| 1 | """YSH Str methods"""
|
| 2 |
|
| 3 | from __future__ import print_function
|
| 4 |
|
| 5 | from _devbuild.gen.syntax_asdl import loc_t, loc
|
| 6 | from _devbuild.gen.runtime_asdl import scope_e
|
| 7 | from _devbuild.gen.value_asdl import (value, value_e, value_t, eggex_ops,
|
| 8 | eggex_ops_t, RegexMatch, LeftName)
|
| 9 | from builtin import pure_ysh
|
| 10 | from core import error
|
| 11 | from core import state
|
| 12 | from core import vm
|
| 13 | from frontend import typed_args
|
| 14 | from mycpp import mops
|
| 15 | from mycpp.mylib import log, tagswitch
|
| 16 | from osh import string_ops
|
| 17 | from ysh import expr_eval
|
| 18 | from ysh import regex_translate
|
| 19 | from ysh import val_ops
|
| 20 |
|
| 21 | import libc
|
| 22 | from libc import REG_NOTBOL
|
| 23 |
|
| 24 | from typing import cast, Any, List, Optional, Tuple
|
| 25 |
|
| 26 | _ = log
|
| 27 |
|
| 28 |
|
| 29 | def _StrMatchStart(s, p):
|
| 30 | # type: (str, str) -> Tuple[bool, int, int]
|
| 31 | """Returns the range of bytes in 's' that match string pattern `p`. the
|
| 32 | pattern matches if 's' starts with all the characters in 'p'.
|
| 33 |
|
| 34 | The returned match result is the tuple "(matched, begin, end)". 'matched'
|
| 35 | is true if the pattern matched. 'begin' and 'end' give the half-open range
|
| 36 | "[begin, end)" of byte indices from 's' for the match, and are a valid but
|
| 37 | empty range if 'match' is false.
|
| 38 |
|
| 39 | Used for shell functions like 'trimStart' when trimming a prefix string.
|
| 40 | """
|
| 41 | if s.startswith(p):
|
| 42 | return (True, 0, len(p))
|
| 43 | else:
|
| 44 | return (False, 0, 0)
|
| 45 |
|
| 46 |
|
| 47 | def _StrMatchEnd(s, p):
|
| 48 | # type: (str, str) -> Tuple[bool, int, int]
|
| 49 | """Returns a match result for the bytes in 's' that match string pattern
|
| 50 | `p`. the pattern matches if 's' ends with all the characters in 'p'.
|
| 51 |
|
| 52 | The returned match result is the tuple "(matched, begin, end)". 'matched'
|
| 53 | is true if the pattern matched. 'begin' and 'end' give the half-open range
|
| 54 | "[begin, end)" of byte indices from 's' for the match, and are a valid but
|
| 55 | empty range if 'match' is false.
|
| 56 |
|
| 57 | Used for shell functions like 'trimEnd' when trimming a suffix string.
|
| 58 | """
|
| 59 | len_s = len(s)
|
| 60 | if s.endswith(p):
|
| 61 | return (True, len_s - len(p), len_s)
|
| 62 | else:
|
| 63 | return (False, len_s, len_s)
|
| 64 |
|
| 65 |
|
| 66 | def _EggexMatchCommon(s, p, ere, empty_p):
|
| 67 | # type: (str, value.Eggex, str, int) -> Tuple[bool, int, int]
|
| 68 | cflags = regex_translate.LibcFlags(p.canonical_flags)
|
| 69 | eflags = 0
|
| 70 | indices = libc.regex_search(ere, cflags, s, eflags)
|
| 71 | if indices is None:
|
| 72 | return (False, empty_p, empty_p)
|
| 73 |
|
| 74 | start = indices[0]
|
| 75 | end = indices[1]
|
| 76 |
|
| 77 | return (True, start, end)
|
| 78 |
|
| 79 |
|
| 80 | def _EggexMatchStart(s, p):
|
| 81 | # type: (str, value.Eggex) -> Tuple[bool, int, int]
|
| 82 | """Returns a match result for the bytes in 's' that match Eggex pattern
|
| 83 | `p` when constrained to match at the start of the string.
|
| 84 |
|
| 85 | Any capturing done by the Eggex pattern is ignored.
|
| 86 |
|
| 87 | The returned match result is the tuple "(matched, begin, end)". 'matched'
|
| 88 | is true if the pattern matched. 'begin' and 'end' give the half-open range
|
| 89 | "[begin, end)" of byte indices from 's' for the match, and are a valid but
|
| 90 | empty range if 'match' is false.
|
| 91 |
|
| 92 | Used for shell functions like 'trimStart' when trimming with an Eggex
|
| 93 | pattern.
|
| 94 | """
|
| 95 | ere = regex_translate.AsPosixEre(p)
|
| 96 | if not ere.startswith('^'):
|
| 97 | ere = '^' + ere
|
| 98 | return _EggexMatchCommon(s, p, ere, 0)
|
| 99 |
|
| 100 |
|
| 101 | def _EggexMatchEnd(s, p):
|
| 102 | # type: (str, value.Eggex) -> Tuple[bool, int, int]
|
| 103 | """Like _EggexMatchStart, but matches against the end of the
|
| 104 | string.
|
| 105 | """
|
| 106 | ere = regex_translate.AsPosixEre(p)
|
| 107 | if not ere.endswith('$'):
|
| 108 | ere = ere + '$'
|
| 109 | return _EggexMatchCommon(s, p, ere, len(s))
|
| 110 |
|
| 111 |
|
| 112 | START = 0b01
|
| 113 | END = 0b10
|
| 114 |
|
| 115 |
|
| 116 | class HasAffix(vm._Callable):
|
| 117 | """ Implements `startsWith()`, `endsWith()`. """
|
| 118 |
|
| 119 | def __init__(self, anchor):
|
| 120 | # type: (int) -> None
|
| 121 | assert anchor in (START, END), ("Anchor must be START or END")
|
| 122 | self.anchor = anchor
|
| 123 |
|
| 124 | def Call(self, rd):
|
| 125 | # type: (typed_args.Reader) -> value_t
|
| 126 | """
|
| 127 | string => startsWith(pattern_str) # => bool
|
| 128 | string => startsWith(pattern_eggex) # => bool
|
| 129 | string => endsWith(pattern_str) # => bool
|
| 130 | string => endsWith(pattern_eggex) # => bool
|
| 131 | """
|
| 132 |
|
| 133 | string = rd.PosStr()
|
| 134 | pattern_val = rd.PosValue()
|
| 135 | pattern_str = None # type: str
|
| 136 | pattern_eggex = None # type: value.Eggex
|
| 137 | with tagswitch(pattern_val) as case:
|
| 138 | if case(value_e.Eggex):
|
| 139 | pattern_eggex = cast(value.Eggex, pattern_val)
|
| 140 | elif case(value_e.Str):
|
| 141 | pattern_str = cast(value.Str, pattern_val).s
|
| 142 | else:
|
| 143 | raise error.TypeErr(pattern_val,
|
| 144 | 'expected pattern to be Eggex or Str',
|
| 145 | rd.LeftParenToken())
|
| 146 | rd.Done()
|
| 147 |
|
| 148 | matched = False
|
| 149 | try:
|
| 150 | if pattern_str is not None:
|
| 151 | if self.anchor & START:
|
| 152 | matched, _, _ = _StrMatchStart(string, pattern_str)
|
| 153 | else:
|
| 154 | matched, _, _ = _StrMatchEnd(string, pattern_str)
|
| 155 | else:
|
| 156 | assert pattern_eggex is not None
|
| 157 | if self.anchor & START:
|
| 158 | matched, _, _ = _EggexMatchStart(string, pattern_eggex)
|
| 159 | else:
|
| 160 | matched, _, _ = _EggexMatchEnd(string, pattern_eggex)
|
| 161 | except error.Strict as e:
|
| 162 | raise error.Expr(e.msg, e.location)
|
| 163 |
|
| 164 | return value.Bool(matched)
|
| 165 |
|
| 166 |
|
| 167 | class Trim(vm._Callable):
|
| 168 | """ Implements `trimStart()`, `trimEnd()`, and `trim()` """
|
| 169 |
|
| 170 | def __init__(self, anchor):
|
| 171 | # type: (int) -> None
|
| 172 | assert anchor in (START, END, START
|
| 173 | | END), ("Anchor must be START, END, or START|END")
|
| 174 | self.anchor = anchor
|
| 175 |
|
| 176 | def Call(self, rd):
|
| 177 | # type: (typed_args.Reader) -> value_t
|
| 178 | """
|
| 179 | string => trimStart() # => Str
|
| 180 | string => trimEnd() # => Str
|
| 181 | string => trim() # => Str
|
| 182 | string => trimStart(pattern_str) # => Str
|
| 183 | string => trimEnd(pattern_str) # => Str
|
| 184 | string => trim(pattern_str) # => Str
|
| 185 | string => trimStart(pattern_eggex) # => Str
|
| 186 | string => trimEnd(pattern_eggex) # => Str
|
| 187 | string => trim(pattern_eggex) # => Str
|
| 188 | """
|
| 189 |
|
| 190 | string = rd.PosStr()
|
| 191 | pattern_val = rd.OptionalValue()
|
| 192 | pattern_str = None # type: str
|
| 193 | pattern_eggex = None # type: value.Eggex
|
| 194 | if pattern_val:
|
| 195 | with tagswitch(pattern_val) as case:
|
| 196 | if case(value_e.Eggex):
|
| 197 | pattern_eggex = cast(value.Eggex, pattern_val)
|
| 198 | elif case(value_e.Str):
|
| 199 | pattern_str = cast(value.Str, pattern_val).s
|
| 200 | else:
|
| 201 | raise error.TypeErr(pattern_val,
|
| 202 | 'expected pattern to be Eggex or Str',
|
| 203 | rd.LeftParenToken())
|
| 204 | rd.Done()
|
| 205 |
|
| 206 | start = 0
|
| 207 | end = len(string)
|
| 208 | try:
|
| 209 | if pattern_str is not None:
|
| 210 | if self.anchor & START:
|
| 211 | _, _, start = _StrMatchStart(string, pattern_str)
|
| 212 | if self.anchor & END:
|
| 213 | _, end, _ = _StrMatchEnd(string, pattern_str)
|
| 214 | elif pattern_eggex is not None:
|
| 215 | if self.anchor & START:
|
| 216 | _, _, start = _EggexMatchStart(string, pattern_eggex)
|
| 217 | if self.anchor & END:
|
| 218 | _, end, _ = _EggexMatchEnd(string, pattern_eggex)
|
| 219 | else:
|
| 220 | if self.anchor & START:
|
| 221 | _, start = string_ops.StartsWithWhitespaceByteRange(string)
|
| 222 | if self.anchor & END:
|
| 223 | end, _ = string_ops.EndsWithWhitespaceByteRange(string)
|
| 224 | except error.Strict as e:
|
| 225 | raise error.Expr(e.msg, e.location)
|
| 226 |
|
| 227 | res = string[start:end]
|
| 228 | return value.Str(res)
|
| 229 |
|
| 230 |
|
| 231 | class Upper(vm._Callable):
|
| 232 |
|
| 233 | def __init__(self):
|
| 234 | # type: () -> None
|
| 235 | pass
|
| 236 |
|
| 237 | def Call(self, rd):
|
| 238 | # type: (typed_args.Reader) -> value_t
|
| 239 |
|
| 240 | string = rd.PosStr()
|
| 241 | rd.Done()
|
| 242 |
|
| 243 | res = string.upper()
|
| 244 | return value.Str(res)
|
| 245 |
|
| 246 |
|
| 247 | SEARCH = 0
|
| 248 | LEFT_MATCH = 1
|
| 249 |
|
| 250 |
|
| 251 | class SearchMatch(vm._Callable):
|
| 252 |
|
| 253 | def __init__(self, which_method):
|
| 254 | # type: (int) -> None
|
| 255 | self.which_method = which_method
|
| 256 |
|
| 257 | def Call(self, rd):
|
| 258 | # type: (typed_args.Reader) -> value_t
|
| 259 | """
|
| 260 | s => search(eggex, pos=0)
|
| 261 | """
|
| 262 | string = rd.PosStr()
|
| 263 |
|
| 264 | pattern = rd.PosValue() # Eggex or ERE Str
|
| 265 | with tagswitch(pattern) as case:
|
| 266 | if case(value_e.Eggex):
|
| 267 | eggex_val = cast(value.Eggex, pattern)
|
| 268 |
|
| 269 | # lazily converts to ERE
|
| 270 | ere = regex_translate.AsPosixEre(eggex_val)
|
| 271 | cflags = regex_translate.LibcFlags(eggex_val.canonical_flags)
|
| 272 | capture = eggex_ops.Yes(
|
| 273 | eggex_val.convert_funcs, eggex_val.convert_toks,
|
| 274 | eggex_val.capture_names) # type: eggex_ops_t
|
| 275 |
|
| 276 | elif case(value_e.Str):
|
| 277 | ere = cast(value.Str, pattern).s
|
| 278 | cflags = 0
|
| 279 | capture = eggex_ops.No
|
| 280 |
|
| 281 | else:
|
| 282 | # TODO: add method name to this error
|
| 283 | raise error.TypeErr(pattern, 'expected Eggex or Str',
|
| 284 | rd.LeftParenToken())
|
| 285 |
|
| 286 | # It's called 'pos', not 'start' like Python. Python has 2 kinds of
|
| 287 | # 'start' in its regex API, which can be confusing.
|
| 288 | pos = mops.BigTruncate(rd.NamedInt('pos', 0))
|
| 289 | rd.Done()
|
| 290 |
|
| 291 | # Make it anchored
|
| 292 | if self.which_method == LEFT_MATCH and not ere.startswith('^'):
|
| 293 | ere = '^' + ere
|
| 294 |
|
| 295 | if self.which_method == LEFT_MATCH:
|
| 296 | eflags = 0 # ^ matches beginning even if pos=5
|
| 297 | else:
|
| 298 | eflags = 0 if pos == 0 else REG_NOTBOL # ^ only matches when pos=0
|
| 299 |
|
| 300 | indices = libc.regex_search(ere, cflags, string, eflags, pos)
|
| 301 |
|
| 302 | if indices is None:
|
| 303 | return value.Null
|
| 304 |
|
| 305 | return RegexMatch(string, indices, capture)
|
| 306 |
|
| 307 |
|
| 308 | class ctx_EvalReplace(object):
|
| 309 | """For $0, $1, $2, $3, ... replacements in Str => replace()"""
|
| 310 |
|
| 311 | def __init__(self, mem, arg0, argv):
|
| 312 | # type: (state.Mem, str, Optional[List[str]]) -> None
|
| 313 | # argv will be None for Str => replace(Str, Expr)
|
| 314 | if argv is None:
|
| 315 | self.pushed_argv = False
|
| 316 | else:
|
| 317 | mem.argv_stack.append(state._ArgFrame(argv))
|
| 318 | self.pushed_argv = True
|
| 319 |
|
| 320 | # $0 needs to have lexical scoping. So we store it with other locals.
|
| 321 | # As "0" cannot be parsed as an lvalue, we can safely store arg0 there.
|
| 322 | assert mem.GetValue("0", scope_e.LocalOnly).tag() == value_e.Undef
|
| 323 | self.lval = LeftName("0", loc.Missing)
|
| 324 | mem.SetLocalName(self.lval, value.Str(arg0))
|
| 325 |
|
| 326 | self.mem = mem
|
| 327 |
|
| 328 | def __enter__(self):
|
| 329 | # type: () -> None
|
| 330 | pass
|
| 331 |
|
| 332 | def __exit__(self, type, value_, traceback):
|
| 333 | # type: (Any, Any, Any) -> None
|
| 334 | self.mem.SetLocalName(self.lval, value.Undef)
|
| 335 | if self.pushed_argv:
|
| 336 | self.mem.argv_stack.pop()
|
| 337 |
|
| 338 |
|
| 339 | class Replace(vm._Callable):
|
| 340 |
|
| 341 | def __init__(self, mem, expr_ev):
|
| 342 | # type: (state.Mem, expr_eval.ExprEvaluator) -> None
|
| 343 | self.mem = mem
|
| 344 | self.expr_ev = expr_ev
|
| 345 |
|
| 346 | def EvalSubstExpr(self, expr, blame_loc):
|
| 347 | # type: (value.Expr, loc_t) -> str
|
| 348 | res = self.expr_ev.EvalExpr(expr.e, blame_loc)
|
| 349 | if res.tag() == value_e.Str:
|
| 350 | return cast(value.Str, res).s
|
| 351 |
|
| 352 | raise error.TypeErr(res, "expected expr to eval to a Str", blame_loc)
|
| 353 |
|
| 354 | def Call(self, rd):
|
| 355 | # type: (typed_args.Reader) -> value_t
|
| 356 | """
|
| 357 | s => replace(string_val, subst_str, count=-1)
|
| 358 | s => replace(string_val, subst_expr, count=-1)
|
| 359 | s => replace(eggex_val, subst_str, count=-1)
|
| 360 | s => replace(eggex_val, subst_expr, count=-1)
|
| 361 |
|
| 362 | For count in [0, MAX_INT], there will be no more than count
|
| 363 | replacements. Any negative count should read as unset, and replace will
|
| 364 | replace all occurances of the pattern.
|
| 365 | """
|
| 366 | string = rd.PosStr()
|
| 367 |
|
| 368 | string_val = None # type: value.Str
|
| 369 | eggex_val = None # type: value.Eggex
|
| 370 | subst_str = None # type: value.Str
|
| 371 | subst_expr = None # type: value.Expr
|
| 372 |
|
| 373 | pattern = rd.PosValue()
|
| 374 | with tagswitch(pattern) as case:
|
| 375 | if case(value_e.Eggex):
|
| 376 | # HACK: mycpp will otherwise generate:
|
| 377 | # value::Eggex* eggex_val ...
|
| 378 | eggex_val_ = cast(value.Eggex, pattern)
|
| 379 | eggex_val = eggex_val_
|
| 380 |
|
| 381 | elif case(value_e.Str):
|
| 382 | string_val_ = cast(value.Str, pattern)
|
| 383 | string_val = string_val_
|
| 384 |
|
| 385 | else:
|
| 386 | raise error.TypeErr(pattern,
|
| 387 | 'expected pattern to be Eggex or Str',
|
| 388 | rd.LeftParenToken())
|
| 389 |
|
| 390 | subst = rd.PosValue()
|
| 391 | with tagswitch(subst) as case:
|
| 392 | if case(value_e.Str):
|
| 393 | subst_str_ = cast(value.Str, subst)
|
| 394 | subst_str = subst_str_
|
| 395 |
|
| 396 | elif case(value_e.Expr):
|
| 397 | subst_expr_ = cast(value.Expr, subst)
|
| 398 | subst_expr = subst_expr_
|
| 399 |
|
| 400 | else:
|
| 401 | raise error.TypeErr(subst,
|
| 402 | 'expected substitution to be Str or Expr',
|
| 403 | rd.LeftParenToken())
|
| 404 |
|
| 405 | count = mops.BigTruncate(rd.NamedInt("count", -1))
|
| 406 | rd.Done()
|
| 407 |
|
| 408 | if count == 0:
|
| 409 | return value.Str(string)
|
| 410 |
|
| 411 | if string_val:
|
| 412 | if subst_str:
|
| 413 | s = subst_str.s
|
| 414 | if subst_expr:
|
| 415 | # Eval with $0 set to string_val (the matched substring)
|
| 416 | with ctx_EvalReplace(self.mem, string_val.s, None):
|
| 417 | s = self.EvalSubstExpr(subst_expr, rd.LeftParenToken())
|
| 418 | assert s is not None
|
| 419 |
|
| 420 | result = string.replace(string_val.s, s, count)
|
| 421 |
|
| 422 | return value.Str(result)
|
| 423 |
|
| 424 | if eggex_val:
|
| 425 | ere = regex_translate.AsPosixEre(eggex_val)
|
| 426 | cflags = regex_translate.LibcFlags(eggex_val.canonical_flags)
|
| 427 |
|
| 428 | # Walk through the string finding all matches of the compiled ere.
|
| 429 | # Then, collect unmatched substrings and substitutions into the
|
| 430 | # `parts` list.
|
| 431 | pos = 0
|
| 432 | parts = [] # type: List[str]
|
| 433 | replace_count = 0
|
| 434 | while pos < len(string):
|
| 435 | indices = libc.regex_search(ere, cflags, string, 0, pos)
|
| 436 | if indices is None:
|
| 437 | break
|
| 438 |
|
| 439 | # Collect captures
|
| 440 | arg0 = None # type: str
|
| 441 | argv = [] # type: List[str]
|
| 442 | named_vars = [] # type: List[Tuple[str, value_t]]
|
| 443 | num_groups = len(indices) / 2
|
| 444 | for group in xrange(num_groups):
|
| 445 | start = indices[2 * group]
|
| 446 | end = indices[2 * group + 1]
|
| 447 | captured = string[start:end]
|
| 448 | val = value.Str(captured) # type: value_t
|
| 449 |
|
| 450 | if len(eggex_val.convert_funcs) and group != 0:
|
| 451 | convert_func = eggex_val.convert_funcs[group - 1]
|
| 452 | convert_tok = eggex_val.convert_toks[group - 1]
|
| 453 |
|
| 454 | if convert_func:
|
| 455 | val = self.expr_ev.CallConvertFunc(
|
| 456 | convert_func, val, convert_tok,
|
| 457 | rd.LeftParenToken())
|
| 458 |
|
| 459 | # $0, $1, $2 variables are argv values, which must be
|
| 460 | # strings. Furthermore, they can only be used in string
|
| 461 | # contexts
|
| 462 | # eg. "$[1]" != "$1".
|
| 463 | val_str = val_ops.Stringify(val, rd.LeftParenToken())
|
| 464 | if group == 0:
|
| 465 | arg0 = val_str
|
| 466 | else:
|
| 467 | argv.append(val_str)
|
| 468 |
|
| 469 | # $0 cannot be named
|
| 470 | if group != 0:
|
| 471 | name = eggex_val.capture_names[group - 2]
|
| 472 | if name is not None:
|
| 473 | named_vars.append((name, val))
|
| 474 |
|
| 475 | if subst_str:
|
| 476 | s = subst_str.s
|
| 477 | if subst_expr:
|
| 478 | with ctx_EvalReplace(self.mem, arg0, argv):
|
| 479 | with pure_ysh.ctx_Shvar(self.mem, named_vars):
|
| 480 | s = self.EvalSubstExpr(subst_expr,
|
| 481 | rd.LeftParenToken())
|
| 482 | assert s is not None
|
| 483 |
|
| 484 | start = indices[0]
|
| 485 | end = indices[1]
|
| 486 | parts.append(string[pos:start]) # Unmatched substring
|
| 487 | parts.append(s) # Replacement
|
| 488 | pos = end # Move to end of match
|
| 489 |
|
| 490 | replace_count += 1
|
| 491 | if count != -1 and replace_count == count:
|
| 492 | break
|
| 493 |
|
| 494 | parts.append(string[pos:]) # Remaining unmatched substring
|
| 495 |
|
| 496 | return value.Str("".join(parts))
|
| 497 |
|
| 498 | raise AssertionError()
|