c lexer
|
| 323 | if pos != len(s):
|
| 324 | # trailing data isn't allowed
|
| 325 | return (False, mops.BigInt(0))
|
| 326 |
|
| 327 | # Do conversions
|
| 328 |
|
| 329 | if id_ == Id.ShNumber_Dec:
|
| 330 | # Normal base 10 integer.
|
| 331 | ok, big_int = mops.FromStr2(s)
|
| 332 | if not ok:
|
| 333 | e_die('Integer too big: %s' % s, blame_loc)
|
| 334 | return (True, big_int)
|
| 335 |
|
| 336 | elif id_ == Id.ShNumber_Oct:
|
| 337 | # 0123, offset by 1
|
| 338 | ok, big_int = mops.FromStr2(s[1:], 8)
|
| 339 | if not ok:
|
| 340 | e_die('Octal integer too big: %s' % s, blame_loc)
|
| 341 | return (True, big_int)
|
| 342 |
|
| 343 | elif id_ == Id.ShNumber_Hex:
|
| 344 | # 0xff, offset by 2
|
| 345 | ok, big_int = mops.FromStr2(s[2:], 16)
|
| 346 | if not ok:
|
| 347 | e_die('Hex integer too big: %s' % s, blame_loc)
|
| 348 | return (True, big_int)
|
| 349 |
|
| 350 | elif id_ == Id.ShNumber_BaseN:
|
| 351 | b, digits = mylib.split_once(s, '#')
|
| 352 | assert digits is not None, digits # assured by lexer
|
| 353 |
|
| 354 | try:
|
| 355 | base = int(b) # machine integer, not BigInt
|
| 356 | except ValueError:
|
| 357 | # Unreachable per the regex validation above
|
| 358 | raise AssertionError()
|
| 359 |
|
| 360 | if base > 64:
|
| 361 | e_strict('Base %d cannot be larger than 64' % base, blame_loc)
|
| 362 | if base < 2:
|
| 363 | e_strict('Base %d must be larger than 2' % base, blame_loc)
|
| 364 |
|
| 365 | integer = mops.ZERO
|
| 366 | for ch in digits:
|
| 367 | if IsLower(ch):
|
| 368 | digit = ord(ch) - ord('a') + 10
|
| 369 | elif IsUpper(ch):
|
| 370 | digit = ord(ch) - ord('A') + 36
|
| 371 | elif ch == '@': # horrible syntax
|
| 372 | digit = 62
|
| 373 | elif ch == '_':
|
| 374 | digit = 63
|
| 375 | elif ch.isdigit():
|
| 376 | digit = int(ch)
|
| 377 | else:
|
| 378 | # Unreachable per the regex validation above
|
| 379 | raise AssertionError()
|
| 380 |
|
| 381 | if digit >= base:
|
| 382 | e_strict('Digits %r out of range for base %d' % (digits, base),
|
| 383 | blame_loc)
|
| 384 |
|
| 385 | # formula is:
|
| 386 | # integer = integer * base + digit
|
| 387 | integer = mops.Add(mops.Mul(integer, mops.IntWiden(base)),
|
| 388 | mops.IntWiden(digit))
|
| 389 | return (True, integer)
|
| 390 |
|
| 391 | else:
|
| 392 | # Id.Unknown_Tok or Id.Eol_Tok
|
| 393 | return (False, mops.BigInt(0)) # not an integer
|
| 394 |
|
| 395 |
|
| 396 | class ArithEvaluator(bash_impl.ArrayIndexEvaluator):
|
| 397 | """Shared between arith and bool evaluators.
|
| 398 |
|
| 399 | They both:
|
| 400 |
|
| 401 | 1. Convert strings to integers, respecting shopt -s strict_arith.
|
| 402 | 2. Look up variables and evaluate words.
|
| 403 | """
|
| 404 |
|
| 405 | def __init__(
|
| 406 | self,
|
| 407 | mem, # type: state.Mem
|
| 408 | exec_opts, # type: optview.Exec
|
| 409 | mutable_opts, # type: state.MutableOpts
|
| 410 | parse_ctx, # type: Optional[parse_lib.ParseContext]
|
| 411 | errfmt, # type: ui.ErrorFormatter
|
| 412 | ):
|
| 413 | # type: (...) -> None
|
| 414 | bash_impl.ArrayIndexEvaluator.__init__(self)
|
| 415 | self.word_ev = None # type: word_eval.StringWordEvaluator
|
| 416 | self.mem = mem
|
| 417 | self.exec_opts = exec_opts
|
| 418 | self.mutable_opts = mutable_opts
|
| 419 | self.parse_ctx = parse_ctx
|
| 420 | self.errfmt = errfmt
|
| 421 |
|
| 422 | def CheckCircularDeps(self):
|
| 423 | # type: () -> None
|
| 424 | assert self.word_ev is not None
|
| 425 |
|
| 426 | def StringToBigInt(self, s, blame_loc):
|
| 427 | # type: (str, loc_t) -> mops.BigInt
|
| 428 | """Use bash-like rules to coerce a string to an integer.
|
| 429 |
|
| 430 | Runtime parsing enables silly stuff like $(( $(echo 1)$(echo 2) + 1 )) => 13
|
| 431 |
|
| 432 | bare word: variable
|
| 433 | quoted word: string (not done?)
|
| 434 | """
|
| 435 | s = s.strip()
|
| 436 |
|
| 437 | ok, i = _ParseOshInteger(s, blame_loc)
|
| 438 | if ok:
|
| 439 | return i
|
| 440 |
|
| 441 | # Doesn't look like an integer
|
| 442 |
|
| 443 | # note: 'test' and '[' never evaluate recursively
|
| 444 | if self.parse_ctx is None:
|
| 445 | if len(s) == 0 or match.IsValidVarName(s):
|
| 446 | # x42 could evaluate to 0
|
| 447 | e_strict("Invalid integer constant %r" % s, blame_loc)
|
| 448 | else:
|
| 449 | # 42x is always fatal!
|
| 450 | e_die("Invalid integer constant %r" % s, blame_loc)
|
| 451 |
|
| 452 | # Special case so we don't get EOF error
|
| 453 | if len(s) == 0:
|
| 454 | return mops.ZERO
|
| 455 |
|
| 456 | # For compatibility: Try to parse it as an expression and evaluate it.
|
| 457 | a_parser = self.parse_ctx.MakeArithParser(s)
|
| 458 |
|
| 459 | try:
|
| 460 | node2 = a_parser.Parse() # may raise error.Parse
|
| 461 | except error.Parse as e:
|
| 462 | self.errfmt.PrettyPrintError(e)
|
| 463 | e_die('Parse error in recursive arithmetic', e.location)
|
| 464 |
|
| 465 | # Prevent infinite recursion of $(( 1x )) -- it's a word that evaluates
|
| 466 | # to itself, and you don't want to reparse it as a word.
|
| 467 | if node2.tag() == arith_expr_e.Word:
|
| 468 | e_die("Invalid integer constant %r" % s, blame_loc)
|
| 469 |
|
| 470 | if self.exec_opts.eval_unsafe_arith():
|
| 471 | integer = self.EvalToBigInt(node2)
|
| 472 | else:
|
| 473 | # BoolEvaluator doesn't have parse_ctx or mutable_opts
|
| 474 | assert self.mutable_opts is not None
|
| 475 |
|
| 476 | # We don't need to flip _allow_process_sub, because they can't be
|
| 477 | # parsed. See spec/bugs.test.sh.
|
| 478 | with state.ctx_Option(self.mutable_opts,
|
| 479 | [option_i._allow_command_sub], False):
|
| 480 | integer = self.EvalToBigInt(node2)
|
| 481 |
|
| 482 | return integer
|
| 483 |
|
| 484 | def _ValToIntOrError(self, val, blame):
|
| 485 | # type: (value_t, arith_expr_t) -> mops.BigInt
|
| 486 | try:
|
| 487 | UP_val = val
|
| 488 | with tagswitch(val) as case:
|
| 489 | if case(value_e.Undef):
|
| 490 | # 'nounset' already handled before got here
|
| 491 | # Happens upon a[undefined]=42, which unfortunately turns into a[0]=42.
|
| 492 | e_strict('Undefined value in arithmetic context',
|
| 493 | loc.Arith(blame))
|
| 494 |
|
| 495 | elif case(value_e.Int):
|
| 496 | val = cast(value.Int, UP_val)
|
| 497 | return val.i
|
| 498 |
|
| 499 | elif case(value_e.Str):
|
| 500 | val = cast(value.Str, UP_val)
|
| 501 | # calls e_strict
|
| 502 | return self.StringToBigInt(val.s, loc.Arith(blame))
|
| 503 |
|
| 504 | except error.Strict as e:
|
| 505 | if self.exec_opts.strict_arith():
|
| 506 | raise
|
| 507 | else:
|
| 508 | return mops.ZERO
|
| 509 |
|
| 510 | # Arrays and associative arrays always fail -- not controlled by
|
| 511 | # strict_arith.
|
| 512 | # In bash, (( a )) is like (( a[0] )), but I don't want that.
|
| 513 | # And returning '0' gives different results.
|
| 514 | e_die(
|
| 515 | "Expected a value convertible to integer, got %s" %
|
| 516 | ui.ValType(val), loc.Arith(blame))
|
| 517 |
|
| 518 | def _EvalLhsAndLookupArith(self, node):
|
| 519 | # type: (arith_expr_t) -> Tuple[mops.BigInt, sh_lvalue_t]
|
| 520 | """ For x = y and x += y and ++x """
|
| 521 |
|
| 522 | lval = self.EvalArithLhs(node)
|
| 523 | val = OldValue(lval, self.mem, self.exec_opts,
|
| 524 | location.TokenForArith(node))
|
| 525 |
|
| 526 | # BASH_LINENO, arr (array name without strict_array), etc.
|
| 527 | if (val.tag() in (value_e.InternalStringArray, value_e.BashAssoc,
|
| 528 | value_e.BashArray) and
|
| 529 | lval.tag() == sh_lvalue_e.Var):
|
| 530 | named_lval = cast(LeftName, lval)
|
| 531 | if word_eval.ShouldArrayDecay(named_lval.name, self.exec_opts):
|
| 532 | if val.tag() in (value_e.InternalStringArray,
|
| 533 | value_e.BashArray):
|
| 534 | lval = sh_lvalue.Indexed(named_lval.name, 0, loc.Missing)
|
| 535 | elif val.tag() == value_e.BashAssoc:
|
| 536 | lval = sh_lvalue.Keyed(named_lval.name, '0', loc.Missing)
|
| 537 | val = word_eval.DecayArray(val)
|
| 538 |
|
| 539 | # This error message could be better, but we already have one
|
| 540 | #if val.tag() in (value_e.InternalStringArray, value_e.BashArray):
|
| 541 | # e_die("Can't use assignment like ++ or += on arrays")
|
| 542 |
|
| 543 | i = self._ValToIntOrError(val, node)
|
| 544 | return i, lval
|
| 545 |
|
| 546 | def _Store(self, lval, new_int):
|
| 547 | # type: (sh_lvalue_t, mops.BigInt) -> None
|
| 548 | val = value.Str(mops.ToStr(new_int))
|
| 549 | state.OshLanguageSetValue(self.mem, lval, val)
|
| 550 |
|
| 551 | def EvalToBigInt(self, node):
|
| 552 | # type: (arith_expr_t) -> mops.BigInt
|
| 553 | """Used externally by ${a[i+1]} and ${a:start:len}.
|
| 554 |
|
| 555 | Also used internally.
|
| 556 | """
|
| 557 | val = self.Eval(node)
|
| 558 |
|
| 559 | # BASH_LINENO, arr (array name without strict_array), etc.
|
| 560 | if (val.tag() in (value_e.InternalStringArray, value_e.BashAssoc,
|
| 561 | value_e.BashArray) and
|
| 562 | node.tag() == arith_expr_e.VarSub):
|
| 563 | vsub = cast(Token, node)
|
| 564 | if word_eval.ShouldArrayDecay(lexer.LazyStr(vsub), self.exec_opts):
|
| 565 | val = word_eval.DecayArray(val)
|
| 566 |
|
| 567 | i = self._ValToIntOrError(val, node)
|
| 568 | return i
|
| 569 |
|
| 570 | def EvalToInt(self, node):
|
| 571 | # type: (arith_expr_t) -> int
|
| 572 | return mops.BigTruncate(self.EvalToBigInt(node))
|
| 573 |
|
| 574 | def Eval(self, node):
|
| 575 | # type: (arith_expr_t) -> value_t
|
| 576 | """
|
| 577 | Returns:
|
| 578 | None for Undef (e.g. empty cell) TODO: Don't return 0!
|
| 579 | int for Str
|
| 580 | List[int] for InternalStringArray and BashArray
|
| 581 | Dict[str, str] for BashAssoc (TODO: Should we support this?)
|
| 582 |
|
| 583 | NOTE: (( A['x'] = 'x' )) and (( x = A['x'] )) are syntactically valid in
|
| 584 | bash, but don't do what you'd think. 'x' sometimes a variable name and
|
| 585 | sometimes a key.
|
| 586 | """
|
| 587 | # OSH semantics: Variable NAMES cannot be formed dynamically; but INTEGERS
|
| 588 | # can. ${foo:-3}4 is OK. $? will be a compound word too, so we don't have
|
| 589 | # to handle that as a special case.
|
| 590 |
|
| 591 | UP_node = node
|
| 592 | with tagswitch(node) as case:
|
| 593 | if case(arith_expr_e.EmptyZero): # $(( ))
|
| 594 | return value.Int(mops.ZERO) # Weird axiom
|
| 595 |
|
| 596 | elif case(arith_expr_e.EmptyOne): # for (( ; ; ))
|
| 597 | return value.Int(mops.ONE)
|
| 598 |
|
| 599 | elif case(arith_expr_e.VarSub): # $(( x )) (can be array)
|
| 600 | vsub = cast(Token, UP_node)
|
| 601 | var_name = lexer.LazyStr(vsub)
|
| 602 | val = self.mem.GetValue(var_name)
|
| 603 | if val.tag() == value_e.Undef and self.exec_opts.nounset():
|
| 604 | e_die('Undefined variable %r' % var_name, vsub)
|
| 605 | return val
|
| 606 |
|
| 607 | elif case(arith_expr_e.Word): # $(( $x )) $(( ${x}${y} )), etc.
|
| 608 | w = cast(CompoundWord, UP_node)
|
| 609 | return self.word_ev.EvalWordToString(w)
|
| 610 |
|
| 611 | elif case(arith_expr_e.UnaryAssign): # a++
|
| 612 | node = cast(arith_expr.UnaryAssign, UP_node)
|
| 613 |
|
| 614 | op_id = node.op_id
|
| 615 | old_big, lval = self._EvalLhsAndLookupArith(node.child)
|
| 616 |
|
| 617 | if op_id == Id.Node_PostDPlus: # post-increment
|
| 618 | new_big = mops.Add(old_big, mops.ONE)
|
| 619 | result = old_big
|
| 620 |
|
| 621 | elif op_id == Id.Node_PostDMinus: # post-decrement
|
| 622 | new_big = mops.Sub(old_big, mops.ONE)
|
| 623 | result = old_big
|
| 624 |
|
| 625 | elif op_id == Id.Arith_DPlus: # pre-increment
|
| 626 | new_big = mops.Add(old_big, mops.ONE)
|
| 627 | result = new_big
|
| 628 |
|
| 629 | elif op_id == Id.Arith_DMinus: # pre-decrement
|
| 630 | new_big = mops.Sub(old_big, mops.ONE)
|
| 631 | result = new_big
|
| 632 |
|
| 633 | else:
|
| 634 | raise AssertionError(op_id)
|
| 635 |
|
| 636 | self._Store(lval, new_big)
|
| 637 | return value.Int(result)
|
| 638 |
|
| 639 | elif case(arith_expr_e.BinaryAssign): # a=1, a+=5, a[1]+=5
|
| 640 | node = cast(arith_expr.BinaryAssign, UP_node)
|
| 641 | op_id = node.op_id
|
| 642 |
|
| 643 | if op_id == Id.Arith_Equal:
|
| 644 | # Don't really need a span ID here, because tdop.CheckLhsExpr should
|
| 645 | # have done all the validation.
|
| 646 | lval = self.EvalArithLhs(node.left)
|
| 647 | rhs_big = self.EvalToBigInt(node.right)
|
| 648 |
|
| 649 | self._Store(lval, rhs_big)
|
| 650 | return value.Int(rhs_big)
|
| 651 |
|
| 652 | old_big, lval = self._EvalLhsAndLookupArith(node.left)
|
| 653 | rhs_big = self.EvalToBigInt(node.right)
|
| 654 |
|
| 655 | if op_id == Id.Arith_PlusEqual:
|
| 656 | new_big = mops.Add(old_big, rhs_big)
|
| 657 | elif op_id == Id.Arith_MinusEqual:
|
| 658 | new_big = mops.Sub(old_big, rhs_big)
|
| 659 | elif op_id == Id.Arith_StarEqual:
|
| 660 | new_big = mops.Mul(old_big, rhs_big)
|
| 661 |
|
| 662 | elif op_id == Id.Arith_SlashEqual:
|
| 663 | if mops.Equal(rhs_big, mops.ZERO):
|
| 664 | e_die('Divide by zero') # TODO: location
|
| 665 | new_big = mops.Div(old_big, rhs_big)
|
| 666 |
|
| 667 | elif op_id == Id.Arith_PercentEqual:
|
| 668 | if mops.Equal(rhs_big, mops.ZERO):
|
| 669 | e_die('Divide by zero') # TODO: location
|
| 670 | new_big = mops.Rem(old_big, rhs_big)
|
| 671 |
|
| 672 | elif op_id == Id.Arith_DGreatEqual:
|
| 673 | new_big = mops.RShift(old_big, rhs_big)
|
| 674 | elif op_id == Id.Arith_DLessEqual:
|
| 675 | new_big = mops.LShift(old_big, rhs_big)
|
| 676 | elif op_id == Id.Arith_AmpEqual:
|
| 677 | new_big = mops.BitAnd(old_big, rhs_big)
|
| 678 | elif op_id == Id.Arith_PipeEqual:
|
| 679 | new_big = mops.BitOr(old_big, rhs_big)
|
| 680 | elif op_id == Id.Arith_CaretEqual:
|
| 681 | new_big = mops.BitXor(old_big, rhs_big)
|
| 682 | else:
|
| 683 | raise AssertionError(op_id) # shouldn't get here
|
| 684 |
|
| 685 | self._Store(lval, new_big)
|
| 686 | return value.Int(new_big)
|
| 687 |
|
| 688 | elif case(arith_expr_e.Unary):
|
| 689 | node = cast(arith_expr.Unary, UP_node)
|
| 690 | op_id = node.op_id
|
| 691 |
|
| 692 | i = self.EvalToBigInt(node.child)
|
| 693 |
|
| 694 | if op_id == Id.Node_UnaryPlus: # +i
|
| 695 | result = i
|
| 696 | elif op_id == Id.Node_UnaryMinus: # -i
|
| 697 | result = mops.Sub(mops.ZERO, i)
|
| 698 |
|
| 699 | elif op_id == Id.Arith_Bang: # logical negation
|
| 700 | if mops.Equal(i, mops.ZERO):
|
| 701 | result = mops.ONE
|
| 702 | else:
|
| 703 | result = mops.ZERO
|
| 704 | elif op_id == Id.Arith_Tilde: # bitwise complement
|
| 705 | result = mops.BitNot(i)
|
| 706 | else:
|
| 707 | raise AssertionError(op_id) # shouldn't get here
|
| 708 |
|
| 709 | return value.Int(result)
|
| 710 |
|
| 711 | elif case(arith_expr_e.Binary):
|
| 712 | node = cast(arith_expr.Binary, UP_node)
|
| 713 | op_id = node.op.id
|
| 714 |
|
| 715 | # Short-circuit evaluation for || and &&.
|
| 716 | if op_id == Id.Arith_DPipe:
|
| 717 | lhs_big = self.EvalToBigInt(node.left)
|
| 718 | if mops.Equal(lhs_big, mops.ZERO):
|
| 719 | rhs_big = self.EvalToBigInt(node.right)
|
| 720 | if mops.Equal(rhs_big, mops.ZERO):
|
| 721 | result = mops.ZERO # false
|
| 722 | else:
|
| 723 | result = mops.ONE # true
|
| 724 | else:
|
| 725 | result = mops.ONE # true
|
| 726 | return value.Int(result)
|
| 727 |
|
| 728 | if op_id == Id.Arith_DAmp:
|
| 729 | lhs_big = self.EvalToBigInt(node.left)
|
| 730 | if mops.Equal(lhs_big, mops.ZERO):
|
| 731 | result = mops.ZERO # false
|
| 732 | else:
|
| 733 | rhs_big = self.EvalToBigInt(node.right)
|
| 734 | if mops.Equal(rhs_big, mops.ZERO):
|
| 735 | result = mops.ZERO # false
|
| 736 | else:
|
| 737 | result = mops.ONE # true
|
| 738 | return value.Int(result)
|
| 739 |
|
| 740 | if op_id == Id.Arith_LBracket:
|
| 741 | # NOTE: Similar to bracket_op_e.ArrayIndex in osh/word_eval.py
|
| 742 |
|
| 743 | left = self.Eval(node.left)
|
| 744 | UP_left = left
|
| 745 | with tagswitch(left) as case:
|
| 746 | if case(value_e.InternalStringArray):
|
| 747 | array_val = cast(value.InternalStringArray,
|
| 748 | UP_left)
|
| 749 | small_i = mops.BigTruncate(
|
| 750 | self.EvalToBigInt(node.right))
|
| 751 | s, error_code = bash_impl.InternalStringArray_GetElement(
|
| 752 | array_val, small_i)
|
| 753 | if error_code == error_code_e.IndexOutOfRange:
|
| 754 | # Note: Bash outputs warning but does not make
|
| 755 | # it a real error. We follow the Bash behavior
|
| 756 | # here.
|
| 757 | small_length = bash_impl.InternalStringArray_Length(
|
| 758 | array_val)
|
| 759 | self.errfmt.Print_(
|
| 760 | "Index %d out of bounds for array of length %d"
|
| 761 | % (small_i, small_length),
|
| 762 | blame_loc=node.op)
|
| 763 |
|
| 764 | elif case(value_e.BashArray):
|
| 765 | sparse_val = cast(value.BashArray, UP_left)
|
| 766 | i = self.EvalToBigInt(node.right)
|
| 767 | s, error_code = bash_impl.BashArray_GetElement(
|
| 768 | sparse_val, i)
|
| 769 | if error_code == error_code_e.IndexOutOfRange:
|
| 770 | # Note: Bash outputs warning but does not make
|
| 771 | # it a real error. We follow the Bash behavior
|
| 772 | # here.
|
| 773 | length = bash_impl.BashArray_Length(sparse_val)
|
| 774 | self.errfmt.Print_(
|
| 775 | "Index %s out of bounds for array of length %s"
|
| 776 | % (mops.ToStr(i), mops.ToStr(length)),
|
| 777 | blame_loc=node.op)
|
| 778 |
|
| 779 | elif case(value_e.BashAssoc):
|
| 780 | left = cast(value.BashAssoc, UP_left)
|
| 781 | key = self.EvalWordToString(node.right)
|
| 782 | s = bash_impl.BashAssoc_GetElement(left, key)
|
| 783 |
|
| 784 | elif case(value_e.Str):
|
| 785 | left = cast(value.Str, UP_left)
|
| 786 | if self.exec_opts.strict_arith():
|
| 787 | e_die(
|
| 788 | "Value of type Str can't be indexed (strict_arith)",
|
| 789 | node.op)
|
| 790 | index = self.EvalToBigInt(node.right)
|
| 791 | # s[0] evaluates to s
|
| 792 | # s[1] evaluates to Undef
|
| 793 | s = left.s if mops.Equal(index,
|
| 794 | mops.ZERO) else None
|
| 795 |
|
| 796 | elif case(value_e.Undef):
|
| 797 | if self.exec_opts.strict_arith():
|
| 798 | e_die(
|
| 799 | "Value of type Undef can't be indexed (strict_arith)",
|
| 800 | node.op)
|
| 801 | s = None # value.Undef
|
| 802 |
|
| 803 | # There isn't a way to distinguish Undef vs. empty
|
| 804 | # string, even with set -o nounset?
|
| 805 | # s = ''
|
| 806 |
|
| 807 | else:
|
| 808 | # TODO: Add error context
|
| 809 | e_die(
|
| 810 | "Value of type %s can't be indexed" %
|
| 811 | ui.ValType(left), node.op)
|
| 812 |
|
| 813 | if s is None:
|
| 814 | val = value.Undef
|
| 815 | else:
|
| 816 | val = value.Str(s)
|
| 817 |
|
| 818 | return val
|
| 819 |
|
| 820 | if op_id == Id.Arith_Comma:
|
| 821 | self.EvalToBigInt(node.left) # throw away result
|
| 822 | result = self.EvalToBigInt(node.right)
|
| 823 | return value.Int(result)
|
| 824 |
|
| 825 | # Rest are integers
|
| 826 | lhs_big = self.EvalToBigInt(node.left)
|
| 827 | rhs_big = self.EvalToBigInt(node.right)
|
| 828 |
|
| 829 | if op_id == Id.Arith_Plus:
|
| 830 | result = mops.Add(lhs_big, rhs_big)
|
| 831 | elif op_id == Id.Arith_Minus:
|
| 832 | result = mops.Sub(lhs_big, rhs_big)
|
| 833 | elif op_id == Id.Arith_Star:
|
| 834 | result = mops.Mul(lhs_big, rhs_big)
|
| 835 | elif op_id == Id.Arith_Slash:
|
| 836 | if mops.Equal(rhs_big, mops.ZERO):
|
| 837 | e_die('Divide by zero', node.op)
|
| 838 | result = mops.Div(lhs_big, rhs_big)
|
| 839 |
|
| 840 | elif op_id == Id.Arith_Percent:
|
| 841 | if mops.Equal(rhs_big, mops.ZERO):
|
| 842 | e_die('Divide by zero', node.op)
|
| 843 | result = mops.Rem(lhs_big, rhs_big)
|
| 844 |
|
| 845 | elif op_id == Id.Arith_DStar:
|
| 846 | if mops.Greater(mops.ZERO, rhs_big):
|
| 847 | e_die("Exponent can't be a negative number",
|
| 848 | loc.Arith(node.right))
|
| 849 | result = num.Exponent(lhs_big, rhs_big)
|
| 850 |
|
| 851 | elif op_id == Id.Arith_DEqual:
|
| 852 | result = mops.FromBool(mops.Equal(lhs_big, rhs_big))
|
| 853 | elif op_id == Id.Arith_NEqual:
|
| 854 | result = mops.FromBool(not mops.Equal(lhs_big, rhs_big))
|
| 855 | elif op_id == Id.Arith_Great:
|
| 856 | result = mops.FromBool(mops.Greater(lhs_big, rhs_big))
|
| 857 | elif op_id == Id.Arith_GreatEqual:
|
| 858 | result = mops.FromBool(
|
| 859 | mops.Greater(lhs_big, rhs_big) or
|
| 860 | mops.Equal(lhs_big, rhs_big))
|
| 861 | elif op_id == Id.Arith_Less:
|
| 862 | result = mops.FromBool(mops.Greater(rhs_big, lhs_big))
|
| 863 | elif op_id == Id.Arith_LessEqual:
|
| 864 | result = mops.FromBool(
|
| 865 | mops.Greater(rhs_big, lhs_big) or
|
| 866 | mops.Equal(lhs_big, rhs_big))
|
| 867 |
|
| 868 | elif op_id == Id.Arith_Pipe:
|
| 869 | result = mops.BitOr(lhs_big, rhs_big)
|
| 870 | elif op_id == Id.Arith_Amp:
|
| 871 | result = mops.BitAnd(lhs_big, rhs_big)
|
| 872 | elif op_id == Id.Arith_Caret:
|
| 873 | result = mops.BitXor(lhs_big, rhs_big)
|
| 874 |
|
| 875 | # Note: how to define shift of negative numbers?
|
| 876 | elif op_id == Id.Arith_DLess:
|
| 877 | if mops.Greater(mops.ZERO, rhs_big): # rhs_big < 0
|
| 878 | raise error.Expr("Can't left shift by negative number",
|
| 879 | node.op)
|
| 880 | result = mops.LShift(lhs_big, rhs_big)
|
| 881 | elif op_id == Id.Arith_DGreat:
|
| 882 | if mops.Greater(mops.ZERO, rhs_big): # rhs_big < 0
|
| 883 | raise error.Expr(
|
| 884 | "Can't right shift by negative number", node.op)
|
| 885 | result = mops.RShift(lhs_big, rhs_big)
|
| 886 | else:
|
| 887 | raise AssertionError(op_id)
|
| 888 |
|
| 889 | return value.Int(result)
|
| 890 |
|
| 891 | elif case(arith_expr_e.TernaryOp):
|
| 892 | node = cast(arith_expr.TernaryOp, UP_node)
|
| 893 |
|
| 894 | cond = self.EvalToBigInt(node.cond)
|
| 895 | if mops.Equal(cond, mops.ZERO):
|
| 896 | return self.Eval(node.false_expr)
|
| 897 | else:
|
| 898 | return self.Eval(node.true_expr)
|
| 899 |
|
| 900 | else:
|
| 901 | raise AssertionError(node.tag())
|
| 902 |
|
| 903 | raise AssertionError('for -Wreturn-type in C++')
|
| 904 |
|
| 905 | def EvalWordToString(self, node, blame_loc=loc.Missing):
|
| 906 | # type: (arith_expr_t, loc_t) -> str
|
| 907 | """
|
| 908 | Raises:
|
| 909 | error.FatalRuntime if the expression isn't a string
|
| 910 | or if it contains a bare variable like a[x]
|
| 911 |
|
| 912 | These are allowed because they're unambiguous, unlike a[x]
|
| 913 |
|
| 914 | a[$x] a["$x"] a["x"] a['x']
|
| 915 | """
|
| 916 | UP_node = node
|
| 917 | if node.tag() == arith_expr_e.Word: # $(( $x )) $(( ${x}${y} )), etc.
|
| 918 | w = cast(CompoundWord, UP_node)
|
| 919 | val = self.word_ev.EvalWordToString(w)
|
| 920 | return val.s
|
| 921 | else:
|
| 922 | # A[x] is the "Parsing Bash is Undecidable" problem
|
| 923 | # It is a string or var name?
|
| 924 | # (It's parsed as arith_expr.VarSub)
|
| 925 | e_die(
|
| 926 | "Assoc array keys must be strings: $x 'x' \"$x\" etc. (OILS-ERR-101)",
|
| 927 | blame_loc)
|
| 928 |
|
| 929 | def EvalShellLhs(self, node, which_scopes):
|
| 930 | # type: (sh_lhs_t, scope_t) -> sh_lvalue_t
|
| 931 | """Evaluate a shell LHS expression
|
| 932 |
|
| 933 | For a=b and a[x]=b etc.
|
| 934 | """
|
| 935 | assert isinstance(node, sh_lhs_t), node
|
| 936 |
|
| 937 | UP_node = node
|
| 938 | lval = None # type: sh_lvalue_t
|
| 939 | with tagswitch(node) as case:
|
| 940 | if case(sh_lhs_e.Name): # a=x
|
| 941 | node = cast(sh_lhs.Name, UP_node)
|
| 942 | assert node.name is not None
|
| 943 |
|
| 944 | lval1 = LeftName(node.name, node.left)
|
| 945 | lval = lval1
|
| 946 |
|
| 947 | elif case(sh_lhs_e.IndexedName): # a[1+2]=x
|
| 948 | node = cast(sh_lhs.IndexedName, UP_node)
|
| 949 | assert node.name is not None
|
| 950 |
|
| 951 | if self.mem.IsBashAssoc(node.name):
|
| 952 | key = self.EvalWordToString(node.index,
|
| 953 | blame_loc=node.left)
|
| 954 | # node.left points to A[ in A[x]=1
|
| 955 | lval2 = sh_lvalue.Keyed(node.name, key, node.left)
|
| 956 | lval = lval2
|
| 957 | else:
|
| 958 | index = mops.BigTruncate(self.EvalToBigInt(node.index))
|
| 959 | lval3 = sh_lvalue.Indexed(node.name, index, node.left)
|
| 960 | lval = lval3
|
| 961 |
|
| 962 | else:
|
| 963 | raise AssertionError(node.tag())
|
| 964 |
|
| 965 | return lval
|
| 966 |
|
| 967 | def _VarNameOrWord(self, anode):
|
| 968 | # type: (arith_expr_t) -> Tuple[Optional[str], loc_t]
|
| 969 | """
|
| 970 | Returns a variable name if the arith node can be interpreted that way.
|
| 971 | """
|
| 972 | UP_anode = anode
|
| 973 | with tagswitch(anode) as case:
|
| 974 | if case(arith_expr_e.VarSub):
|
| 975 | tok = cast(Token, UP_anode)
|
| 976 | return (lexer.LazyStr(tok), tok)
|
| 977 |
|
| 978 | elif case(arith_expr_e.Word):
|
| 979 | w = cast(CompoundWord, UP_anode)
|
| 980 | var_name = self.EvalWordToString(w)
|
| 981 | return (var_name, w)
|
| 982 |
|
| 983 | no_str = None # type: Optional[str]
|
| 984 | return (no_str, loc.Missing)
|
| 985 |
|
| 986 | def EvalArithLhs(self, anode):
|
| 987 | # type: (arith_expr_t) -> sh_lvalue_t
|
| 988 | """
|
| 989 | For (( a[x] = 1 )) etc.
|
| 990 | """
|
| 991 | UP_anode = anode
|
| 992 | if anode.tag() == arith_expr_e.Binary:
|
| 993 | anode = cast(arith_expr.Binary, UP_anode)
|
| 994 | if anode.op.id == Id.Arith_LBracket:
|
| 995 | var_name, blame_loc = self._VarNameOrWord(anode.left)
|
| 996 |
|
| 997 | # (( 1[2] = 3 )) isn't valid
|
| 998 | if not match.IsValidVarName(var_name):
|
| 999 | e_die('Invalid variable name %r' % var_name, blame_loc)
|
| 1000 |
|
| 1001 | if var_name is not None:
|
| 1002 | if self.mem.IsBashAssoc(var_name):
|
| 1003 | arith_loc = location.TokenForArith(anode)
|
| 1004 | key = self.EvalWordToString(anode.right,
|
| 1005 | blame_loc=arith_loc)
|
| 1006 | return sh_lvalue.Keyed(var_name, key, blame_loc)
|
| 1007 | else:
|
| 1008 | index = mops.BigTruncate(self.EvalToBigInt(
|
| 1009 | anode.right))
|
| 1010 | return sh_lvalue.Indexed(var_name, index, blame_loc)
|
| 1011 |
|
| 1012 | var_name, blame_loc = self._VarNameOrWord(anode)
|
| 1013 | if var_name is not None:
|
| 1014 | return LeftName(var_name, blame_loc)
|
| 1015 |
|
| 1016 | # e.g. unset 'x-y'. status 2 for runtime parse error
|
| 1017 | e_die_status(2, 'Invalid LHS to modify', blame_loc)
|
| 1018 |
|
| 1019 |
|
| 1020 | class BoolEvaluator(ArithEvaluator):
|
| 1021 | """This is also an ArithEvaluator because it has to understand.
|
| 1022 |
|
| 1023 | [[ x -eq 3 ]]
|
| 1024 |
|
| 1025 | where x='1+2'
|
| 1026 | """
|
| 1027 |
|
| 1028 | def __init__(
|
| 1029 | self,
|
| 1030 | mem, # type: state.Mem
|
| 1031 | exec_opts, # type: optview.Exec
|
| 1032 | mutable_opts, # type: Optional[state.MutableOpts]
|
| 1033 | parse_ctx, # type: Optional[parse_lib.ParseContext]
|
| 1034 | errfmt, # type: ui.ErrorFormatter
|
| 1035 | bracket=False # type: bool
|
| 1036 | ):
|
| 1037 | # type: (...) -> None
|
| 1038 | ArithEvaluator.__init__(self, mem, exec_opts, mutable_opts, parse_ctx,
|
| 1039 | errfmt)
|
| 1040 | self.bracket = bracket # [ and [[ are slightly different
|
| 1041 |
|
| 1042 | def _IsDefined(self, s, blame_loc):
|
| 1043 | # type: (str, loc_t) -> bool
|
| 1044 |
|
| 1045 | m = util.RegexSearch(consts.TEST_V_RE, s)
|
| 1046 | if m is None:
|
| 1047 | if self.exec_opts.strict_word_eval():
|
| 1048 | e_die('-v expected name or name[index]', blame_loc)
|
| 1049 | return False
|
| 1050 |
|
| 1051 | var_name = m[1]
|
| 1052 | index_str = m[3]
|
| 1053 |
|
| 1054 | val = self.mem.GetValue(var_name)
|
| 1055 | if len(index_str) == 0: # it's just a variable name
|
| 1056 | return val.tag() != value_e.Undef
|
| 1057 |
|
| 1058 | UP_val = val
|
| 1059 | with tagswitch(val) as case:
|
| 1060 | if case(value_e.InternalStringArray, value_e.BashArray):
|
| 1061 | try:
|
| 1062 | # could use mops.FromStr?
|
| 1063 | index = int(index_str)
|
| 1064 | except ValueError as e:
|
| 1065 | if self.exec_opts.strict_word_eval():
|
| 1066 | e_die(
|
| 1067 | '-v got BashArray and invalid index %r' %
|
| 1068 | index_str, blame_loc)
|
| 1069 | return False
|
| 1070 |
|
| 1071 | if val.tag() == value_e.InternalStringArray:
|
| 1072 | array_val = cast(value.InternalStringArray, UP_val)
|
| 1073 | result, error_code = bash_impl.InternalStringArray_HasElement(
|
| 1074 | array_val, index)
|
| 1075 | if error_code == error_code_e.IndexOutOfRange:
|
| 1076 | length = bash_impl.InternalStringArray_Length(
|
| 1077 | array_val)
|
| 1078 | e_die(
|
| 1079 | '-v got index %s, which is out of bounds for array of length %d'
|
| 1080 | % (index_str, length), blame_loc)
|
| 1081 |
|
| 1082 | elif val.tag() == value_e.BashArray:
|
| 1083 | sparse_val = cast(value.BashArray, UP_val)
|
| 1084 | result, error_code = bash_impl.BashArray_HasElement(
|
| 1085 | sparse_val, mops.IntWiden(index))
|
| 1086 | if error_code == error_code_e.IndexOutOfRange:
|
| 1087 | big_length = bash_impl.BashArray_Length(sparse_val)
|
| 1088 | e_die(
|
| 1089 | '-v got index %s, which is out of bounds for array of length %s'
|
| 1090 | % (index_str, mops.ToStr(big_length)), blame_loc)
|
| 1091 |
|
| 1092 | else:
|
| 1093 | raise AssertionError()
|
| 1094 |
|
| 1095 | return result
|
| 1096 |
|
| 1097 | elif case(value_e.BashAssoc):
|
| 1098 | val = cast(value.BashAssoc, UP_val)
|
| 1099 | return bash_impl.BashAssoc_HasElement(val, index_str)
|
| 1100 |
|
| 1101 | else:
|
| 1102 | # work around mycpp bug! parses as 'elif'
|
| 1103 | pass
|
| 1104 |
|
| 1105 | if self.exec_opts.strict_word_eval():
|
| 1106 | raise error.TypeErr(val, 'Expected BashArray or BashAssoc',
|
| 1107 | blame_loc)
|
| 1108 | return False
|
| 1109 | raise AssertionError()
|
| 1110 |
|
| 1111 | def _StringToBigIntOrError(self, s, blame_loc):
|
| 1112 | # type: (str, loc_t) -> mops.BigInt
|
| 1113 |
|
| 1114 | # Used by [ $x -gt 3 ]
|
| 1115 | if self.bracket:
|
| 1116 | if match.LooksLikeInteger(s):
|
| 1117 | ok, i = mops.FromStr2(s)
|
| 1118 | else:
|
| 1119 | ok = False
|
| 1120 |
|
| 1121 | if not ok:
|
| 1122 | # builtin_bracket.py catches this and return status 2, so it's
|
| 1123 | # not fatal
|
| 1124 | e_die('Invalid integer %r' % s, blame_loc)
|
| 1125 |
|
| 1126 | return i
|
| 1127 |
|
| 1128 | # Used by both [[ $x -gt 3 ]] and $(( x ))
|
| 1129 | else:
|
| 1130 | try:
|
| 1131 | i = self.StringToBigInt(s, blame_loc)
|
| 1132 | except error.Strict as e:
|
| 1133 | if self.bracket or self.exec_opts.strict_arith():
|
| 1134 | raise
|
| 1135 | else:
|
| 1136 | i = mops.ZERO
|
| 1137 | return i
|
| 1138 |
|
| 1139 | def _EvalCompoundWord(self, word, eval_flags=0):
|
| 1140 | # type: (word_t, int) -> str
|
| 1141 | val = self.word_ev.EvalWordToString(word, eval_flags)
|
| 1142 | return val.s
|
| 1143 |
|
| 1144 | def EvalB(self, node):
|
| 1145 | # type: (bool_expr_t) -> bool
|
| 1146 |
|
| 1147 | UP_node = node
|
| 1148 | with tagswitch(node) as case:
|
| 1149 | if case(bool_expr_e.WordTest):
|
| 1150 | node = cast(bool_expr.WordTest, UP_node)
|
| 1151 | s = self._EvalCompoundWord(node.w)
|
| 1152 | return bool(s)
|
| 1153 |
|
| 1154 | elif case(bool_expr_e.LogicalNot):
|
| 1155 | node = cast(bool_expr.LogicalNot, UP_node)
|
| 1156 | b = self.EvalB(node.child)
|
| 1157 | return not b
|
| 1158 |
|
| 1159 | elif case(bool_expr_e.LogicalAnd):
|
| 1160 | node = cast(bool_expr.LogicalAnd, UP_node)
|
| 1161 | # Short-circuit evaluation
|
| 1162 | if self.EvalB(node.left):
|
| 1163 | return self.EvalB(node.right)
|
| 1164 | else:
|
| 1165 | return False
|
| 1166 |
|
| 1167 | elif case(bool_expr_e.LogicalOr):
|
| 1168 | node = cast(bool_expr.LogicalOr, UP_node)
|
| 1169 | if self.EvalB(node.left):
|
| 1170 | return True
|
| 1171 | else:
|
| 1172 | return self.EvalB(node.right)
|
| 1173 |
|
| 1174 | elif case(bool_expr_e.Unary):
|
| 1175 | node = cast(bool_expr.Unary, UP_node)
|
| 1176 | op_id = node.op_id
|
| 1177 | s = self._EvalCompoundWord(node.child)
|
| 1178 |
|
| 1179 | # Now dispatch on arg type. (arg_type could be static in the
|
| 1180 | # LST?)
|
| 1181 | arg_type = consts.BoolArgType(op_id)
|
| 1182 |
|
| 1183 | if arg_type == bool_arg_type_e.Path:
|
| 1184 | return bool_stat.DoUnaryOp(op_id, s)
|
| 1185 |
|
| 1186 | if arg_type == bool_arg_type_e.Str:
|
| 1187 | if op_id == Id.BoolUnary_z:
|
| 1188 | return not bool(s)
|
| 1189 | if op_id == Id.BoolUnary_n:
|
| 1190 | return bool(s)
|
| 1191 | if op_id == Id.BoolUnary_true:
|
| 1192 | return s == 'true'
|
| 1193 | if op_id == Id.BoolUnary_false:
|
| 1194 | return s == 'false'
|
| 1195 |
|
| 1196 | raise AssertionError(op_id) # should never happen
|
| 1197 |
|
| 1198 | if arg_type == bool_arg_type_e.Other:
|
| 1199 | if op_id == Id.BoolUnary_t:
|
| 1200 | return bool_stat.isatty(s, node.child)
|
| 1201 |
|
| 1202 | # See whether 'set -o' options have been set
|
| 1203 | if op_id == Id.BoolUnary_o:
|
| 1204 | index = consts.OptionNum(s)
|
| 1205 | if index == 0:
|
| 1206 | return False
|
| 1207 | else:
|
| 1208 | return self.exec_opts.opt0_array[index]
|
| 1209 |
|
| 1210 | if op_id == Id.BoolUnary_v:
|
| 1211 | return self._IsDefined(s, loc.Word(node.child))
|
| 1212 |
|
| 1213 | e_die("%s isn't implemented" %
|
| 1214 | ui.PrettyId(op_id)) # implicit location
|
| 1215 |
|
| 1216 | raise AssertionError(arg_type)
|
| 1217 |
|
| 1218 | elif case(bool_expr_e.Binary):
|
| 1219 | node = cast(bool_expr.Binary, UP_node)
|
| 1220 |
|
| 1221 | op_id = node.op_id
|
| 1222 | # Whether to glob escape
|
| 1223 | eval_flags = 0
|
| 1224 | with switch(op_id) as case2:
|
| 1225 | if case2(Id.BoolBinary_GlobEqual, Id.BoolBinary_GlobDEqual,
|
| 1226 | Id.BoolBinary_GlobNEqual):
|
| 1227 | eval_flags |= word_eval.QUOTE_FNMATCH
|
| 1228 | elif case2(Id.BoolBinary_EqualTilde):
|
| 1229 | eval_flags |= word_eval.QUOTE_ERE
|
| 1230 |
|
| 1231 | s1 = self._EvalCompoundWord(node.left)
|
| 1232 | s2 = self._EvalCompoundWord(node.right, eval_flags)
|
| 1233 |
|
| 1234 | # Now dispatch on arg type
|
| 1235 | arg_type = consts.BoolArgType(op_id)
|
| 1236 |
|
| 1237 | if arg_type == bool_arg_type_e.Path:
|
| 1238 | return bool_stat.DoBinaryOp(op_id, s1, s2)
|
| 1239 |
|
| 1240 | if arg_type == bool_arg_type_e.Int:
|
| 1241 | # NOTE: We assume they are constants like [[ 3 -eq 3 ]].
|
| 1242 | # Bash also allows [[ 1+2 -eq 3 ]].
|
| 1243 | i1 = self._StringToBigIntOrError(s1, loc.Word(node.left))
|
| 1244 | i2 = self._StringToBigIntOrError(s2, loc.Word(node.right))
|
| 1245 |
|
| 1246 | if op_id == Id.BoolBinary_eq:
|
| 1247 | return mops.Equal(i1, i2)
|
| 1248 | if op_id == Id.BoolBinary_ne:
|
| 1249 | return not mops.Equal(i1, i2)
|
| 1250 | if op_id == Id.BoolBinary_gt:
|
| 1251 | return mops.Greater(i1, i2)
|
| 1252 | if op_id == Id.BoolBinary_ge:
|
| 1253 | return mops.Greater(i1, i2) or mops.Equal(i1, i2)
|
| 1254 | if op_id == Id.BoolBinary_lt:
|
| 1255 | return mops.Greater(i2, i1)
|
| 1256 | if op_id == Id.BoolBinary_le:
|
| 1257 | return mops.Greater(i2, i1) or mops.Equal(i1, i2)
|
| 1258 |
|
| 1259 | raise AssertionError(op_id) # should never happen
|
| 1260 |
|
| 1261 | if arg_type == bool_arg_type_e.Str:
|
| 1262 | fnmatch_flags = (FNM_CASEFOLD
|
| 1263 | if self.exec_opts.nocasematch() else 0)
|
| 1264 |
|
| 1265 | if op_id in (Id.BoolBinary_GlobEqual,
|
| 1266 | Id.BoolBinary_GlobDEqual):
|
| 1267 | #log('Matching %s against pattern %s', s1, s2)
|
| 1268 | return libc.fnmatch(s2, s1, fnmatch_flags)
|
| 1269 |
|
| 1270 | if op_id == Id.BoolBinary_GlobNEqual:
|
| 1271 | return not libc.fnmatch(s2, s1, fnmatch_flags)
|
| 1272 |
|
| 1273 | if op_id in (Id.BoolBinary_Equal, Id.BoolBinary_DEqual):
|
| 1274 | return s1 == s2
|
| 1275 |
|
| 1276 | if op_id == Id.BoolBinary_NEqual:
|
| 1277 | return s1 != s2
|
| 1278 |
|
| 1279 | if op_id == Id.BoolBinary_EqualTilde:
|
| 1280 | # TODO: This should go to --debug-file
|
| 1281 | #log('Matching %r against regex %r', s1, s2)
|
| 1282 | regex_flags = (REG_ICASE
|
| 1283 | if self.exec_opts.nocasematch() else 0)
|
| 1284 |
|
| 1285 | try:
|
| 1286 | indices = libc.regex_search(s2, regex_flags, s1, 0)
|
| 1287 | except ValueError as e:
|
| 1288 | # Status 2 indicates a regex parse error. This is
|
| 1289 | # fatal in OSH but not in bash, which treats [[
|
| 1290 | # like a command with an exit code.
|
| 1291 | e_die_status(2, e.message, loc.Word(node.right))
|
| 1292 |
|
| 1293 | if indices is not None:
|
| 1294 | self.mem.SetRegexMatch(
|
| 1295 | RegexMatch(s1, indices, eggex_ops.No))
|
| 1296 | return True
|
| 1297 | else:
|
| 1298 | self.mem.SetRegexMatch(regex_match.No)
|
| 1299 | return False
|
| 1300 |
|
| 1301 | if op_id == Id.Op_Less:
|
| 1302 | return str_cmp(s1, s2) < 0
|
| 1303 |
|
| 1304 | if op_id == Id.Op_Great:
|
| 1305 | return str_cmp(s1, s2) > 0
|
| 1306 |
|
| 1307 | raise AssertionError(op_id) # should never happen
|
| 1308 |
|
| 1309 | raise AssertionError(node.tag())
|