| 1 | from __future__ import print_function
|
| 2 | """
|
| 3 | ysh_ify.py: Roughly translate OSH to YSH. Doesn't respect semantics.
|
| 4 |
|
| 5 | ESSENTIAL
|
| 6 |
|
| 7 | Command:
|
| 8 |
|
| 9 | then/fi, do/done -> { }
|
| 10 |
|
| 11 | new case statement
|
| 12 |
|
| 13 | f() { } -> proc f { } (changes scope)
|
| 14 |
|
| 15 | subshell -> forkwait, because () is taken
|
| 16 | { } to fopen { }?
|
| 17 |
|
| 18 | Approximate: var declaration:
|
| 19 | local a=b -> var a = 'b', I think
|
| 20 |
|
| 21 | <<EOF here docs to '''
|
| 22 |
|
| 23 | Word:
|
| 24 | "$@" -> @ARGV
|
| 25 |
|
| 26 | Not common: unquoted $x -> @[split(x)]
|
| 27 |
|
| 28 | LEGACY that I don't personally use
|
| 29 |
|
| 30 | Builtins:
|
| 31 | [ -> test
|
| 32 | . -> source
|
| 33 |
|
| 34 | Word:
|
| 35 | backticks -> $() (I don't use this)
|
| 36 | quote removal "$foo" -> $foo
|
| 37 | brace removal ${foo} and "${foo}" -> $foo
|
| 38 |
|
| 39 | --tool format
|
| 40 |
|
| 41 | fix indentation and spacing, like clang-format
|
| 42 | can "lower" the LST to a rough representation with keywords / "first words",
|
| 43 | { } ( ), and comments
|
| 44 | - the "atoms" should not have newlines
|
| 45 | """
|
| 46 |
|
| 47 | from _devbuild.gen.id_kind_asdl import Id, Id_str
|
| 48 | from _devbuild.gen.runtime_asdl import word_style_e, word_style_t
|
| 49 | from _devbuild.gen.syntax_asdl import (
|
| 50 | loc,
|
| 51 | CompoundWord,
|
| 52 | Token,
|
| 53 | SimpleVarSub,
|
| 54 | BracedVarSub,
|
| 55 | CommandSub,
|
| 56 | DoubleQuoted,
|
| 57 | SingleQuoted,
|
| 58 | word_e,
|
| 59 | word_t,
|
| 60 | #word_str,
|
| 61 | word_part,
|
| 62 | word_part_e,
|
| 63 | word_part_t,
|
| 64 | rhs_word_e,
|
| 65 | rhs_word_t,
|
| 66 | sh_lhs,
|
| 67 | sh_lhs_e,
|
| 68 | command,
|
| 69 | command_e,
|
| 70 | #command_str,
|
| 71 | BraceGroup,
|
| 72 | for_iter_e,
|
| 73 | case_arg_e,
|
| 74 | case_arg,
|
| 75 | condition_e,
|
| 76 | redir_param,
|
| 77 | redir_param_e,
|
| 78 | Redir,
|
| 79 | List_of_command,
|
| 80 | ShFunction,
|
| 81 | )
|
| 82 | from asdl import runtime
|
| 83 | from core.error import p_die
|
| 84 | from display import ui
|
| 85 | from frontend import lexer
|
| 86 | from frontend import location
|
| 87 | from osh import word_
|
| 88 | from mycpp import mylib
|
| 89 | from mycpp.mylib import log, print_stderr, tagswitch
|
| 90 |
|
| 91 | from typing import Dict, cast, TYPE_CHECKING
|
| 92 | if TYPE_CHECKING:
|
| 93 | from _devbuild.gen.syntax_asdl import command_t
|
| 94 | from core import alloc
|
| 95 |
|
| 96 | _ = log
|
| 97 |
|
| 98 |
|
| 99 | class Cursor(object):
|
| 100 | """
|
| 101 | API to print/transform a complete source file, stored in a single arena.
|
| 102 |
|
| 103 | In, core/alloc.py, SnipCodeBlock() and SnipCodeString work on lines. They
|
| 104 | don't iterate over tokens.
|
| 105 |
|
| 106 | Or add a separate hash table of Token -> span ID? That makes sense because
|
| 107 | we need that kind of "address hash" for type checking anyway.
|
| 108 |
|
| 109 | You use the hash table to go from next_token_id .. TokenId(until_token).
|
| 110 | """
|
| 111 |
|
| 112 | def __init__(self, arena, f):
|
| 113 | # type: (alloc.Arena, mylib.Writer) -> None
|
| 114 | self.arena = arena
|
| 115 | self.f = f
|
| 116 | self.next_span_id = 0
|
| 117 |
|
| 118 | def _PrintUntilSpid(self, until_span_id):
|
| 119 | # type: (int) -> None
|
| 120 |
|
| 121 | # Sometimes we add +1
|
| 122 | if until_span_id == runtime.NO_SPID:
|
| 123 | assert 0, 'Missing span ID, got %d' % until_span_id
|
| 124 |
|
| 125 | for span_id in xrange(self.next_span_id, until_span_id):
|
| 126 | span = self.arena.GetToken(span_id)
|
| 127 |
|
| 128 | # A span for Eof may not have a line when the file is completely empty.
|
| 129 | if span.line is None:
|
| 130 | continue
|
| 131 |
|
| 132 | # Special case for recovering stripped leading space!
|
| 133 | # See osh/word_compile.py
|
| 134 | start_index = (0 if span.id == Id.Lit_CharsWithoutPrefix else
|
| 135 | span.col)
|
| 136 | end_index = span.col + span.length
|
| 137 |
|
| 138 | piece = span.line.content[start_index:end_index]
|
| 139 | self.f.write(piece)
|
| 140 |
|
| 141 | self.next_span_id = until_span_id
|
| 142 |
|
| 143 | def _SkipUntilSpid(self, next_span_id):
|
| 144 | # type: (int) -> None
|
| 145 | """Skip everything before next_span_id.
|
| 146 |
|
| 147 | Printing will start at next_span_id
|
| 148 | """
|
| 149 | if (next_span_id == runtime.NO_SPID or
|
| 150 | next_span_id == runtime.NO_SPID + 1):
|
| 151 | assert 0, 'Missing span ID, got %d' % next_span_id
|
| 152 | self.next_span_id = next_span_id
|
| 153 |
|
| 154 | def SkipUntil(self, tok):
|
| 155 | # type: (Token) -> None
|
| 156 | span_id = self.arena.GetSpanId(tok)
|
| 157 | self._SkipUntilSpid(span_id)
|
| 158 |
|
| 159 | def SkipPast(self, tok):
|
| 160 | # type: (Token) -> None
|
| 161 | span_id = self.arena.GetSpanId(tok)
|
| 162 | self._SkipUntilSpid(span_id + 1)
|
| 163 |
|
| 164 | def PrintUntil(self, tok):
|
| 165 | # type: (Token) -> None
|
| 166 | span_id = self.arena.GetSpanId(tok)
|
| 167 |
|
| 168 | # Test invariant
|
| 169 | if mylib.PYTHON:
|
| 170 | arena_tok = self.arena.GetToken(span_id)
|
| 171 | if tok != arena_tok:
|
| 172 | raise AssertionError(
|
| 173 | '%s %d %d != %s %d %d' %
|
| 174 | (tok, span_id, id(tok), arena_tok,
|
| 175 | self.arena.GetSpanId(arena_tok), id(arena_tok)))
|
| 176 |
|
| 177 | self._PrintUntilSpid(span_id)
|
| 178 |
|
| 179 | def PrintIncluding(self, tok):
|
| 180 | # type: (Token) -> None
|
| 181 | span_id = self.arena.GetSpanId(tok)
|
| 182 | self._PrintUntilSpid(span_id + 1)
|
| 183 |
|
| 184 | def PrintUntilEnd(self):
|
| 185 | # type: () -> None
|
| 186 | self._PrintUntilSpid(self.arena.LastSpanId())
|
| 187 |
|
| 188 |
|
| 189 | def LosslessCat(arena):
|
| 190 | # type: (alloc.Arena) -> None
|
| 191 | """
|
| 192 | For testing the lossless invariant: the tokens "add up" to the original
|
| 193 | doc.
|
| 194 | """
|
| 195 | cursor = Cursor(arena, mylib.Stdout())
|
| 196 | cursor.PrintUntilEnd()
|
| 197 |
|
| 198 |
|
| 199 | def PrintTokens(arena):
|
| 200 | # type: (alloc.Arena) -> None
|
| 201 | """Debugging tool to see tokens."""
|
| 202 |
|
| 203 | if len(arena.tokens) == 1: # Special case for line_id == -1
|
| 204 | print('Empty file with EOF token on invalid line:')
|
| 205 | print('%s' % arena.tokens[0])
|
| 206 | return
|
| 207 |
|
| 208 | # TODO:
|
| 209 | # - TSV8: end position, token type
|
| 210 | # - then an option to print token text, as a J8 string
|
| 211 | # - and then there can be a separate tool to number the columns
|
| 212 | #
|
| 213 | # - Do we also have JSON8 / HTM8 / TSV8 tokens?
|
| 214 | # - And mini-languages like glob, etc.
|
| 215 | for i, tok in enumerate(arena.tokens):
|
| 216 | piece = tok.line.content[tok.col:tok.col + tok.length]
|
| 217 | print('%5d %-20s %r' % (i, Id_str(tok.id, dot=False), piece))
|
| 218 | print_stderr('(%d tokens)' % len(arena.tokens))
|
| 219 |
|
| 220 |
|
| 221 | def TreeFind(arena, node, errfmt):
|
| 222 | # type: (alloc.Arena, command_t, ui.ErrorFormatter) -> None
|
| 223 | """
|
| 224 | Find various constructs in the tree. TODO: could this be like query
|
| 225 | language?
|
| 226 | TreeSitter has a query language for CSTs. But those are untyped, whereas
|
| 227 | we are strongly typed.
|
| 228 | """
|
| 229 | fi = Finder(arena, errfmt)
|
| 230 | fi.DoCommand(node)
|
| 231 |
|
| 232 |
|
| 233 | class Finder(object):
|
| 234 | """
|
| 235 | Walk an OSH command_t syntax tree, looking for certain constructs.
|
| 236 |
|
| 237 | Note: it might be nice to have some kind of visitor, so we don't have to
|
| 238 | repeat the traversal logic?
|
| 239 |
|
| 240 | Or even better would be a homogeneous traversal, or query language.
|
| 241 | """
|
| 242 |
|
| 243 | def __init__(self, arena, errfmt):
|
| 244 | # type: (alloc.Arena, ui.ErrorFormatter) -> None
|
| 245 | self.arena = arena
|
| 246 | self.errfmt = errfmt
|
| 247 |
|
| 248 | def DoWordPart(self, p):
|
| 249 | # type: (word_part_t) -> None
|
| 250 | UP_p = p
|
| 251 | with tagswitch(p) as case:
|
| 252 | if case(word_part_e.Literal):
|
| 253 | tok = cast(Token, UP_p)
|
| 254 | if tok.id == Id.Lit_ArrayLhsOpen:
|
| 255 | #log('*** %s', tok)
|
| 256 | self.errfmt.Print_('BAD parse?', tok)
|
| 257 | elif case(word_part_e.CommandSub):
|
| 258 | p = cast(CommandSub, UP_p)
|
| 259 | self.DoCommand(p.child)
|
| 260 |
|
| 261 | def DoWord(self, w):
|
| 262 | # type: (word_t) -> None
|
| 263 | UP_w = w
|
| 264 | with tagswitch(w) as case:
|
| 265 | if case(word_e.Compound):
|
| 266 | w = cast(CompoundWord, UP_w)
|
| 267 | part0 = w.parts[0]
|
| 268 | self.DoWordPart(part0)
|
| 269 | #log('p %r', part0)
|
| 270 | else:
|
| 271 | #log('?Word? %s', word_str(w.tag()))
|
| 272 | pass
|
| 273 |
|
| 274 | def DoRhsWord(self, w):
|
| 275 | # type: (rhs_word_t) -> None
|
| 276 | UP_w = w
|
| 277 | with tagswitch(w) as case:
|
| 278 | if case(rhs_word_e.Compound):
|
| 279 | w = cast(CompoundWord, UP_w)
|
| 280 | self.DoWord(w)
|
| 281 | elif case(rhs_word_e.Empty):
|
| 282 | pass
|
| 283 | else:
|
| 284 | raise AssertionError()
|
| 285 |
|
| 286 | def DoCommand(self, node):
|
| 287 | # type: (command_t) -> None
|
| 288 |
|
| 289 | UP_node = node
|
| 290 | with tagswitch(node) as case:
|
| 291 | if case(command_e.Simple):
|
| 292 | node = cast(command.Simple, UP_node)
|
| 293 | # Only first word has a[
|
| 294 | if len(node.words):
|
| 295 | self.DoWord(node.words[0])
|
| 296 |
|
| 297 | elif case(command_e.Sentence):
|
| 298 | node = cast(command.Sentence, UP_node)
|
| 299 | self.DoCommand(node.child)
|
| 300 |
|
| 301 | elif case(command_e.ShAssignment):
|
| 302 | node = cast(command.ShAssignment, UP_node)
|
| 303 | for pair in node.pairs:
|
| 304 | # FYI
|
| 305 | if 0:
|
| 306 | if pair.left.id == Id.Lit_ArrayLhsOpen:
|
| 307 | self.errfmt.Print_('OK', pair.left)
|
| 308 |
|
| 309 | self.DoRhsWord(pair.rhs)
|
| 310 |
|
| 311 | elif case(command_e.CommandList):
|
| 312 | node = cast(command.CommandList, UP_node)
|
| 313 | for child in node.children:
|
| 314 | self.DoCommand(child)
|
| 315 |
|
| 316 | elif case(command_e.Redirect):
|
| 317 | node = cast(command.Redirect, UP_node)
|
| 318 | self.DoCommand(node.child)
|
| 319 |
|
| 320 | elif case(command_e.Pipeline):
|
| 321 | node = cast(command.Pipeline, UP_node)
|
| 322 | for child in node.children:
|
| 323 | self.DoCommand(child)
|
| 324 |
|
| 325 | elif case(command_e.AndOr):
|
| 326 | node = cast(command.AndOr, UP_node)
|
| 327 | for child in node.children:
|
| 328 | self.DoCommand(child)
|
| 329 |
|
| 330 | # This has to be different in the function case.
|
| 331 | elif case(command_e.BraceGroup):
|
| 332 | node = cast(BraceGroup, UP_node)
|
| 333 | for child in node.children:
|
| 334 | self.DoCommand(child)
|
| 335 |
|
| 336 | elif case(command_e.Subshell):
|
| 337 | node = cast(command.Subshell, UP_node)
|
| 338 | self.DoCommand(node.child)
|
| 339 |
|
| 340 | elif case(command_e.ShFunction):
|
| 341 | node = cast(ShFunction, UP_node)
|
| 342 | self.DoCommand(node.body)
|
| 343 |
|
| 344 | elif case(command_e.DoGroup):
|
| 345 | node = cast(command.DoGroup, UP_node)
|
| 346 | for child in node.children:
|
| 347 | self.DoCommand(child)
|
| 348 |
|
| 349 | elif case(command_e.ForEach):
|
| 350 | node = cast(command.ForEach, UP_node)
|
| 351 | self.DoCommand(node.body)
|
| 352 |
|
| 353 | elif case(command_e.WhileUntil):
|
| 354 | node = cast(command.WhileUntil, UP_node)
|
| 355 | # TODO: cond
|
| 356 | self.DoCommand(node.body)
|
| 357 |
|
| 358 | elif case(command_e.If):
|
| 359 | node = cast(command.If, UP_node)
|
| 360 |
|
| 361 | for i, arm in enumerate(node.arms):
|
| 362 | # TODO: cond
|
| 363 | for child in arm.action:
|
| 364 | self.DoCommand(child)
|
| 365 |
|
| 366 | # else -> } else {
|
| 367 | if len(node.else_action):
|
| 368 | for child in node.else_action:
|
| 369 | self.DoCommand(child)
|
| 370 |
|
| 371 | elif case(command_e.Case):
|
| 372 | node = cast(command.Case, UP_node)
|
| 373 |
|
| 374 | for case_arm in node.arms:
|
| 375 | for child in case_arm.action:
|
| 376 | self.DoCommand(child)
|
| 377 |
|
| 378 | elif case(command_e.TimeBlock):
|
| 379 | node = cast(command.TimeBlock, UP_node)
|
| 380 | self.DoCommand(node.pipeline)
|
| 381 |
|
| 382 | elif case(command_e.DParen):
|
| 383 | node = cast(command.DParen, UP_node)
|
| 384 | # TODO: arith expressions can words with command subs
|
| 385 | pass
|
| 386 |
|
| 387 | elif case(command_e.DBracket):
|
| 388 | node = cast(command.DBracket, UP_node)
|
| 389 |
|
| 390 | # TODO: bool_expr_t can have words with command subs
|
| 391 | pass
|
| 392 |
|
| 393 | else:
|
| 394 | #log('?Command? %s', command_str(node.tag()))
|
| 395 | pass
|
| 396 |
|
| 397 | #cursor = Cursor(arena, mylib.Stdout())
|
| 398 | #cursor.PrintUntilEnd()
|
| 399 |
|
| 400 |
|
| 401 | def Ysh_ify(arena, node):
|
| 402 | # type: (alloc.Arena, command_t) -> None
|
| 403 | cursor = Cursor(arena, mylib.Stdout())
|
| 404 | fixer = YshPrinter(cursor, arena, mylib.Stdout())
|
| 405 | fixer.DoCommand(node, None, at_top_level=True) # no local symbols yet
|
| 406 | fixer.End()
|
| 407 |
|
| 408 |
|
| 409 | # PROBLEM: ~ substitution. That is disabled by "".
|
| 410 | # You can turn it into $HOME I guess
|
| 411 | # const foo = "$HOME/src"
|
| 412 | # const foo = %( ~/src )[0] # does this make sense?
|
| 413 |
|
| 414 |
|
| 415 | def _GetRhsStyle(w):
|
| 416 | # type: (rhs_word_t) -> word_style_t
|
| 417 | """Determine what style an assignment should use. '' or "", or an
|
| 418 | expression.
|
| 419 |
|
| 420 | SQ foo= setglobal foo = ''
|
| 421 | SQ foo='' setglobal foo = ''
|
| 422 | DQ foo="" setglobal foo = "" # Or we could normalize it if no subs?
|
| 423 | DQ foo="" setglobal foo = "" # Or we could normalize it if no subs?
|
| 424 |
|
| 425 | # Need these too.
|
| 426 | # Or honestly should C strings be the default? And then raw strings are
|
| 427 | # optional? Because most usages of \n and \0 can turn into Oil?
|
| 428 | # Yeah I want the default to be statically parseable, so we subvert the \t
|
| 429 | # and \n of command line tools?
|
| 430 | # As long as we are fully analyzing the strings, we might as well go all the
|
| 431 | # way!
|
| 432 | # I think I need a PartialStaticEval() to paper over this.
|
| 433 | #
|
| 434 | # The main issue is regex and globs, because they use escape for a different
|
| 435 | # purpose. I think just do
|
| 436 | # grep r'foo\tbar' or something.
|
| 437 |
|
| 438 | C_SQ foo=$'\n' setglobal foo = C'\n'
|
| 439 | C_DQ foo=$'\n'"$bar" setglobal foo = C"\n$(bar)"
|
| 440 |
|
| 441 | Expr path=${1:-} setglobal path = $1 or ''
|
| 442 | Expr host=${2:-$(hostname)} setglobal host = $2 or $[hostname]
|
| 443 |
|
| 444 | What's the difference between Expr and Unquoted? I think they're the same/
|
| 445 | """
|
| 446 | # Actually splitting NEVER HAPPENS ON ASSIGNMENT. LEAVE IT OFF.
|
| 447 |
|
| 448 | UP_w = w
|
| 449 | with tagswitch(w) as case:
|
| 450 | if case(rhs_word_e.Empty):
|
| 451 | return word_style_e.SQ
|
| 452 |
|
| 453 | elif case(rhs_word_e.Compound):
|
| 454 | w = cast(CompoundWord, UP_w)
|
| 455 | if len(w.parts) == 0:
|
| 456 | raise AssertionError(w)
|
| 457 |
|
| 458 | elif len(w.parts) == 1:
|
| 459 | part0 = w.parts[0]
|
| 460 | UP_part0 = part0
|
| 461 | with tagswitch(part0) as case:
|
| 462 | # VAR_SUBS
|
| 463 | if case(word_part_e.TildeSub):
|
| 464 | # x=~andy/src
|
| 465 | # -> setvar x = homedir('andy') + '/src'
|
| 466 | return word_style_e.Expr
|
| 467 |
|
| 468 | elif case(word_part_e.Literal):
|
| 469 | # local x=y
|
| 470 | # -> var x = 'y'
|
| 471 | return word_style_e.SQ
|
| 472 |
|
| 473 | elif case(word_part_e.SimpleVarSub):
|
| 474 | # local x=$myvar
|
| 475 | # -> var x = "$myvar"
|
| 476 | # or var x = ${myvar}
|
| 477 | # or var x = myvar
|
| 478 | return word_style_e.DQ
|
| 479 |
|
| 480 | elif case(word_part_e.BracedVarSub, word_part_e.CommandSub,
|
| 481 | word_part_e.ArithSub):
|
| 482 | # x=$(hostname)
|
| 483 | # -> setvar x = $(hostname)
|
| 484 | return word_style_e.Unquoted
|
| 485 |
|
| 486 | elif case(word_part_e.DoubleQuoted):
|
| 487 | part0 = cast(DoubleQuoted, UP_part0)
|
| 488 |
|
| 489 | # TODO: remove quotes in single part like "$(hostname)" -> $(hostname)
|
| 490 | return word_style_e.DQ
|
| 491 |
|
| 492 | else:
|
| 493 | # multiple parts use YSTR in general?
|
| 494 | # Depends if there are subs
|
| 495 | return word_style_e.DQ
|
| 496 |
|
| 497 | # Default
|
| 498 | return word_style_e.SQ
|
| 499 |
|
| 500 |
|
| 501 | class YshPrinter(object):
|
| 502 | """Prettify OSH to YSH."""
|
| 503 |
|
| 504 | def __init__(self, cursor, arena, f):
|
| 505 | # type: (Cursor, alloc.Arena, mylib.Writer) -> None
|
| 506 | self.cursor = cursor
|
| 507 | self.arena = arena
|
| 508 | self.f = f
|
| 509 |
|
| 510 | def _DebugSpid(self, spid):
|
| 511 | # type: (int) -> None
|
| 512 | span = self.arena.GetToken(spid)
|
| 513 | s = span.line.content[span.col:span.col + span.length]
|
| 514 | print_stderr('SPID %d = %r' % (spid, s))
|
| 515 |
|
| 516 | def End(self):
|
| 517 | # type: () -> None
|
| 518 | """Make sure we print until the end of the file."""
|
| 519 | self.cursor.PrintUntilEnd()
|
| 520 |
|
| 521 | def DoRedirect(self, node, local_symbols):
|
| 522 | # type: (Redir, Dict[str, bool]) -> None
|
| 523 | """
|
| 524 | Change here docs to <<< '''
|
| 525 | """
|
| 526 | #print(node, file=sys.stderr)
|
| 527 | op_id = node.op.id
|
| 528 | self.cursor.PrintUntil(node.op)
|
| 529 |
|
| 530 | if node.arg.tag() == redir_param_e.HereDoc:
|
| 531 | here_doc = cast(redir_param.HereDoc, node.arg)
|
| 532 |
|
| 533 | here_begin = here_doc.here_begin
|
| 534 | ok, delimiter, delim_quoted = word_.StaticEval(here_begin)
|
| 535 | if not ok:
|
| 536 | p_die('Invalid here doc delimiter', loc.Word(here_begin))
|
| 537 |
|
| 538 | # Turn everything into <<<. We just change the quotes
|
| 539 | self.f.write('<<<')
|
| 540 |
|
| 541 | if delim_quoted:
|
| 542 | self.f.write(" '''")
|
| 543 | else:
|
| 544 | self.f.write(' """')
|
| 545 |
|
| 546 | delim_end_tok = location.RightTokenForWord(here_begin)
|
| 547 | self.cursor.SkipPast(delim_end_tok)
|
| 548 |
|
| 549 | # Now print the lines. TODO: Have a flag to indent these to the level of
|
| 550 | # the owning command, e.g.
|
| 551 | # cat <<EOF
|
| 552 | # EOF
|
| 553 | # Or since most here docs are the top level, you could just have a hack
|
| 554 | # for a fixed indent? TODO: Look at real use cases.
|
| 555 | for part in here_doc.stdin_parts:
|
| 556 | self.DoWordPart(part, local_symbols)
|
| 557 |
|
| 558 | self.cursor.SkipPast(here_doc.here_end_tok)
|
| 559 | if delim_quoted:
|
| 560 | self.f.write("'''\n")
|
| 561 | else:
|
| 562 | self.f.write('"""\n')
|
| 563 |
|
| 564 | else:
|
| 565 | pass
|
| 566 |
|
| 567 | # cat << EOF
|
| 568 | # hello $name
|
| 569 | # EOF
|
| 570 | # cat <<< """
|
| 571 | # hello $name
|
| 572 | # """
|
| 573 |
|
| 574 | # cat << 'EOF'
|
| 575 | # no expansion
|
| 576 | # EOF
|
| 577 |
|
| 578 | # cat <<< '''
|
| 579 | # no expansion
|
| 580 | # '''
|
| 581 |
|
| 582 | def DoShAssignment(self, node, at_top_level, local_symbols):
|
| 583 | # type: (command.ShAssignment, bool, Dict[str, bool]) -> None
|
| 584 | """
|
| 585 | local_symbols:
|
| 586 | - Add every 'local' declaration to it
|
| 587 | - problem: what if you have local in an "if" ?
|
| 588 | - we could treat it like nested scope and see what happens? Do any
|
| 589 | programs have a problem with it?
|
| 590 | case/if/for/while/BraceGroup all define scopes or what?
|
| 591 | You don't want inconsistency of variables that could be defined at
|
| 592 | any point.
|
| 593 | - or maybe you only need it within "if / case" ? Well I guess
|
| 594 | for/while can break out of the loop and cause problems. A break is
|
| 595 | an "if".
|
| 596 |
|
| 597 | - for subsequent
|
| 598 | """
|
| 599 | # Change RHS to expression language. Bare words not allowed. foo -> 'foo'
|
| 600 |
|
| 601 | has_rhs = False # TODO: Should be on a per-variable basis.
|
| 602 | # local a=b c=d, or just punt on those
|
| 603 | defined_locally = False # is it a local variable in this function?
|
| 604 | # can't tell if global
|
| 605 |
|
| 606 | if True:
|
| 607 | self.cursor.PrintUntil(node.pairs[0].left)
|
| 608 |
|
| 609 | # For now, just detect whether the FIRST assignment on the line has been
|
| 610 | # declared locally. We might want to split every line into separate
|
| 611 | # statements.
|
| 612 | if local_symbols is not None:
|
| 613 | lhs0 = node.pairs[0].lhs
|
| 614 | #if lhs0.tag() == sh_lhs_e.Name and lhs0.name in local_symbols:
|
| 615 | # defined_locally = True
|
| 616 |
|
| 617 | #print("CHECKING NAME", lhs0.name, defined_locally, local_symbols)
|
| 618 |
|
| 619 | # TODO: Avoid translating these
|
| 620 | has_array_index = [
|
| 621 | pair.lhs.tag() == sh_lhs_e.UnparsedIndex for pair in node.pairs
|
| 622 | ]
|
| 623 |
|
| 624 | # need semantic analysis.
|
| 625 | # Would be nice to assume that it's a local though.
|
| 626 | if at_top_level:
|
| 627 | self.f.write('setvar ')
|
| 628 | elif defined_locally:
|
| 629 | self.f.write('set ')
|
| 630 | #self.f.write('[local mutated]')
|
| 631 | else:
|
| 632 | # We're in a function, but it's not defined locally, so we must be
|
| 633 | # mutating a global.
|
| 634 | self.f.write('setvar ')
|
| 635 |
|
| 636 | # foo=bar spam=eggs -> foo = 'bar', spam = 'eggs'
|
| 637 | n = len(node.pairs)
|
| 638 | for i, pair in enumerate(node.pairs):
|
| 639 | lhs = pair.lhs
|
| 640 | UP_lhs = lhs
|
| 641 | with tagswitch(lhs) as case:
|
| 642 | if case(sh_lhs_e.Name):
|
| 643 | lhs = cast(sh_lhs.Name, UP_lhs)
|
| 644 |
|
| 645 | self.cursor.PrintUntil(pair.left)
|
| 646 | # Assume skipping over one Lit_VarLike token
|
| 647 | self.cursor.SkipPast(pair.left)
|
| 648 |
|
| 649 | # Replace name. I guess it's Lit_Chars.
|
| 650 | self.f.write(lhs.name)
|
| 651 | self.f.write(' = ')
|
| 652 |
|
| 653 | # TODO: This should be translated from Empty.
|
| 654 | if pair.rhs.tag() == rhs_word_e.Empty:
|
| 655 | self.f.write("''") # local i -> var i = ''
|
| 656 | else:
|
| 657 | self.DoRhsWord(pair.rhs, local_symbols)
|
| 658 |
|
| 659 | elif case(sh_lhs_e.UnparsedIndex):
|
| 660 | # --one-pass-parse gives us this node, instead of IndexedName
|
| 661 | pass
|
| 662 |
|
| 663 | else:
|
| 664 | raise AssertionError(pair.lhs.__class__.__name__)
|
| 665 |
|
| 666 | if i != n - 1:
|
| 667 | self.f.write(',')
|
| 668 |
|
| 669 | def _DoSimple(self, node, local_symbols):
|
| 670 | # type: (command.Simple, Dict[str, bool]) -> None
|
| 671 |
|
| 672 | # How to preserve spaces between words? Do you want to do it?
|
| 673 | # Well you need to test this:
|
| 674 | #
|
| 675 | # echo foo \
|
| 676 | # bar
|
| 677 |
|
| 678 | if len(node.more_env):
|
| 679 | # We only need to transform the right side, not left side.
|
| 680 | for pair in node.more_env:
|
| 681 | self.DoRhsWord(pair.val, local_symbols)
|
| 682 |
|
| 683 | if len(node.words):
|
| 684 | first_word = node.words[0]
|
| 685 | ok, val, quoted = word_.StaticEval(first_word)
|
| 686 | word0_tok = location.LeftTokenForWord(first_word)
|
| 687 | if ok and not quoted:
|
| 688 | if val == '[' and len(node.words) >= 3:
|
| 689 | word2 = node.words[-2]
|
| 690 | last_word = node.words[-1]
|
| 691 |
|
| 692 | # Check if last word is ]
|
| 693 | ok, val, quoted = word_.StaticEval(last_word)
|
| 694 | if ok and not quoted and val == ']':
|
| 695 | # Replace [ with 'test'
|
| 696 | self.cursor.PrintUntil(word0_tok)
|
| 697 | self.cursor.SkipPast(word0_tok)
|
| 698 | self.f.write('test')
|
| 699 |
|
| 700 | for w in node.words[1:-1]:
|
| 701 | self.DoWordInCommand(w, local_symbols)
|
| 702 |
|
| 703 | # Now omit ]
|
| 704 | tok2 = location.RightTokenForWord(word2)
|
| 705 | rbrack_tok = location.LeftTokenForWord(last_word)
|
| 706 |
|
| 707 | # Skip the space token before ]
|
| 708 | self.cursor.PrintIncluding(tok2)
|
| 709 | # ] takes one spid
|
| 710 | self.cursor.SkipPast(rbrack_tok)
|
| 711 | return
|
| 712 | else:
|
| 713 | raise RuntimeError('Got [ without ]')
|
| 714 |
|
| 715 | elif val == '.':
|
| 716 | self.cursor.PrintUntil(word0_tok)
|
| 717 | self.cursor.SkipPast(word0_tok)
|
| 718 | self.f.write('source')
|
| 719 | return
|
| 720 |
|
| 721 | for w in node.words:
|
| 722 | self.DoWordInCommand(w, local_symbols)
|
| 723 |
|
| 724 | # TODO: Print the terminator. Could be \n or ;
|
| 725 | # Need to print env like PYTHONPATH = 'foo' && ls
|
| 726 | # Need to print redirects:
|
| 727 | # < > are the same. << is here string, and >> is assignment.
|
| 728 | # append is >+
|
| 729 |
|
| 730 | # TODO: static_eval of simple command
|
| 731 | # - [ -> "test". Eliminate trailing ].
|
| 732 | # - . -> source, etc.
|
| 733 |
|
| 734 | def DoCommand(self, node, local_symbols, at_top_level=False):
|
| 735 | # type: (command_t, Dict[str, bool], bool) -> None
|
| 736 |
|
| 737 | UP_node = node
|
| 738 |
|
| 739 | with tagswitch(node) as case:
|
| 740 | if case(command_e.CommandList):
|
| 741 | node = cast(command.CommandList, UP_node)
|
| 742 |
|
| 743 | # TODO: How to distinguish between echo hi; echo bye; and on
|
| 744 | # separate lines
|
| 745 | for child in node.children:
|
| 746 | self.DoCommand(child,
|
| 747 | local_symbols,
|
| 748 | at_top_level=at_top_level)
|
| 749 |
|
| 750 | elif case(command_e.Redirect):
|
| 751 | node = cast(command.Redirect, UP_node)
|
| 752 |
|
| 753 | self.DoCommand(node.child,
|
| 754 | local_symbols,
|
| 755 | at_top_level=at_top_level)
|
| 756 | for r in node.redirects:
|
| 757 | self.DoRedirect(r, local_symbols)
|
| 758 |
|
| 759 | elif case(command_e.Simple):
|
| 760 | node = cast(command.Simple, UP_node)
|
| 761 |
|
| 762 | self._DoSimple(node, local_symbols)
|
| 763 |
|
| 764 | elif case(command_e.ShAssignment):
|
| 765 | node = cast(command.ShAssignment, UP_node)
|
| 766 |
|
| 767 | self.DoShAssignment(node, at_top_level, local_symbols)
|
| 768 |
|
| 769 | elif case(command_e.Pipeline):
|
| 770 | node = cast(command.Pipeline, UP_node)
|
| 771 |
|
| 772 | for child in node.children:
|
| 773 | self.DoCommand(child, local_symbols)
|
| 774 |
|
| 775 | elif case(command_e.AndOr):
|
| 776 | node = cast(command.AndOr, UP_node)
|
| 777 |
|
| 778 | for child in node.children:
|
| 779 | self.DoCommand(child, local_symbols)
|
| 780 |
|
| 781 | elif case(command_e.Sentence):
|
| 782 | node = cast(command.Sentence, UP_node)
|
| 783 |
|
| 784 | # 'ls &' to 'fork ls'
|
| 785 | # Keep ; the same.
|
| 786 | self.DoCommand(node.child, local_symbols)
|
| 787 |
|
| 788 | # This has to be different in the function case.
|
| 789 | elif case(command_e.BraceGroup):
|
| 790 | node = cast(BraceGroup, UP_node)
|
| 791 |
|
| 792 | # { echo hi; } -> do { echo hi }
|
| 793 | # For now it might be OK to keep 'do { echo hi; }
|
| 794 | self.cursor.PrintUntil(node.left)
|
| 795 | self.cursor.SkipPast(node.left)
|
| 796 | self.f.write('do {')
|
| 797 |
|
| 798 | for child in node.children:
|
| 799 | self.DoCommand(child, local_symbols)
|
| 800 |
|
| 801 | elif case(command_e.Subshell):
|
| 802 | node = cast(command.Subshell, UP_node)
|
| 803 |
|
| 804 | # (echo hi) -> shell echo hi
|
| 805 | # (echo hi; echo bye) -> shell {echo hi; echo bye}
|
| 806 |
|
| 807 | self.cursor.PrintUntil(node.left)
|
| 808 | self.cursor.SkipPast(node.left)
|
| 809 | self.f.write('shell {')
|
| 810 |
|
| 811 | self.DoCommand(node.child, local_symbols)
|
| 812 |
|
| 813 | #self._DebugSpid(right_spid)
|
| 814 | #self._DebugSpid(right_spid + 1)
|
| 815 |
|
| 816 | #print('RIGHT SPID', right_spid)
|
| 817 | self.cursor.PrintUntil(node.right)
|
| 818 | self.cursor.SkipPast(node.right)
|
| 819 | self.f.write('}')
|
| 820 |
|
| 821 | elif case(command_e.ShFunction):
|
| 822 | node = cast(ShFunction, UP_node)
|
| 823 |
|
| 824 | # TODO: skip name
|
| 825 | #self.f.write('proc %s' % node.name)
|
| 826 |
|
| 827 | # New symbol table for every function.
|
| 828 | new_local_symbols = {} # type: Dict[str, bool]
|
| 829 |
|
| 830 | # Should be the left most span, including 'function'
|
| 831 | if node.keyword: # function foo { ...
|
| 832 | self.cursor.PrintUntil(node.keyword)
|
| 833 | else: # foo() { ...
|
| 834 | self.cursor.PrintUntil(node.name_tok)
|
| 835 |
|
| 836 | self.f.write('proc %s ' % node.name)
|
| 837 |
|
| 838 | UP_body = node.body
|
| 839 | with tagswitch(UP_body) as case:
|
| 840 | if case(command_e.BraceGroup):
|
| 841 | body = cast(BraceGroup, UP_body)
|
| 842 | self.cursor.SkipUntil(body.left)
|
| 843 |
|
| 844 | # Don't add "do" like a standalone brace group. Just use {}.
|
| 845 | for child in body.children:
|
| 846 | self.DoCommand(child, new_local_symbols)
|
| 847 | else:
|
| 848 | # very rare cases like f() ( subshell )
|
| 849 | pass
|
| 850 |
|
| 851 | elif case(command_e.DoGroup):
|
| 852 | node = cast(command.DoGroup, UP_node)
|
| 853 |
|
| 854 | self.cursor.PrintUntil(node.left)
|
| 855 | self.cursor.SkipPast(node.left)
|
| 856 | self.f.write('{')
|
| 857 |
|
| 858 | for child in node.children:
|
| 859 | self.DoCommand(child, local_symbols)
|
| 860 |
|
| 861 | self.cursor.PrintUntil(node.right)
|
| 862 | self.cursor.SkipPast(node.right)
|
| 863 | self.f.write('}')
|
| 864 |
|
| 865 | elif case(command_e.ForEach):
|
| 866 | node = cast(command.ForEach, UP_node)
|
| 867 |
|
| 868 | # Need to preserve spaces between words, because there can be line
|
| 869 | # wrapping.
|
| 870 | # for x in a b c \
|
| 871 | # d e f; do
|
| 872 |
|
| 873 | UP_iterable = node.iterable
|
| 874 | with tagswitch(node.iterable) as case:
|
| 875 | if case(for_iter_e.Args):
|
| 876 | self.f.write('for %s in @ARGV ' % node.iter_names[0])
|
| 877 |
|
| 878 | # note: command_t doesn't have .spids
|
| 879 | body_tok = location.TokenForCommand(node.body)
|
| 880 | self.cursor.SkipUntil(body_tok)
|
| 881 |
|
| 882 | elif case(for_iter_e.Words):
|
| 883 | pass
|
| 884 |
|
| 885 | elif case(for_iter_e.YshExpr):
|
| 886 | pass
|
| 887 |
|
| 888 | if node.semi_tok is not None:
|
| 889 | self.cursor.PrintUntil(node.semi_tok)
|
| 890 | self.cursor.SkipPast(node.semi_tok)
|
| 891 |
|
| 892 | self.DoCommand(node.body, local_symbols)
|
| 893 |
|
| 894 | elif case(command_e.WhileUntil):
|
| 895 | node = cast(command.WhileUntil, UP_node)
|
| 896 |
|
| 897 | # Skip 'until', and replace it with 'while not'
|
| 898 | if node.keyword.id == Id.KW_Until:
|
| 899 | self.cursor.PrintUntil(node.keyword)
|
| 900 | self.cursor.SkipPast(node.keyword)
|
| 901 | self.f.write('while !')
|
| 902 |
|
| 903 | if node.cond.tag() == condition_e.Shell:
|
| 904 | commands = cast(List_of_command, node.cond)
|
| 905 | # Skip the semi-colon in the condition, which is usually a Sentence
|
| 906 | if (len(commands) == 1 and
|
| 907 | commands[0].tag() == command_e.Sentence):
|
| 908 | sentence = cast(command.Sentence, commands[0])
|
| 909 | self.DoCommand(sentence.child, local_symbols)
|
| 910 | self.cursor.SkipPast(sentence.terminator)
|
| 911 |
|
| 912 | self.DoCommand(node.body, local_symbols)
|
| 913 |
|
| 914 | elif case(command_e.If):
|
| 915 | node = cast(command.If, UP_node)
|
| 916 |
|
| 917 | # if foo; then -> if foo {
|
| 918 | # elif foo; then -> } elif foo {
|
| 919 | for i, arm in enumerate(node.arms):
|
| 920 | elif_tok = arm.keyword
|
| 921 | then_tok = arm.then_tok
|
| 922 |
|
| 923 | if i != 0: # 'if' not 'elif' on the first arm
|
| 924 | self.cursor.PrintUntil(elif_tok)
|
| 925 | self.f.write('} ')
|
| 926 |
|
| 927 | cond = arm.cond
|
| 928 | if cond.tag() == condition_e.Shell:
|
| 929 | commands = cast(List_of_command, cond)
|
| 930 | if (len(commands) == 1 and
|
| 931 | commands[0].tag() == command_e.Sentence):
|
| 932 | sentence = cast(command.Sentence, commands[0])
|
| 933 | self.DoCommand(sentence, local_symbols)
|
| 934 |
|
| 935 | # Remove semi-colon
|
| 936 | self.cursor.PrintUntil(sentence.terminator)
|
| 937 | self.cursor.SkipPast(sentence.terminator)
|
| 938 | else:
|
| 939 | for child in commands:
|
| 940 | self.DoCommand(child, local_symbols)
|
| 941 |
|
| 942 | self.cursor.PrintUntil(then_tok)
|
| 943 | self.cursor.SkipPast(then_tok)
|
| 944 | self.f.write('{')
|
| 945 |
|
| 946 | for child in arm.action:
|
| 947 | self.DoCommand(child, local_symbols)
|
| 948 |
|
| 949 | # else -> } else {
|
| 950 | if len(node.else_action):
|
| 951 | self.cursor.PrintUntil(node.else_kw)
|
| 952 | self.f.write('} ')
|
| 953 | self.cursor.PrintIncluding(node.else_kw)
|
| 954 | self.f.write(' {')
|
| 955 |
|
| 956 | for child in node.else_action:
|
| 957 | self.DoCommand(child, local_symbols)
|
| 958 |
|
| 959 | # fi -> }
|
| 960 | self.cursor.PrintUntil(node.fi_kw)
|
| 961 | self.cursor.SkipPast(node.fi_kw)
|
| 962 | self.f.write('}')
|
| 963 |
|
| 964 | elif case(command_e.Case):
|
| 965 | node = cast(command.Case, UP_node)
|
| 966 |
|
| 967 | to_match = None # type: word_t
|
| 968 | with tagswitch(node.to_match) as case:
|
| 969 | if case(case_arg_e.YshExpr):
|
| 970 | return
|
| 971 | elif case(case_arg_e.Word):
|
| 972 | to_match = cast(case_arg.Word, node.to_match).w
|
| 973 | else:
|
| 974 | raise AssertionError()
|
| 975 |
|
| 976 | self.cursor.PrintIncluding(node.case_kw)
|
| 977 |
|
| 978 | # Figure out the variable name, so we can translate
|
| 979 | # - $var to (var)
|
| 980 | # - "$var" to (var)
|
| 981 | var_part = None # type: SimpleVarSub
|
| 982 | with tagswitch(to_match) as case:
|
| 983 | if case(word_e.Compound):
|
| 984 | w = cast(CompoundWord, to_match)
|
| 985 | part0 = w.parts[0]
|
| 986 |
|
| 987 | with tagswitch(part0) as case2:
|
| 988 | if case2(word_part_e.SimpleVarSub):
|
| 989 | var_part = cast(SimpleVarSub, part0)
|
| 990 |
|
| 991 | elif case2(word_part_e.DoubleQuoted):
|
| 992 | dq_part = cast(DoubleQuoted, part0)
|
| 993 | if len(dq_part.parts) == 1:
|
| 994 | dq_part0 = dq_part.parts[0]
|
| 995 |
|
| 996 | # Nesting is annoying -- it would be nice to use pattern
|
| 997 | # matching, but mycpp won't like it.
|
| 998 | # TODO: extract into a common function
|
| 999 | with tagswitch(dq_part0) as case3:
|
| 1000 | if case3(word_part_e.SimpleVarSub):
|
| 1001 | var_part = cast(
|
| 1002 | SimpleVarSub, dq_part0)
|
| 1003 | #log("VAR PART %s", var_part)
|
| 1004 |
|
| 1005 | if var_part:
|
| 1006 | self.f.write(' (')
|
| 1007 | self.f.write(lexer.LazyStr(var_part.tok))
|
| 1008 | self.f.write(') ')
|
| 1009 |
|
| 1010 | self.cursor.SkipPast(node.arms_start) # Skip past 'in'
|
| 1011 | self.f.write('{')
|
| 1012 |
|
| 1013 | missing_last_dsemi = False
|
| 1014 |
|
| 1015 | for case_arm in node.arms:
|
| 1016 | # Replace ) with {
|
| 1017 | self.cursor.PrintUntil(case_arm.middle)
|
| 1018 | self.f.write(' {')
|
| 1019 | self.cursor.SkipPast(case_arm.middle)
|
| 1020 |
|
| 1021 | for child in case_arm.action:
|
| 1022 | self.DoCommand(child, local_symbols)
|
| 1023 |
|
| 1024 | if case_arm.right:
|
| 1025 | # Change ;; to }
|
| 1026 | self.cursor.PrintUntil(case_arm.right)
|
| 1027 | self.f.write('}')
|
| 1028 | self.cursor.SkipPast(case_arm.right)
|
| 1029 | else:
|
| 1030 | # valid: case $x in pat) echo hi ; esac
|
| 1031 | missing_last_dsemi = True
|
| 1032 |
|
| 1033 | self.cursor.PrintUntil(node.arms_end) # 'esac' or }
|
| 1034 |
|
| 1035 | if missing_last_dsemi: # Print } for missing ;;
|
| 1036 | self.f.write('}\n')
|
| 1037 |
|
| 1038 | self.cursor.SkipPast(node.arms_end) # 'esac' or }
|
| 1039 |
|
| 1040 | self.f.write('}') # in place of 'esac'
|
| 1041 |
|
| 1042 | elif case(command_e.TimeBlock):
|
| 1043 | node = cast(command.TimeBlock, UP_node)
|
| 1044 |
|
| 1045 | self.DoCommand(node.pipeline, local_symbols)
|
| 1046 |
|
| 1047 | elif case(command_e.DParen):
|
| 1048 | node = cast(command.DParen, UP_node)
|
| 1049 | # TODO: arith expressions can words with command subs
|
| 1050 | pass
|
| 1051 |
|
| 1052 | elif case(command_e.DBracket):
|
| 1053 | node = cast(command.DBracket, UP_node)
|
| 1054 |
|
| 1055 | # TODO: bool_expr_t can have words with command subs
|
| 1056 | pass
|
| 1057 |
|
| 1058 | else:
|
| 1059 | pass
|
| 1060 | #log('Command not handled: %s', node)
|
| 1061 | #raise AssertionError(node.__class__.__name__)
|
| 1062 |
|
| 1063 | def DoRhsWord(self, node, local_symbols):
|
| 1064 | # type: (rhs_word_t, Dict[str, bool]) -> None
|
| 1065 | """For the RHS of assignments.
|
| 1066 |
|
| 1067 | TODO: for complex cases of word joining:
|
| 1068 | local a=unquoted'single'"double"'"'
|
| 1069 |
|
| 1070 | We can try to handle it:
|
| 1071 | var a = y"unquotedsingledouble\""
|
| 1072 |
|
| 1073 | Or simply abort and LEAVE IT ALONE. We should only translate things we
|
| 1074 | recognize.
|
| 1075 | """
|
| 1076 | UP_node = node
|
| 1077 | with tagswitch(node) as case:
|
| 1078 | if case(rhs_word_e.Empty):
|
| 1079 | self.f.write("''")
|
| 1080 |
|
| 1081 | elif case(rhs_word_e.Compound):
|
| 1082 | node = cast(CompoundWord, UP_node)
|
| 1083 |
|
| 1084 | # TODO: This is wrong!
|
| 1085 | style = _GetRhsStyle(node)
|
| 1086 | if style == word_style_e.SQ:
|
| 1087 | self.f.write("'")
|
| 1088 | self.DoWordInCommand(node, local_symbols)
|
| 1089 | self.f.write("'")
|
| 1090 | elif style == word_style_e.DQ:
|
| 1091 | self.f.write('"')
|
| 1092 | self.DoWordInCommand(node, local_symbols)
|
| 1093 | self.f.write('"')
|
| 1094 | # TODO: Put these back
|
| 1095 | #elif style == word_style_e.Expr:
|
| 1096 | # pass
|
| 1097 | #elif style == word_style_e.Unquoted:
|
| 1098 | # pass
|
| 1099 | else:
|
| 1100 | # "${foo:-default}" -> foo or 'default'
|
| 1101 | # ${foo:-default} -> @split(foo or 'default')
|
| 1102 | # @(foo or 'default') -- implicit split.
|
| 1103 |
|
| 1104 | if word_.IsVarSub(node): # ${1} or "$1"
|
| 1105 | # Do it in expression mode
|
| 1106 | pass
|
| 1107 | # NOTE: ArithSub with $(1 +2 ) is different than 1 + 2 because of
|
| 1108 | # conversion to string.
|
| 1109 |
|
| 1110 | # For now, just stub it out
|
| 1111 | self.DoWordInCommand(node, local_symbols)
|
| 1112 |
|
| 1113 | def DoWordInCommand(self, node, local_symbols):
|
| 1114 | # type: (word_t, Dict[str, bool]) -> None
|
| 1115 | """E.g. remove unquoted.
|
| 1116 |
|
| 1117 | echo "$x" -> echo $x
|
| 1118 | """
|
| 1119 | UP_node = node
|
| 1120 |
|
| 1121 | with tagswitch(node) as case:
|
| 1122 | if case(word_e.Compound):
|
| 1123 | node = cast(CompoundWord, UP_node)
|
| 1124 |
|
| 1125 | # UNQUOTE simple var subs
|
| 1126 |
|
| 1127 | # Special case for "$@".
|
| 1128 | # TODO:
|
| 1129 | # "$foo" -> $foo
|
| 1130 | # "${foo}" -> $foo
|
| 1131 |
|
| 1132 | if (len(node.parts) == 1 and
|
| 1133 | node.parts[0].tag() == word_part_e.DoubleQuoted):
|
| 1134 | dq_part = cast(DoubleQuoted, node.parts[0])
|
| 1135 |
|
| 1136 | # NOTE: In double quoted case, this is the begin and end quote.
|
| 1137 | # Do we need a HereDoc part?
|
| 1138 |
|
| 1139 | if len(dq_part.parts) == 1:
|
| 1140 | part0 = dq_part.parts[0]
|
| 1141 | if part0.tag() == word_part_e.SimpleVarSub:
|
| 1142 | vsub_part = cast(SimpleVarSub, dq_part.parts[0])
|
| 1143 | if vsub_part.tok.id == Id.VSub_At:
|
| 1144 | self.cursor.PrintUntil(dq_part.left)
|
| 1145 | self.cursor.SkipPast(
|
| 1146 | dq_part.right) # " then $@ then "
|
| 1147 | self.f.write('@ARGV')
|
| 1148 | return # Done replacing
|
| 1149 |
|
| 1150 | # "$1" -> $1, "$foo" -> $foo
|
| 1151 | if vsub_part.tok.id in (Id.VSub_Number,
|
| 1152 | Id.VSub_DollarName):
|
| 1153 | self.cursor.PrintUntil(dq_part.left)
|
| 1154 | self.cursor.SkipPast(dq_part.right)
|
| 1155 | self.f.write(lexer.TokenVal(vsub_part.tok))
|
| 1156 | return
|
| 1157 |
|
| 1158 | # Single arith sub, command sub, etc.
|
| 1159 | # On the other hand, an unquoted one needs to turn into
|
| 1160 | #
|
| 1161 | # $(echo one two) -> @[echo one two]
|
| 1162 | # `echo one two` -> @[echo one two]
|
| 1163 | #
|
| 1164 | # ${var:-'the default'} -> @$(var or 'the default')
|
| 1165 | #
|
| 1166 | # $((1 + 2)) -> $(1 + 2) -- this is OK unquoted
|
| 1167 |
|
| 1168 | elif part0.tag() == word_part_e.BracedVarSub:
|
| 1169 | # Skip over quote
|
| 1170 | self.cursor.PrintUntil(dq_part.left)
|
| 1171 | self.cursor.SkipPast(dq_part.left)
|
| 1172 | self.DoWordPart(part0, local_symbols)
|
| 1173 | self.cursor.SkipPast(dq_part.right)
|
| 1174 | return
|
| 1175 |
|
| 1176 | elif part0.tag() == word_part_e.CommandSub:
|
| 1177 | self.cursor.PrintUntil(dq_part.left)
|
| 1178 | self.cursor.SkipPast(dq_part.left)
|
| 1179 | self.DoWordPart(part0, local_symbols)
|
| 1180 | self.cursor.SkipPast(dq_part.right)
|
| 1181 | return
|
| 1182 |
|
| 1183 | # TODO: 'foo'"bar" should be "foobar", etc.
|
| 1184 | # If any part is double quoted, you can always double quote the whole
|
| 1185 | # thing?
|
| 1186 | for part in node.parts:
|
| 1187 | self.DoWordPart(part, local_symbols)
|
| 1188 |
|
| 1189 | elif case(word_e.BracedTree):
|
| 1190 | # Not doing anything now
|
| 1191 | pass
|
| 1192 |
|
| 1193 | else:
|
| 1194 | raise AssertionError(node.__class__.__name__)
|
| 1195 |
|
| 1196 | def DoWordPart(self, node, local_symbols, quoted=False):
|
| 1197 | # type: (word_part_t, Dict[str, bool], bool) -> None
|
| 1198 |
|
| 1199 | left_tok = location.LeftTokenForWordPart(node)
|
| 1200 | if left_tok:
|
| 1201 | self.cursor.PrintUntil(left_tok)
|
| 1202 |
|
| 1203 | UP_node = node
|
| 1204 |
|
| 1205 | with tagswitch(node) as case:
|
| 1206 | if case(word_part_e.YshArrayLiteral,
|
| 1207 | word_part_e.InitializerLiteral, word_part_e.TildeSub,
|
| 1208 | word_part_e.ExtGlob):
|
| 1209 | pass
|
| 1210 |
|
| 1211 | elif case(word_part_e.EscapedLiteral):
|
| 1212 | node = cast(word_part.EscapedLiteral, UP_node)
|
| 1213 | if quoted:
|
| 1214 | pass
|
| 1215 | else:
|
| 1216 | # If unquoted \e, it should quoted instead. ' ' vs. \<invisible space>
|
| 1217 | # Hm is this necessary though? I think the only motivation is changing
|
| 1218 | # \{ and \( for macros. And ' ' to be readable/visible.
|
| 1219 | t = node.token
|
| 1220 | val = lexer.TokenSliceLeft(t, 1)
|
| 1221 | assert len(val) == 1, val
|
| 1222 | if val != '\n':
|
| 1223 | self.cursor.PrintUntil(t)
|
| 1224 | self.cursor.SkipPast(t)
|
| 1225 | self.f.write("'%s'" % val)
|
| 1226 |
|
| 1227 | elif case(word_part_e.Literal):
|
| 1228 | node = cast(Token, UP_node)
|
| 1229 | self.cursor.PrintIncluding(node)
|
| 1230 |
|
| 1231 | elif case(word_part_e.SingleQuoted):
|
| 1232 | node = cast(SingleQuoted, UP_node)
|
| 1233 |
|
| 1234 | # TODO:
|
| 1235 | # '\n' is '\\n'
|
| 1236 | # $'\n' is '\n'
|
| 1237 | # TODO: Should print until right_spid
|
| 1238 | # left_spid, right_spid = node.spids
|
| 1239 | self.cursor.PrintUntil(node.right)
|
| 1240 |
|
| 1241 | elif case(word_part_e.DoubleQuoted):
|
| 1242 | node = cast(DoubleQuoted, UP_node)
|
| 1243 | for part in node.parts:
|
| 1244 | self.DoWordPart(part, local_symbols, quoted=True)
|
| 1245 |
|
| 1246 | elif case(word_part_e.SimpleVarSub):
|
| 1247 | node = cast(SimpleVarSub, UP_node)
|
| 1248 |
|
| 1249 | op_id = node.tok.id
|
| 1250 |
|
| 1251 | if op_id == Id.VSub_DollarName:
|
| 1252 | self.cursor.PrintIncluding(node.tok)
|
| 1253 |
|
| 1254 | elif op_id == Id.VSub_Number:
|
| 1255 | self.cursor.PrintIncluding(node.tok)
|
| 1256 |
|
| 1257 | elif op_id == Id.VSub_At: # $@ -- handled quoted case above
|
| 1258 | self.f.write('$[join(ARGV)]')
|
| 1259 | self.cursor.SkipPast(node.tok)
|
| 1260 |
|
| 1261 | elif op_id == Id.VSub_Star: # $*
|
| 1262 | # PEDANTIC: Depends if quoted or unquoted
|
| 1263 | self.f.write('$[join(ARGV)]')
|
| 1264 | self.cursor.SkipPast(node.tok)
|
| 1265 |
|
| 1266 | elif op_id == Id.VSub_Pound: # $#
|
| 1267 | # len(ARGV) ?
|
| 1268 | self.f.write('$Argc')
|
| 1269 | self.cursor.SkipPast(node.tok)
|
| 1270 |
|
| 1271 | else:
|
| 1272 | pass
|
| 1273 |
|
| 1274 | elif case(word_part_e.BracedVarSub):
|
| 1275 | node = cast(BracedVarSub, UP_node)
|
| 1276 |
|
| 1277 | # NOTE: Why do we need this but we don't need it in command sub?
|
| 1278 | self.cursor.PrintUntil(node.left)
|
| 1279 |
|
| 1280 | if node.bracket_op:
|
| 1281 | # a[1]
|
| 1282 | # These two change the sigil! ${a[@]} is now @a!
|
| 1283 | # a[@]
|
| 1284 | # a[*]
|
| 1285 | pass
|
| 1286 |
|
| 1287 | if node.prefix_op:
|
| 1288 | # len()
|
| 1289 | pass
|
| 1290 | if node.suffix_op:
|
| 1291 | pass
|
| 1292 |
|
| 1293 | op_id = node.name_tok.id
|
| 1294 | if op_id == Id.VSub_QMark:
|
| 1295 | self.cursor.PrintIncluding(node.name_tok)
|
| 1296 |
|
| 1297 | self.cursor.PrintIncluding(node.right)
|
| 1298 |
|
| 1299 | elif case(word_part_e.CommandSub):
|
| 1300 | node = cast(CommandSub, UP_node)
|
| 1301 |
|
| 1302 | if node.left_token.id == Id.Left_Backtick:
|
| 1303 | self.cursor.PrintUntil(node.left_token)
|
| 1304 | self.f.write('$(')
|
| 1305 | self.cursor.SkipPast(node.left_token)
|
| 1306 |
|
| 1307 | self.DoCommand(node.child, local_symbols)
|
| 1308 |
|
| 1309 | # Skip over right `
|
| 1310 | self.cursor.SkipPast(node.right)
|
| 1311 | self.f.write(')')
|
| 1312 |
|
| 1313 | else:
|
| 1314 | self.cursor.PrintIncluding(node.right)
|
| 1315 |
|
| 1316 | else:
|
| 1317 | pass
|