| 1 | """
|
| 2 | lexer_def.py - Lexing for OSH, YSH, and J8 Notation.
|
| 3 |
|
| 4 | The OSH/YSH lexer has lexer modes, each with a regex -> Id mapping.
|
| 5 |
|
| 6 | After changing this file, run:
|
| 7 |
|
| 8 | build/py.sh all
|
| 9 |
|
| 10 | or at least:
|
| 11 |
|
| 12 | build/py.sh fastlex
|
| 13 |
|
| 14 | Input Handling
|
| 15 | --------------
|
| 16 |
|
| 17 | Every line is NUL terminated:
|
| 18 |
|
| 19 | 'one\n\0' 'last line\0'
|
| 20 |
|
| 21 | which means that no regexes below should match \0.
|
| 22 |
|
| 23 | For example, use [^'\0]+ instead of [^']+ .
|
| 24 |
|
| 25 | If this rule isn't followed, we would read uninitialized memory past the
|
| 26 | sentinel. Python's regex engine knows where the end of the input string is, so
|
| 27 | it doesn't require need a sentinel like \0.
|
| 28 |
|
| 29 | The frontend/lexer_gen.py generator adds a pattern mapping \0 to Id.Eol_Tok.
|
| 30 | """
|
| 31 |
|
| 32 | from _devbuild.gen.id_kind_asdl import Id, Id_t, Kind
|
| 33 | from _devbuild.gen.types_asdl import lex_mode_e
|
| 34 |
|
| 35 | from frontend import id_kind_def
|
| 36 |
|
| 37 | from typing import Tuple
|
| 38 |
|
| 39 | # Initialize spec that the lexer depends on.
|
| 40 | ID_SPEC = id_kind_def.IdSpec({}, {})
|
| 41 |
|
| 42 | id_kind_def.AddKinds(ID_SPEC)
|
| 43 | id_kind_def.AddBoolKinds(ID_SPEC) # must come second
|
| 44 | id_kind_def.SetupTestBuiltin(ID_SPEC, {}, {}, {})
|
| 45 |
|
| 46 |
|
| 47 | def C(pat, tok_type):
|
| 48 | # type: (str, Id_t) -> Tuple[bool, str, Id_t]
|
| 49 | """Lexer rule with a constant string, e.g. C('$*', VSub_Star)"""
|
| 50 | return (False, pat, tok_type)
|
| 51 |
|
| 52 |
|
| 53 | def R(pat, tok_type):
|
| 54 | # type: (str, Id_t) -> Tuple[bool, str, Id_t]
|
| 55 | """Lexer rule with a regex string, e.g. R('\$[0-9]', VSub_Number)"""
|
| 56 | return (True, pat, tok_type)
|
| 57 |
|
| 58 |
|
| 59 | # See unit tests in frontend/match_test.py.
|
| 60 | # We need the [^\0]* because the re2c translation assumes it's anchored like $.
|
| 61 | SHOULD_HIJACK_RE = r'#![^\0]*sh[ \t\r\n][^\0]*'
|
| 62 |
|
| 63 | # Separates words (\r it not whitespace here)
|
| 64 | _SIGNIFICANT_SPACE = R(r'[ \t]+', Id.WS_Space)
|
| 65 |
|
| 66 | _BACKSLASH = [
|
| 67 | # To be conservative, we could deny a set of chars similar to
|
| 68 | # _LITERAL_WHITELIST_REGEX, rather than allowing all the operator characters
|
| 69 | # like \( and \;.
|
| 70 | #
|
| 71 | # strict_backslash makes this stricter.
|
| 72 | R(r'\\[^\n\0]', Id.Lit_EscapedChar),
|
| 73 | C('\\\n', Id.Ignored_LineCont),
|
| 74 | ]
|
| 75 |
|
| 76 | # Only 4 characters are backslash escaped inside "".
|
| 77 | # https://www.gnu.org/software/bash/manual/bash.html#Double-Quotes
|
| 78 | _DQ_BACKSLASH = [
|
| 79 | R(r'\\[$`"\\]', Id.Lit_EscapedChar),
|
| 80 | C('\\', Id.Lit_BadBackslash), # syntax error in YSH, but NOT in OSH
|
| 81 | ]
|
| 82 |
|
| 83 | VAR_NAME_RE = r'[a-zA-Z_][a-zA-Z0-9_]*'
|
| 84 |
|
| 85 | # All Kind.VSub
|
| 86 | _VARS = [
|
| 87 | # Unbraced variables
|
| 88 | R(r'\$' + VAR_NAME_RE, Id.VSub_DollarName),
|
| 89 | R(r'\$[0-9]', Id.VSub_Number),
|
| 90 | C(r'$!', Id.VSub_Bang),
|
| 91 | C(r'$@', Id.VSub_At),
|
| 92 | C(r'$#', Id.VSub_Pound),
|
| 93 | C(r'$$', Id.VSub_Dollar),
|
| 94 | C(r'$*', Id.VSub_Star),
|
| 95 | C(r'$-', Id.VSub_Hyphen),
|
| 96 | C(r'$?', Id.VSub_QMark),
|
| 97 | ]
|
| 98 |
|
| 99 | # Kind.Left that are valid in double-quoted modes.
|
| 100 |
|
| 101 | _LEFT_SUBS = [
|
| 102 | C('`', Id.Left_Backtick),
|
| 103 | C('$(', Id.Left_DollarParen),
|
| 104 | C('${', Id.Left_DollarBrace),
|
| 105 | # Parse zsh syntax, but don't execute it.
|
| 106 | # The examples we've seen so far are like ${(%):-} and ${(m)
|
| 107 | R(r'\$\{\([^)\0]+\)', Id.Left_DollarBraceZsh),
|
| 108 | C('$((', Id.Left_DollarDParen),
|
| 109 | C('$[', Id.Left_DollarBracket),
|
| 110 | ]
|
| 111 |
|
| 112 | # Additional Kind.Left that are valid in unquoted modes.
|
| 113 | _LEFT_UNQUOTED = [
|
| 114 | C('"', Id.Left_DoubleQuote),
|
| 115 | C("'", Id.Left_SingleQuote),
|
| 116 | C('$"', Id.Left_DollarDoubleQuote),
|
| 117 | C("$'", Id.Left_DollarSingleQuote),
|
| 118 | ]
|
| 119 |
|
| 120 | _LEFT_PROCSUB = [
|
| 121 | C('<(', Id.Left_ProcSubIn),
|
| 122 | C('>(', Id.Left_ProcSubOut),
|
| 123 | ]
|
| 124 |
|
| 125 | # The regexes below are in Python syntax, but are translate to re2c syntax by
|
| 126 | # frontend/lexer_gen.py.
|
| 127 | #
|
| 128 | # http://re2c.org/manual/syntax/syntax.html
|
| 129 | # https://docs.python.org/2/library/re.html
|
| 130 | #
|
| 131 | # We use a limited set of constructs:
|
| 132 | # - + and * for repetition
|
| 133 | # - Character classes [] with simple ranges and negation
|
| 134 | # - Escapes like \n \0
|
| 135 |
|
| 136 | LEXER_DEF = {} # TODO: Should be a list so we enforce order.
|
| 137 |
|
| 138 | # Anything until the end of the line is a comment. Does not match the newline
|
| 139 | # itself. We want to switch modes and possibly process Op_Newline for here
|
| 140 | # docs, etc.
|
| 141 | LEXER_DEF[lex_mode_e.Comment] = [R(r'[^\n\0]*', Id.Ignored_Comment)]
|
| 142 |
|
| 143 | # A whitelist to make bigger Lit_Chars tokens. We don't want one byte at a time.
|
| 144 | #
|
| 145 | # The shell language says that "anything other byte" is a literal character --
|
| 146 | # for example, unquoted $ \ ! are literal, not a syntax error.
|
| 147 | #
|
| 148 | # That is, a literal is defined NEGATIVELY, for a single characters. But here
|
| 149 | # we define a SUBSET of literal chars POSITIVELY.
|
| 150 |
|
| 151 | # The range \x80-\xff makes sure that UTF-8 sequences are a single token.
|
| 152 | _LITERAL_WHITELIST_REGEX = r'[\x80-\xffa-zA-Z0-9_.\-]+'
|
| 153 |
|
| 154 | _UNQUOTED = _BACKSLASH + _LEFT_SUBS + _LEFT_UNQUOTED + _LEFT_PROCSUB + _VARS + [
|
| 155 | # NOTE: We could add anything 128 and above to this character class? So
|
| 156 | # utf-8 characters don't get split?
|
| 157 | R(_LITERAL_WHITELIST_REGEX, Id.Lit_Chars),
|
| 158 | C('~', Id.Lit_Tilde), # for tilde sub
|
| 159 | C('/', Id.Lit_Slash), # also for tilde sub
|
| 160 | C(':', Id.Lit_Colon), # for special PATH=a:~foo tilde detection
|
| 161 | C('$', Id.Lit_Dollar), # shopt -u parse_dollar
|
| 162 | C('#', Id.Lit_Pound), # For comments
|
| 163 | _SIGNIFICANT_SPACE,
|
| 164 | C('\n', Id.Op_Newline),
|
| 165 | C('&', Id.Op_Amp),
|
| 166 | C('|', Id.Op_Pipe),
|
| 167 | C('|&', Id.Op_PipeAmp),
|
| 168 | C('&&', Id.Op_DAmp),
|
| 169 | C('||', Id.Op_DPipe),
|
| 170 | C(';', Id.Op_Semi),
|
| 171 | # Case terminators
|
| 172 | C(';;', Id.Op_DSemi),
|
| 173 | C(';&', Id.Op_SemiAmp),
|
| 174 | C(';;&', Id.Op_DSemiAmp),
|
| 175 | C('(', Id.Op_LParen),
|
| 176 | C(')', Id.Op_RParen),
|
| 177 | R(r'[^\0]', Id.Lit_Other), # any other single char is a literal
|
| 178 | ]
|
| 179 |
|
| 180 | # In ShCommand and DBracket states.
|
| 181 | _EXTGLOB_BEGIN = [
|
| 182 | C(',(', Id.ExtGlob_Comma), # YSH synonym for @(...)
|
| 183 | C('@(', Id.ExtGlob_At),
|
| 184 | C('*(', Id.ExtGlob_Star),
|
| 185 | C('+(', Id.ExtGlob_Plus),
|
| 186 | C('?(', Id.ExtGlob_QMark),
|
| 187 | C('!(', Id.ExtGlob_Bang),
|
| 188 | ]
|
| 189 |
|
| 190 | KEYWORDS = [
|
| 191 | # NOTE: { is matched elsewhere
|
| 192 | C('[[', Id.KW_DLeftBracket),
|
| 193 | C('!', Id.KW_Bang),
|
| 194 | C('for', Id.KW_For),
|
| 195 | C('while', Id.KW_While),
|
| 196 | C('until', Id.KW_Until),
|
| 197 | C('do', Id.KW_Do),
|
| 198 | C('done', Id.KW_Done),
|
| 199 | C('in', Id.KW_In),
|
| 200 | C('case', Id.KW_Case),
|
| 201 | C('esac', Id.KW_Esac),
|
| 202 | C('if', Id.KW_If),
|
| 203 | C('fi', Id.KW_Fi),
|
| 204 | C('then', Id.KW_Then),
|
| 205 | C('else', Id.KW_Else),
|
| 206 | C('elif', Id.KW_Elif),
|
| 207 | C('function', Id.KW_Function),
|
| 208 | C('time', Id.KW_Time),
|
| 209 |
|
| 210 | # YSH
|
| 211 | C('const', Id.KW_Const), # maybe remove this
|
| 212 | C('var', Id.KW_Var),
|
| 213 | C('setvar', Id.KW_SetVar),
|
| 214 | C('setglobal', Id.KW_SetGlobal),
|
| 215 | C('call', Id.KW_Call),
|
| 216 | C('proc', Id.KW_Proc),
|
| 217 | C('typed', Id.KW_Typed),
|
| 218 | C('func', Id.KW_Func),
|
| 219 | ]
|
| 220 |
|
| 221 | # These are treated like builtins in bash, but keywords in OSH. However, we
|
| 222 | # maintain compatibility with bash for the 'type' builtin.
|
| 223 | CONTROL_FLOW = [
|
| 224 | C('break', Id.ControlFlow_Break),
|
| 225 | C('continue', Id.ControlFlow_Continue),
|
| 226 | C('return', Id.ControlFlow_Return),
|
| 227 | C('exit', Id.ControlFlow_Exit),
|
| 228 | ]
|
| 229 |
|
| 230 | # Used by ysh/grammar_gen.py too
|
| 231 | EXPR_WORDS = [
|
| 232 | C('null', Id.Expr_Null),
|
| 233 | C('true', Id.Expr_True),
|
| 234 | C('false', Id.Expr_False),
|
| 235 | C('and', Id.Expr_And),
|
| 236 | C('or', Id.Expr_Or),
|
| 237 | C('not', Id.Expr_Not),
|
| 238 | C('for', Id.Expr_For),
|
| 239 | C('is', Id.Expr_Is),
|
| 240 | C('in', Id.Expr_In),
|
| 241 | C('if', Id.Expr_If),
|
| 242 | C('else', Id.Expr_Else),
|
| 243 |
|
| 244 | # Unused: could be for function literals, although we also have
|
| 245 | # |x| x+1 lambdas
|
| 246 | C('func', Id.Expr_Func),
|
| 247 |
|
| 248 | # / <capture d+/
|
| 249 | C('capture', Id.Expr_Capture),
|
| 250 | # / <capture d+ as date> /
|
| 251 | C('as', Id.Expr_As),
|
| 252 | ]
|
| 253 |
|
| 254 | FD_VAR_NAME = r'\{' + VAR_NAME_RE + r'\}'
|
| 255 |
|
| 256 | # file descriptors can only have two digits, like mksh
|
| 257 | # dash/zsh/etc. can have one
|
| 258 | FD_NUM = r'[0-9]?[0-9]?'
|
| 259 |
|
| 260 | # These two can must be recognized in the ShCommand state, but can't nested
|
| 261 | # within [[.
|
| 262 | # Keywords have to be checked before _UNQUOTED so we get <KW_If "if"> instead
|
| 263 | # of <Lit_Chars "if">.
|
| 264 | LEXER_DEF[lex_mode_e.ShCommand] = [
|
| 265 | # These four are not allowed within [[, so they are in ShCommand but not
|
| 266 | # _UNQUOTED.
|
| 267 |
|
| 268 | # e.g. beginning of NAME=val, which will always be longer than
|
| 269 | # _LITERAL_WHITELIST_REGEX.
|
| 270 | R(VAR_NAME_RE + '\+?=', Id.Lit_VarLike),
|
| 271 | R(VAR_NAME_RE + '\[', Id.Lit_ArrayLhsOpen),
|
| 272 | R(r'\]\+?=', Id.Lit_ArrayLhsClose),
|
| 273 | C('((', Id.Op_DLeftParen),
|
| 274 |
|
| 275 | # For static globbing, and [] for array literals
|
| 276 | C('[', Id.Lit_LBracket), # e.g. A=(['x']=1)
|
| 277 | C(']', Id.Lit_RBracket), # e.g. *.[ch]
|
| 278 | # NOTE: Glob_Star and Glob_QMark are for dynamic parsing
|
| 279 | C('*', Id.Lit_Star),
|
| 280 | C('?', Id.Lit_QMark),
|
| 281 | C('###', Id.Lit_TPound), # like Lit_Pound, for doc comments
|
| 282 | C('...', Id.Lit_TDot), # ... for multiline commands
|
| 283 |
|
| 284 | # For brace expansion {a,b}
|
| 285 | C('{', Id.Lit_LBrace),
|
| 286 | C('}', Id.Lit_RBrace), # Also for var sub ${a}
|
| 287 | C(',', Id.Lit_Comma),
|
| 288 | C('=', Id.Lit_Equals), # for = f(x) and x = 1+2*3
|
| 289 | C('@', Id.Lit_At), # for detecting @[, @' etc. shopt -s parse_at_all
|
| 290 |
|
| 291 | # @array and @func(1, c)
|
| 292 | R('@' + VAR_NAME_RE, Id.Lit_Splice), # for YSH splicing
|
| 293 | C('@[', Id.Lit_AtLBracket), # @[split(x)]
|
| 294 | C('@{.', Id.Lit_AtLBraceDot), # for split builtin sub @{.myproc arg1}
|
| 295 | R(FD_NUM + r'<', Id.Redir_Less),
|
| 296 | R(FD_NUM + r'>', Id.Redir_Great),
|
| 297 | R(FD_NUM + r'<<', Id.Redir_DLess),
|
| 298 | R(FD_NUM + r'<<<', Id.Redir_TLess),
|
| 299 | R(FD_NUM + r'>>', Id.Redir_DGreat),
|
| 300 | R(FD_NUM + r'<<-', Id.Redir_DLessDash),
|
| 301 | R(FD_NUM + r'>&', Id.Redir_GreatAnd),
|
| 302 | R(FD_NUM + r'<&', Id.Redir_LessAnd),
|
| 303 | R(FD_NUM + r'<>', Id.Redir_LessGreat),
|
| 304 | R(FD_NUM + r'>\|', Id.Redir_Clobber),
|
| 305 | R(FD_VAR_NAME + r'<', Id.Redir_Less),
|
| 306 | R(FD_VAR_NAME + r'>', Id.Redir_Great),
|
| 307 | R(FD_VAR_NAME + r'<<', Id.Redir_DLess),
|
| 308 | R(FD_VAR_NAME + r'<<<', Id.Redir_TLess),
|
| 309 | R(FD_VAR_NAME + r'>>', Id.Redir_DGreat),
|
| 310 | R(FD_VAR_NAME + r'<<-', Id.Redir_DLessDash),
|
| 311 | R(FD_VAR_NAME + r'>&', Id.Redir_GreatAnd),
|
| 312 | R(FD_VAR_NAME + r'<&', Id.Redir_LessAnd),
|
| 313 | R(FD_VAR_NAME + r'<>', Id.Redir_LessGreat),
|
| 314 | R(FD_VAR_NAME + r'>\|', Id.Redir_Clobber),
|
| 315 |
|
| 316 | # No leading descriptor (2 is implied)
|
| 317 | C(r'&>', Id.Redir_AndGreat),
|
| 318 | C(r'&>>', Id.Redir_AndDGreat),
|
| 319 | ] + KEYWORDS + CONTROL_FLOW + _UNQUOTED + _EXTGLOB_BEGIN
|
| 320 |
|
| 321 | # Preprocessing before ShCommand
|
| 322 | LEXER_DEF[lex_mode_e.Backtick] = [
|
| 323 | C(r'`', Id.Backtick_Right),
|
| 324 | # A backslash, and then $ or ` or \
|
| 325 | R(r'\\[$`\\]', Id.Backtick_Quoted),
|
| 326 | # \" treated specially, depending on whether bacticks are double-quoted!
|
| 327 | R(r'\\"', Id.Backtick_DoubleQuote),
|
| 328 | R(r'[^`\\\0]+', Id.Backtick_Other), # contiguous run of literals
|
| 329 | R(r'[^\0]', Id.Backtick_Other), # anything else
|
| 330 | ]
|
| 331 |
|
| 332 | # DBRACKET: can be like ShCommand, except:
|
| 333 | # - Don't really need redirects either... Redir_Less could be Op_Less
|
| 334 | # - Id.Op_DLeftParen can't be nested inside.
|
| 335 | LEXER_DEF[lex_mode_e.DBracket] = [
|
| 336 | C(']]', Id.Lit_DRightBracket),
|
| 337 | # Must be KW and not Op, because we can have stuff like [[ $foo == !* ]]
|
| 338 | # in addition to [[ ! a && b ]]
|
| 339 | C('!', Id.KW_Bang),
|
| 340 | C('<', Id.Op_Less),
|
| 341 | C('>', Id.Op_Great),
|
| 342 | ] + ID_SPEC.LexerPairs(Kind.BoolUnary) + \
|
| 343 | ID_SPEC.LexerPairs(Kind.BoolBinary) + \
|
| 344 | _UNQUOTED + _EXTGLOB_BEGIN
|
| 345 |
|
| 346 | # Inside an extended glob, most characters are literals, including spaces and
|
| 347 | # punctuation. We also accept \, $var, ${var}, "", etc. They can also be
|
| 348 | # nested, so _EXTGLOB_BEGIN appears here.
|
| 349 | #
|
| 350 | # Example: echo @(<> <>|&&|'foo'|$bar)
|
| 351 | LEXER_DEF[lex_mode_e.ExtGlob] = \
|
| 352 | _BACKSLASH + _LEFT_SUBS + _LEFT_UNQUOTED + _VARS + _EXTGLOB_BEGIN + [
|
| 353 | R(r'[^\\$`"\'|)@*+!?\0]+', Id.Lit_Chars),
|
| 354 | C('|', Id.Op_Pipe),
|
| 355 | C(')', Id.Op_RParen), # maybe be translated to Id.ExtGlob_RParen
|
| 356 | R(r'[^\0]', Id.Lit_Other), # everything else is literal
|
| 357 | ]
|
| 358 |
|
| 359 | # Notes on BASH_REGEX states
|
| 360 | #
|
| 361 | # From bash manual:
|
| 362 | #
|
| 363 | # - Any part of the pattern may be quoted to force the quoted portion to be
|
| 364 | # matched as a string.
|
| 365 | # - Bracket expressions in regular expressions must be treated carefully, since
|
| 366 | # normal quoting characters lose their meanings between brackets.
|
| 367 | # - If the pattern is stored in a shell variable, quoting the variable
|
| 368 | # expansion forces the entire pattern to be matched as a string.
|
| 369 | #
|
| 370 | # Is there a re.escape function? It's just like EscapeGlob and UnescapeGlob.
|
| 371 | #
|
| 372 | # TODO: For testing, write a script to extract and save regexes... and compile
|
| 373 | # them with regcomp. I've only seen constant regexes.
|
| 374 | #
|
| 375 | # bash code: ( | ) are special
|
| 376 |
|
| 377 | LEXER_DEF[lex_mode_e.BashRegex] = _LEFT_SUBS + _LEFT_UNQUOTED + _VARS + [
|
| 378 | # Like lex_mode_e.ShCommand
|
| 379 | R(_LITERAL_WHITELIST_REGEX, Id.Lit_Chars),
|
| 380 |
|
| 381 | # Tokens for Tilde sub. bash weirdness: RHS of [[ x =~ ~ ]] is expanded
|
| 382 | C('~', Id.Lit_Tilde),
|
| 383 | C('/', Id.Lit_Slash),
|
| 384 |
|
| 385 | # Id.WS_Space delimits words. In lex_mode_e.BashRegexFakeInner, we
|
| 386 | # translate them to Id.Lit_Chars.
|
| 387 | _SIGNIFICANT_SPACE,
|
| 388 |
|
| 389 | # Analogous to Id.ExtGlob_* - we need to change lexer modes when we hit this
|
| 390 | C('(', Id.BashRegex_LParen),
|
| 391 |
|
| 392 | # Not special, this is like lex_mode_e.Outer
|
| 393 | C(')', Id.Op_RParen),
|
| 394 |
|
| 395 | # Copied and adapted from _UNQUOTED
|
| 396 | # \n & ; < > are parse errors OUTSIDE a group [[ s =~ ; ]]
|
| 397 | # but become allowed INSIDE a group [[ s =~ (;) ]]
|
| 398 | C('\n', Id.BashRegex_AllowedInParens),
|
| 399 | C('&', Id.BashRegex_AllowedInParens),
|
| 400 | C(';', Id.BashRegex_AllowedInParens),
|
| 401 | C('>', Id.BashRegex_AllowedInParens),
|
| 402 | C('<', Id.BashRegex_AllowedInParens),
|
| 403 |
|
| 404 | # e.g. | is Id.Lit_Other, not pipe operator
|
| 405 | R(r'[^\0]', Id.Lit_Other), # like _UNQUOTED, any other byte is literal
|
| 406 | ] + _BACKSLASH # These have to come after RegexMeta
|
| 407 |
|
| 408 | LEXER_DEF[lex_mode_e.DQ] = _DQ_BACKSLASH + [
|
| 409 | C('\\\n', Id.Ignored_LineCont),
|
| 410 | ] + _LEFT_SUBS + _VARS + [
|
| 411 | R(r'[^$`"\0\\]+', Id.Lit_Chars), # matches a line at most
|
| 412 | C('$', Id.Lit_Dollar), # completion of var names relies on this
|
| 413 | # NOTE: When parsing here doc line, this token doesn't end it.
|
| 414 | C('"', Id.Right_DoubleQuote),
|
| 415 | ]
|
| 416 |
|
| 417 | _VS_ARG_COMMON = [
|
| 418 | C('/', Id.Lit_Slash), # for patsub (not Id.VOp2_Slash)
|
| 419 | C('#', Id.Lit_Pound), # for patsub prefix (not Id.VOp1_Pound)
|
| 420 | C('%', Id.Lit_Percent), # for patsdub suffix (not Id.VOp1_Percent)
|
| 421 | C('}', Id.Right_DollarBrace), # For var sub "${a}"
|
| 422 | C('$', Id.Lit_Dollar), # completion of var names relies on this
|
| 423 | ]
|
| 424 |
|
| 425 | # We don't execute zsh var subs, but to find the closing } properly, we need to
|
| 426 | # to recognize \} and '}' and "}" $'}' etc.
|
| 427 | LEXER_DEF[lex_mode_e.VSub_Zsh] = \
|
| 428 | _BACKSLASH + _LEFT_SUBS + _LEFT_UNQUOTED + _LEFT_PROCSUB + \
|
| 429 | [
|
| 430 | C('}', Id.Right_DollarBrace), # For var sub "${a}"
|
| 431 | R(r'[^\0]', Id.Lit_Other), # e.g. "$", must be last
|
| 432 | ]
|
| 433 |
|
| 434 | # Kind.{Lit,Ignored,VSub,Left,Right,Eof}
|
| 435 | LEXER_DEF[lex_mode_e.VSub_ArgUnquoted] = \
|
| 436 | _BACKSLASH + _VS_ARG_COMMON + _LEFT_SUBS + _LEFT_UNQUOTED + _LEFT_PROCSUB + \
|
| 437 | _VARS + _EXTGLOB_BEGIN + [
|
| 438 |
|
| 439 | # Token for Tilde sub
|
| 440 | C('~', Id.Lit_Tilde),
|
| 441 |
|
| 442 | # - doesn't match ~ for tilde sub
|
| 443 | # - doesn't match < and > so it doesn't eat <()
|
| 444 | # - doesn't match @ ! ? + * so it doesn't eat _EXTGLOB_BEGIN -- ( alone it
|
| 445 | # not enough
|
| 446 | R(r'[^$`~/}"\'\0\\#%<>@!?+*]+', Id.Lit_Chars),
|
| 447 | R(r'[^\0]', Id.Lit_Other), # e.g. "$", must be last
|
| 448 | ]
|
| 449 |
|
| 450 | # Kind.{Lit,Ignored,VSub,Left,Right,Eof}
|
| 451 | LEXER_DEF[lex_mode_e.VSub_ArgDQ] = \
|
| 452 | _DQ_BACKSLASH + _VS_ARG_COMMON + _LEFT_SUBS + _VARS + [
|
| 453 |
|
| 454 | C(r'\}', Id.Lit_EscapedChar), # For "${var-\}}"
|
| 455 |
|
| 456 | R(r'[^$`/}"\0\\#%]+', Id.Lit_Chars), # matches a line at most
|
| 457 |
|
| 458 | # Weird wart: even in double quoted state, double quotes are allowed
|
| 459 | C('"', Id.Left_DoubleQuote),
|
| 460 |
|
| 461 | # Another weird wart of bash/mksh: $'' is recognized but NOT ''!
|
| 462 | C("$'", Id.Left_DollarSingleQuote),
|
| 463 | ]
|
| 464 |
|
| 465 | # NOTE: Id.Ignored_LineCont is NOT supported in SQ state, as opposed to DQ
|
| 466 | # state.
|
| 467 | LEXER_DEF[lex_mode_e.SQ_Raw] = [
|
| 468 | R(r"[^'\0]+", Id.Lit_Chars), # matches a line at most
|
| 469 | C("'", Id.Right_SingleQuote),
|
| 470 | ]
|
| 471 |
|
| 472 | # The main purpose for EXPR_CHARS is in regex literals, e.g. [a-z \t \n].
|
| 473 | #
|
| 474 | # In YSH expressions, Chars are code point integers, so \u{1234} is the same as
|
| 475 | # 0x1234. And \0 is 0x0.
|
| 476 |
|
| 477 | # In Python:
|
| 478 | # chr(0x00012345) == u'\U00012345'
|
| 479 | #
|
| 480 | # In YSH:
|
| 481 | # 0x00012345 == \u{12345}
|
| 482 | # chr(0x00012345) == chr(\u{12345}) == $'\u{012345}'
|
| 483 |
|
| 484 | _U_BRACED_CHAR = R(r'\\[uU]\{[0-9a-fA-F]{1,6}\}', Id.Char_UBraced)
|
| 485 |
|
| 486 | _X_CHAR_LOOSE = R(r'\\x[0-9a-fA-F]{1,2}', Id.Char_Hex) # bash
|
| 487 | _CHAR_YHEX = R(r'\\y[0-9a-fA-F]{2}', Id.Char_YHex) # \yff - J8 only
|
| 488 |
|
| 489 | _U4_CHAR_LOOSE = R(r'\\u[0-9a-fA-F]{1,4}', Id.Char_Unicode4) # bash
|
| 490 |
|
| 491 | _U4_CHAR_STRICT = R(r'\\u[0-9a-fA-F]{4}', Id.Char_Unicode4) # JSON-only
|
| 492 |
|
| 493 | #_JSON_ONE_CHAR = R(r'\\[\\"/bfnrt]', Id.Char_OneChar)
|
| 494 | EXPR_CHARS = [
|
| 495 | # Allow same backslash escapes as J8 strings, except;
|
| 496 | # - legacy \b \f
|
| 497 | # - unnecessary \/
|
| 498 | #
|
| 499 | # Note that \0 should be written \y00.
|
| 500 | R(r'''\\[\\"'nrt]''', Id.Char_OneChar),
|
| 501 | _CHAR_YHEX,
|
| 502 |
|
| 503 | # Eggex. This is a LITERAL translation to \xff in ERE? So it's not \yff
|
| 504 | # It doesn't have semantics; it's just syntax.
|
| 505 | R(r'\\x[0-9a-fA-F]{2}', Id.Char_Hex),
|
| 506 | _U_BRACED_CHAR,
|
| 507 | ]
|
| 508 |
|
| 509 | # Shared between echo -e and $''.
|
| 510 | _C_STRING_COMMON = [
|
| 511 |
|
| 512 | # \x6 is valid in bash
|
| 513 | _X_CHAR_LOOSE,
|
| 514 | _U4_CHAR_LOOSE,
|
| 515 | R(r'\\U[0-9a-fA-F]{1,8}', Id.Char_Unicode8),
|
| 516 | R(r'\\[0abeEfrtnv\\]', Id.Char_OneChar),
|
| 517 |
|
| 518 | # e.g. \A is not an escape, and \x doesn't match a hex escape. We allow it,
|
| 519 | # but a lint tool could warn about it.
|
| 520 | C('\\', Id.Unknown_Backslash),
|
| 521 | ]
|
| 522 |
|
| 523 | ECHO_E_DEF = _C_STRING_COMMON + [
|
| 524 | # Note: tokens above \0377 can either be truncated or be flagged a syntax
|
| 525 | # error in strict mode.
|
| 526 | R(r'\\0[0-7]{1,3}', Id.Char_Octal4),
|
| 527 | C(r'\c', Id.Char_Stop),
|
| 528 |
|
| 529 | # e.g. 'foo', anything that's not a backslash escape
|
| 530 | R(r'[^\\\0]+', Id.Lit_Chars),
|
| 531 | ]
|
| 532 |
|
| 533 | # https://json.org/
|
| 534 |
|
| 535 | # Note that [0-9] has to come second, because Python chooses the first match.
|
| 536 | _JSON_INT = r'-?([1-9][0-9]*|[0-9])' # Numbers can't start with leading 0
|
| 537 | _JSON_FRACTION = r'(\.[0-9]+)?'
|
| 538 | _JSON_EXP = r'([eE][-+]?[0-9]+)?'
|
| 539 |
|
| 540 | # R5RS extended alphabetic characters
|
| 541 | # https://groups.csail.mit.edu/mac/ftpdir/scheme-reports/r5rs-html/r5rs_4.html
|
| 542 | #
|
| 543 | # ! $ % & * + - . / : < = > ? @ ^ _ ~
|
| 544 |
|
| 545 | # Description from Guile Scheme - https://www.gnu.org/software/guile/manual/html_node/Symbol-Read-Syntax.html
|
| 546 | #
|
| 547 | # "The read syntax for a symbol is a sequence of letters, digits, and extended
|
| 548 | # alphabetic characters, beginning with a character that cannot begin a
|
| 549 | # number. In addition, the special cases of +, -, and ... are read as symbols
|
| 550 | # even though numbers can begin with +, - or ."
|
| 551 | #
|
| 552 | # (They should have used regular languages!)
|
| 553 |
|
| 554 | # We take out $ and @ for our splicing syntax, i.e. $unquote and
|
| 555 | # @unquote-splicing. And : for now because we use it for name:value.
|
| 556 |
|
| 557 | # Also note Scheme allows |a b| for symbols with funny chars, and Guile scheme
|
| 558 | # allows #{a b}#. We could use `a b` or (symbol "a b").
|
| 559 |
|
| 560 | J8_SYMBOL_CHARS = r'!%&*+./<=>?^_~-' # - is last for regex char class
|
| 561 |
|
| 562 | # yapf: disable
|
| 563 | J8_SYMBOL_RE = (
|
| 564 | r'[a-zA-Z' + J8_SYMBOL_CHARS + ']' +
|
| 565 | r'[a-zA-Z0-9' + J8_SYMBOL_CHARS + ']*')
|
| 566 | # yapf: enable
|
| 567 |
|
| 568 | _J8_LEFT = [
|
| 569 | C('"', Id.Left_DoubleQuote), # JSON string
|
| 570 | C('j"', Id.Left_JDoubleQuote), # JSON string with explicit J8 prefix
|
| 571 | # Three left quotes that are J8 only
|
| 572 | C("u'", Id.Left_USingleQuote), # unicode string
|
| 573 | C("'", Id.Left_USingleQuote), # '' is alias for u'' in data, not in code
|
| 574 | C("b'", Id.Left_BSingleQuote), # byte string
|
| 575 | ]
|
| 576 |
|
| 577 | J8_DEF = _J8_LEFT + [
|
| 578 | C('[', Id.J8_LBracket),
|
| 579 | C(']', Id.J8_RBracket),
|
| 580 | C('{', Id.J8_LBrace),
|
| 581 | C('}', Id.J8_RBrace),
|
| 582 | C('(', Id.J8_LParen), # NIL8 only
|
| 583 | C(')', Id.J8_RParen), # NIL8 only
|
| 584 | C(',', Id.J8_Comma),
|
| 585 | C(':', Id.J8_Colon),
|
| 586 | C('null', Id.J8_Null),
|
| 587 | C('true', Id.J8_Bool),
|
| 588 | C('false', Id.J8_Bool),
|
| 589 | R(_JSON_INT, Id.J8_Int),
|
| 590 | R(_JSON_INT + _JSON_FRACTION + _JSON_EXP, Id.J8_Float),
|
| 591 |
|
| 592 | # Identifier names come AFTER null true false.
|
| 593 | # - Happens to be the same as shell identifier # names.
|
| 594 | # - Note that JS allows $ as an identifier, but we don't.
|
| 595 | # - Used for dict keys / NIL8 field names.
|
| 596 | R(VAR_NAME_RE, Id.J8_Identifier),
|
| 597 |
|
| 598 | # Symbol is a SUPERSET of Identifier. The first word in NIL8 can be can
|
| 599 | # be either Symbol or plain Identifier, but field names can only be
|
| 600 | # Identifier. JSON8 only has Identifier.
|
| 601 | #R(J8_SYMBOL_RE, Id.J8_Symbol), # NIL8 only
|
| 602 | R(r'[~!@$%^&*+=|;./<>?-]+', Id.J8_Operator), # NIL8 only
|
| 603 | R(r'[ \r\t]+', Id.Ignored_Space),
|
| 604 | # A separate token, to count lines for error messages
|
| 605 | C('\n', Id.Ignored_Newline),
|
| 606 | # comment is # until end of line
|
| 607 | # // comments are JavaScript style, but right now we might want them as
|
| 608 | # symbols?
|
| 609 | R(r'#[^\n\0]*', Id.Ignored_Comment), # J8 only (JSON8, NIL8)
|
| 610 |
|
| 611 | # This will reject ASCII control chars
|
| 612 | R(r'[^\0]', Id.Unknown_Tok),
|
| 613 | ]
|
| 614 |
|
| 615 | # Exclude control characters 0x00-0x1f, aka 0-31 in J8 data only (not YSH code)
|
| 616 | _ASCII_CONTROL = R(r'[\x01-\x1F]', Id.Char_AsciiControl)
|
| 617 |
|
| 618 | J8_LINES_DEF = _J8_LEFT + [
|
| 619 | # not sure if we want \r here - same with lex_mode_e.Expr
|
| 620 | R(r'[ \r\t]+', Id.WS_Space),
|
| 621 | R(r'[\n]', Id.J8_Newline),
|
| 622 |
|
| 623 | # doesn't match \t, which means tabs are allowed in the middle of unquoted
|
| 624 | # lines
|
| 625 | _ASCII_CONTROL,
|
| 626 |
|
| 627 | # not space or ' or " or ASCII control or EOF
|
| 628 | R(r'''[^ \t\r\n'"\x00-\x1F]+''', Id.Lit_Chars),
|
| 629 | ]
|
| 630 |
|
| 631 | # https://json.org list of chars, plus '
|
| 632 | _JSON_ONE_CHAR = R(r'\\[\\"/bfnrt]', Id.Char_OneChar)
|
| 633 |
|
| 634 | # b'' u'' strings - what's common between code and data.
|
| 635 | _J8_STR_COMMON = [
|
| 636 | C("'", Id.Right_SingleQuote), # end for J8
|
| 637 | _JSON_ONE_CHAR,
|
| 638 | C("\\'", Id.Char_OneChar), # since ' ends, allow \'
|
| 639 | _CHAR_YHEX,
|
| 640 | _U_BRACED_CHAR, # \u{123456} - J8 only
|
| 641 |
|
| 642 | # osh/word_parse.py relies on this. It has to be consistent with $''
|
| 643 | # lexing, which uses _C_STRING_COMMON
|
| 644 | C('\\', Id.Unknown_Backslash),
|
| 645 | ]
|
| 646 |
|
| 647 | # Lexer for J8 strings in CODE.
|
| 648 | LEXER_DEF[lex_mode_e.J8_Str] = _J8_STR_COMMON + [
|
| 649 | # Don't produce Char_AsciiControl tokens - that's only for data
|
| 650 |
|
| 651 | # will match invalid UTF-8 - we have a separate validation step
|
| 652 | R(r"[^\\'\0]+", Id.Lit_Chars),
|
| 653 | ]
|
| 654 |
|
| 655 | # Lexer for J8 string data.
|
| 656 | # ASCII control characters are disallowed in DATA, but not CODE!
|
| 657 | J8_STR_DEF = _J8_STR_COMMON + [
|
| 658 | _ASCII_CONTROL,
|
| 659 | # will match invalid UTF-8 - we have a separate validation step
|
| 660 | R(r"[^\\'\x00-\x1F]+", Id.Lit_Chars),
|
| 661 | ]
|
| 662 |
|
| 663 | # Lexer for JSON string data - e.g. "json \" \u1234"
|
| 664 | JSON_STR_DEF = [
|
| 665 | C('"', Id.Right_DoubleQuote), # end for JSON
|
| 666 | _JSON_ONE_CHAR,
|
| 667 | _U4_CHAR_STRICT, # \u1234 - JSON only
|
| 668 |
|
| 669 | # High surrogate [\uD800, \uDC00)
|
| 670 | # Low surrogate [\uDC00, \uE000)
|
| 671 | # This pattern makes it easier to decode. Unpaired surrogates because Id.Char_Unicode4.
|
| 672 | R(
|
| 673 | r'\\u[dD][89aAbB][0-9a-fA-F][0-9a-fA-F]\\u[dD][cCdDeEfF][0-9a-fA-F][0-9a-fA-F]',
|
| 674 | Id.Char_SurrogatePair),
|
| 675 | C('\\', Id.Unknown_Backslash), # e.g. the \ before bad \z
|
| 676 | _ASCII_CONTROL,
|
| 677 |
|
| 678 | # Note: This will match INVALID UTF-8. UTF-8 validation is another step.
|
| 679 | R(r'[^\\"\x00-\x1F]+', Id.Lit_Chars),
|
| 680 | ]
|
| 681 |
|
| 682 | _WHITESPACE = r'[ \t\r\n]*' # ASCII whitespace doesn't have legacy \f \v
|
| 683 |
|
| 684 | SH_NUMBER_DEF = [
|
| 685 | R('0', Id.ShNumber_Dec), # not octal
|
| 686 | R(r'[1-9][0-9]*', Id.ShNumber_Dec),
|
| 687 | R(r'0[0-7]+', Id.ShNumber_Oct),
|
| 688 | R(r'0x[0-9A-Fa-f]+', Id.ShNumber_Hex),
|
| 689 | R(r'[1-9][0-9]*#[0-9a-zA-Z@_]+', Id.ShNumber_BaseN),
|
| 690 | R(r'[^\0]', Id.Unknown_Tok), # any other char
|
| 691 | ]
|
| 692 |
|
| 693 | OCTAL3_RE = r'\\[0-7]{1,3}'
|
| 694 |
|
| 695 | # https://www.gnu.org/software/bash/manual/html_node/Controlling-the-PromptEvaluator.html#Controlling-the-PromptEvaluator
|
| 696 | PS1_DEF = [
|
| 697 | R(OCTAL3_RE, Id.PS_Octal3),
|
| 698 | R(r'\\[adehHjlnrstT@AuvVwW!#$\\]', Id.PS_Subst),
|
| 699 | # \D{%H:%M} strftime format
|
| 700 | R(r'\\D\{[^}\0]*\}', Id.PS_Subst),
|
| 701 | C(r'\[', Id.PS_LBrace), # non-printing
|
| 702 | C(r'\]', Id.PS_RBrace),
|
| 703 | R(r'[^\\\0]+', Id.PS_Literals),
|
| 704 | # e.g. \x is not a valid escape.
|
| 705 | C('\\', Id.PS_BadBackslash),
|
| 706 | ]
|
| 707 |
|
| 708 | # NOTE: Id.Ignored_LineCont is also not supported here, even though the whole
|
| 709 | # point of it is that supports other backslash escapes like \n! It just
|
| 710 | # becomes a regular backslash.
|
| 711 | LEXER_DEF[lex_mode_e.SQ_C] = _C_STRING_COMMON + [
|
| 712 | # Weird special case matching bash: backslash that ends a line. We emit
|
| 713 | # this token literally in OSH, but disable it in YSH.
|
| 714 | C('\\\n', Id.Unknown_Backslash),
|
| 715 |
|
| 716 | # Silly difference! In echo -e, the syntax is \0377, but here it's $'\377',
|
| 717 | # with no leading 0.
|
| 718 | R(OCTAL3_RE, Id.Char_Octal3),
|
| 719 |
|
| 720 | # ' and " are escaped in $'' mode, but not echo -e.
|
| 721 | C(r"\'", Id.Char_OneChar),
|
| 722 | C(r'\"', Id.Char_OneChar),
|
| 723 |
|
| 724 | # e.g. 'foo', anything that's not a backslash escape or '
|
| 725 | R(r"[^\\'\0]+", Id.Lit_Chars),
|
| 726 | C("'", Id.Right_SingleQuote),
|
| 727 | ]
|
| 728 |
|
| 729 | LEXER_DEF[lex_mode_e.PrintfOuter] = _C_STRING_COMMON + [
|
| 730 | R(OCTAL3_RE, Id.Char_Octal3),
|
| 731 | R(r"[^%\\\0]+", Id.Lit_Chars),
|
| 732 | C('%%', Id.Format_EscapedPercent),
|
| 733 | C('%', Id.Format_Percent),
|
| 734 | ]
|
| 735 |
|
| 736 | # Maybe: bash also supports %(strftime)T
|
| 737 | LEXER_DEF[lex_mode_e.PrintfPercent] = [
|
| 738 | # Flags
|
| 739 | R('[- +#]', Id.Format_Flag),
|
| 740 | C('0', Id.Format_Zero),
|
| 741 | R('[1-9][0-9]*', Id.Format_Num),
|
| 742 | C('*', Id.Format_Star),
|
| 743 | C('.', Id.Format_Dot),
|
| 744 | # We support dsq. The others we parse to display an error message.
|
| 745 | R('[disqbcouxXeEfFgG]', Id.Format_Type),
|
| 746 | R('\([^()\0]*\)T', Id.Format_Time),
|
| 747 | R(r'[^\0]', Id.Unknown_Tok), # any other char
|
| 748 | ]
|
| 749 |
|
| 750 | LEXER_DEF[lex_mode_e.VSub_1] = [
|
| 751 | R(VAR_NAME_RE, Id.VSub_Name),
|
| 752 | # ${11} is valid, compared to $11 which is $1 and then literal 1.
|
| 753 | R(r'[0-9]+', Id.VSub_Number),
|
| 754 | C('!', Id.VSub_Bang),
|
| 755 | C('@', Id.VSub_At),
|
| 756 | C('#', Id.VSub_Pound),
|
| 757 | C('$', Id.VSub_Dollar),
|
| 758 | C('*', Id.VSub_Star),
|
| 759 | C('-', Id.VSub_Hyphen),
|
| 760 | C('?', Id.VSub_QMark),
|
| 761 | C('.', Id.VSub_Dot), # ${.myproc builtin sub}
|
| 762 | C('}', Id.Right_DollarBrace),
|
| 763 | C('\\\n', Id.Ignored_LineCont),
|
| 764 | C('\n', Id.Unknown_Tok), # newline not allowed inside ${}
|
| 765 | R(r'[^\0]', Id.Unknown_Tok), # any char except newline
|
| 766 | ]
|
| 767 |
|
| 768 | LEXER_DEF[lex_mode_e.VSub_2] = \
|
| 769 | ID_SPEC.LexerPairs(Kind.VTest) + \
|
| 770 | ID_SPEC.LexerPairs(Kind.VOp0) + \
|
| 771 | ID_SPEC.LexerPairs(Kind.VOpYsh) + \
|
| 772 | ID_SPEC.LexerPairs(Kind.VOp1) + \
|
| 773 | ID_SPEC.LexerPairs(Kind.VOp2) + \
|
| 774 | ID_SPEC.LexerPairs(Kind.VOp3) + [
|
| 775 | C('}', Id.Right_DollarBrace),
|
| 776 |
|
| 777 | C('\\\n', Id.Ignored_LineCont),
|
| 778 | C('\n', Id.Unknown_Tok), # newline not allowed inside ${}
|
| 779 | R(r'[^\0]', Id.Unknown_Tok), # any char except newline
|
| 780 | ]
|
| 781 |
|
| 782 | _EXPR_ARITH_SHARED = [
|
| 783 | C('\\\n', Id.Ignored_LineCont),
|
| 784 | R(r'[^\0]', Id.Unknown_Tok) # any char. This should be a syntax error.
|
| 785 | ]
|
| 786 |
|
| 787 | # https://www.gnu.org/software/bash/manual/html_node/Shell-Arithmetic.html#Shell-Arithmetic
|
| 788 | LEXER_DEF[lex_mode_e.Arith] = \
|
| 789 | _LEFT_SUBS + _VARS + _LEFT_UNQUOTED + [
|
| 790 |
|
| 791 | # Arithmetic expressions can cross newlines.
|
| 792 | R(r'[ \t\r\n]+', Id.Ignored_Space),
|
| 793 |
|
| 794 | # Examples of arith constants:
|
| 795 | # 64#azAZ
|
| 796 | # 0xabc 0xABC
|
| 797 | # 0123
|
| 798 | # A separate digits token makes this easier to parse STATICALLY. But this
|
| 799 | # doesn't help with DYNAMIC parsing.
|
| 800 | R(VAR_NAME_RE, Id.Lit_ArithVarLike), # for variable names or 64#_
|
| 801 | R(r'[0-9]+', Id.Lit_Digits),
|
| 802 | C('@', Id.Lit_At), # for 64#@ or ${a[@]}
|
| 803 | C('#', Id.Lit_Pound), # for 64#a
|
| 804 |
|
| 805 | # TODO: 64#@ interferes with VS_AT. Hm.
|
| 806 | ] + ID_SPEC.LexerPairs(Kind.Arith) + _EXPR_ARITH_SHARED
|
| 807 |
|
| 808 | # A lexer for the parser that converts globs to extended regexes. Since we're
|
| 809 | # only parsing character classes ([^[:space:][:alpha:]]) as opaque blobs, we
|
| 810 | # don't need lexer modes here.
|
| 811 | GLOB_DEF = [
|
| 812 | # These could be operators in the glob, or just literals in a char class,
|
| 813 | # e.g. touch '?'; echo [?].
|
| 814 | C('*', Id.Glob_Star),
|
| 815 | C('?', Id.Glob_QMark),
|
| 816 |
|
| 817 | # For negation. Treated as operators inside [], but literals outside.
|
| 818 | C('!', Id.Glob_Bang),
|
| 819 | C('^', Id.Glob_Caret),
|
| 820 |
|
| 821 | # Character classes.
|
| 822 | C('[', Id.Glob_LBracket),
|
| 823 | C(']', Id.Glob_RBracket),
|
| 824 |
|
| 825 | # There is no whitelist of characters; backslashes are unconditionally
|
| 826 | # removed. With libc.fnmatch(), the pattern r'\f' matches 'f' but not '\\f'.
|
| 827 | # See libc_test.py.
|
| 828 | R(r'\\[^\0]', Id.Glob_EscapedChar),
|
| 829 | C('\\', Id.Glob_BadBackslash), # Trailing single backslash
|
| 830 |
|
| 831 | # For efficiency, combine other characters into a single token, e.g. 'py' in
|
| 832 | # '*.py' or 'alpha' in '[[:alpha:]]'.
|
| 833 | R(r'[a-zA-Z0-9_]+', Id.Glob_CleanLiterals), # no regex escaping
|
| 834 | R(r'[^\0]', Id.Glob_OtherLiteral), # anything else -- examine the char
|
| 835 | ]
|
| 836 |
|
| 837 | # History expansion. We're doing this as "pre-lexing" since that's what bash
|
| 838 | # and zsh seem to do. Example:
|
| 839 | #
|
| 840 | # $ foo=x
|
| 841 | # $ echo $
|
| 842 | # $ !!foo # expands to echo $foo and prints x
|
| 843 | #
|
| 844 | # We can also reuse this in the RootCompleter to expand history interactively.
|
| 845 | #
|
| 846 | # bash note: handled in lib/readline/histexpand.c. Quite messy and handles
|
| 847 | # quotes AGAIN.
|
| 848 | #
|
| 849 | # Note: \! gets expanded to literal \! for the real lexer, but no history
|
| 850 | # expansion occurs.
|
| 851 |
|
| 852 | HISTORY_DEF = [
|
| 853 | # Common operators.
|
| 854 | R(r'![!*^$]', Id.History_Op),
|
| 855 |
|
| 856 | # By command number.
|
| 857 | R(r'!-?[0-9]+', Id.History_Num),
|
| 858 |
|
| 859 | # Search by prefix of substring (optional '?').
|
| 860 | # NOTE: there are no numbers allowed here! Bash doesn't seem to support it.
|
| 861 | # No hyphen since it conflits with $-1 too.
|
| 862 | #
|
| 863 | # Required trailing whitespace is there to avoid conflict with [!charclass]
|
| 864 | # and ${!indirect}. This is a simpler hack than the one bash has. See
|
| 865 | # frontend/lex_test.py.
|
| 866 | R(r'!\??[a-zA-Z_/.][0-9a-zA-Z_/.]+[ \t\r\n]', Id.History_Search),
|
| 867 |
|
| 868 | # Comment is until end of line
|
| 869 | R(r"#[^\0]*", Id.History_Other),
|
| 870 |
|
| 871 | # Single quoted, e.g. 'a' or $'\n'. Terminated by another single quote or
|
| 872 | # end of string.
|
| 873 | R(r"'[^'\0]*'?", Id.History_Other),
|
| 874 |
|
| 875 | # Runs of chars that are definitely not special
|
| 876 | R(r"[^!\\'#\0]+", Id.History_Other),
|
| 877 |
|
| 878 | # Escaped characters. \! disables history
|
| 879 | R(r'\\[^\0]', Id.History_Other),
|
| 880 | # Other single chars, like a trailing \ or !
|
| 881 | R(r'[^\0]', Id.History_Other),
|
| 882 | ]
|
| 883 |
|
| 884 | BRACE_RANGE_DEF = [
|
| 885 | R(r'-?[0-9]+', Id.Range_Int),
|
| 886 | R(r'[a-zA-Z]', Id.Range_Char), # just a single character
|
| 887 | R(r'\.\.', Id.Range_Dots),
|
| 888 | R(r'[^\0]', Id.Range_Other), # invalid
|
| 889 | ]
|
| 890 |
|
| 891 | #
|
| 892 | # YSH lexing
|
| 893 | #
|
| 894 |
|
| 895 | # Valid in lex_mode_e.{Expr,DQ}
|
| 896 | # Used by ysh/grammar_gen.py
|
| 897 | YSH_LEFT_SUBS = [
|
| 898 | C('$(', Id.Left_DollarParen),
|
| 899 | C('${', Id.Left_DollarBrace),
|
| 900 | C('$[', Id.Left_DollarBracket), # TODO: Implement $[x]
|
| 901 | ]
|
| 902 |
|
| 903 | # Valid in lex_mode_e.Expr, but not valid in DQ
|
| 904 | # Used by ysh/grammar_gen.py
|
| 905 |
|
| 906 | YSH_LEFT_UNQUOTED = [
|
| 907 | # Double quoted
|
| 908 | C('"', Id.Left_DoubleQuote),
|
| 909 | C('$"', Id.Left_DollarDoubleQuote), # $"" is synonym for ""
|
| 910 | C('j"', Id.Left_JDoubleQuote), # for printing ERROR
|
| 911 | # Single quoted
|
| 912 | C("'", Id.Left_SingleQuote),
|
| 913 | C("r'", Id.Left_RSingleQuote),
|
| 914 | C("u'", Id.Left_USingleQuote),
|
| 915 | C("b'", Id.Left_BSingleQuote),
|
| 916 | C("$'", Id.Left_DollarSingleQuote), # legacy
|
| 917 | C('^"', Id.Left_CaretDoubleQuote),
|
| 918 | C('"""', Id.Left_TDoubleQuote),
|
| 919 | C('$"""', Id.Left_DollarTDoubleQuote),
|
| 920 | # In expression mode, we add the r'' and c'' prefixes for '' and $''.
|
| 921 | C("'''", Id.Left_TSingleQuote),
|
| 922 | C("r'''", Id.Left_RTSingleQuote),
|
| 923 | C("u'''", Id.Left_UTSingleQuote),
|
| 924 | C("b'''", Id.Left_BTSingleQuote),
|
| 925 | C('@(', Id.Left_AtParen), # Split Command Sub
|
| 926 | C('^(', Id.Left_CaretParen), # Block literals in expression mode
|
| 927 | C('^[', Id.Left_CaretBracket), # Expr literals
|
| 928 | C('^{', Id.Left_CaretBrace), # Unused
|
| 929 | C(':|', Id.Left_ColonPipe), # shell-like word arrays.
|
| 930 | C('%(', Id.Left_PercentParen), # old syntax for shell-like word arrays.
|
| 931 | C('%[', Id.Expr_Reserved), # Maybe: like %() without unquoted [], {}
|
| 932 | C('%{', Id.Expr_Reserved), # Table literals
|
| 933 | # t = %{
|
| 934 | # name:Str age:Int
|
| 935 | # 'andy c' 10
|
| 936 | # }
|
| 937 | # Significant newlines. No unquoted [], {}
|
| 938 |
|
| 939 | # Not sure if we'll use these
|
| 940 | C('@{', Id.Expr_Reserved),
|
| 941 | C('@[', Id.Expr_Reserved),
|
| 942 |
|
| 943 | # Idea: Set literals are #{a, b} like Clojure
|
| 944 | ]
|
| 945 |
|
| 946 | # Used by ysh/grammar_gen.py
|
| 947 | EXPR_OPS = [
|
| 948 | # Terminator
|
| 949 | C(';', Id.Op_Semi),
|
| 950 | C('(', Id.Op_LParen),
|
| 951 | C(')', Id.Op_RParen),
|
| 952 | # NOTE: type expressions are expressions, e.g. Dict[Str, Int]
|
| 953 | C('[', Id.Op_LBracket),
|
| 954 | C(']', Id.Op_RBracket),
|
| 955 | C('{', Id.Op_LBrace),
|
| 956 | C('}', Id.Op_RBrace),
|
| 957 | ]
|
| 958 |
|
| 959 | # Newline is significant, but sometimes elided by expr_parse.py.
|
| 960 | _EXPR_NEWLINE_COMMENT = [
|
| 961 | C('\n', Id.Op_Newline),
|
| 962 | R(r'#[^\n\0]*', Id.Ignored_Comment),
|
| 963 | # Like lex_mode_e.Arith, \r is whitespace even without \n
|
| 964 | R(r'[ \t\r]+', Id.Ignored_Space),
|
| 965 | ]
|
| 966 |
|
| 967 | # Note: we often check match.LooksLikeInteger(s), call mops.FromStr(s), and
|
| 968 | # ASSUME it will not throw ValueError
|
| 969 | LOOKS_LIKE_INTEGER = _WHITESPACE + '-?[0-9]+' + _WHITESPACE
|
| 970 |
|
| 971 | # TODO: use for YSH comparison operators > >= < <=
|
| 972 | #
|
| 973 | # Python allows 0 to be written 00 or 0_0_0, which is weird. But let's be
|
| 974 | # consistent, and avoid '00' turning into a float!
|
| 975 | _YSH_DECIMAL_INT_RE = r'[0-9](_?[0-9])*'
|
| 976 |
|
| 977 | LOOKS_LIKE_YSH_INT = _WHITESPACE + '-?' + _YSH_DECIMAL_INT_RE + _WHITESPACE
|
| 978 |
|
| 979 | _YSH_FLOAT_RE = (
|
| 980 | _YSH_DECIMAL_INT_RE +
|
| 981 | # Unlike Python, exponent can't be like 42e5_000. There's no use because
|
| 982 | # 1e309 is already inf. Let's keep our code simple.
|
| 983 | r'(\.' + _YSH_DECIMAL_INT_RE + ')?([eE][+\-]?[0-9]+)?')
|
| 984 |
|
| 985 | # Ditto, used for YSH comparison operators
|
| 986 | # Added optional Optional -?
|
| 987 | # Example: -3_000_000.000_001e12
|
| 988 | LOOKS_LIKE_YSH_FLOAT = _WHITESPACE + '-?' + _YSH_FLOAT_RE + _WHITESPACE
|
| 989 |
|
| 990 | # Python 3 float literals:
|
| 991 |
|
| 992 | # digitpart ::= digit (["_"] digit)*
|
| 993 | # fraction ::= "." digitpart
|
| 994 | # exponent ::= ("e" | "E") ["+" | "-"] digitpart
|
| 995 | # pointfloat ::= [digitpart] fraction | digitpart "."
|
| 996 | # exponentfloat ::= (digitpart | pointfloat) exponent
|
| 997 | # floatnumber ::= pointfloat | exponentfloat
|
| 998 |
|
| 999 | # NOTE: Borrowing tokens from Arith (i.e. $(( )) ), but not using LexerPairs().
|
| 1000 | LEXER_DEF[lex_mode_e.Expr] = \
|
| 1001 | _VARS + YSH_LEFT_SUBS + YSH_LEFT_UNQUOTED + EXPR_OPS + EXPR_WORDS + \
|
| 1002 | EXPR_CHARS + [
|
| 1003 |
|
| 1004 | # https://docs.python.org/3/reference/lexical_analysis.html#integer-literals
|
| 1005 | #
|
| 1006 | # integer ::= decinteger | bininteger | octinteger | hexinteger
|
| 1007 | # decinteger ::= nonzerodigit (["_"] digit)* | "0"+ (["_"] "0")*
|
| 1008 | # bininteger ::= "0" ("b" | "B") (["_"] bindigit)+
|
| 1009 | # octinteger ::= "0" ("o" | "O") (["_"] octdigit)+
|
| 1010 | # hexinteger ::= "0" ("x" | "X") (["_"] hexdigit)+
|
| 1011 | # nonzerodigit ::= "1"..."9"
|
| 1012 | # digit ::= "0"..."9"
|
| 1013 | # bindigit ::= "0" | "1"
|
| 1014 | # octdigit ::= "0"..."7"
|
| 1015 | # hexdigit ::= digit | "a"..."f" | "A"..."F"
|
| 1016 |
|
| 1017 | R(_YSH_DECIMAL_INT_RE, Id.Expr_DecInt),
|
| 1018 |
|
| 1019 | R(r'0[bB](_?[01])+', Id.Expr_BinInt),
|
| 1020 | R(r'0[oO](_?[0-7])+', Id.Expr_OctInt),
|
| 1021 | R(r'0[xX](_?[0-9a-fA-F])+', Id.Expr_HexInt),
|
| 1022 |
|
| 1023 | R(_YSH_FLOAT_RE, Id.Expr_Float),
|
| 1024 |
|
| 1025 | # These can be looked up as keywords separately, so you enforce that they have
|
| 1026 | # space around them?
|
| 1027 | R(VAR_NAME_RE, Id.Expr_Name),
|
| 1028 |
|
| 1029 | R('%' + VAR_NAME_RE, Id.Expr_Symbol),
|
| 1030 |
|
| 1031 | #
|
| 1032 | # Arith
|
| 1033 | #
|
| 1034 |
|
| 1035 | C(',', Id.Arith_Comma),
|
| 1036 | C(':', Id.Arith_Colon), # for slicing a[1:2], and mylist:pop()
|
| 1037 |
|
| 1038 | C('?', Id.Arith_QMark), # regex postfix
|
| 1039 |
|
| 1040 | C('+', Id.Arith_Plus), # arith infix, regex postfix
|
| 1041 | C('-', Id.Arith_Minus), # arith infix, regex postfix
|
| 1042 | C('*', Id.Arith_Star),
|
| 1043 | C('^', Id.Arith_Caret), # xor
|
| 1044 | C('/', Id.Arith_Slash),
|
| 1045 | C('%', Id.Arith_Percent),
|
| 1046 |
|
| 1047 | C('**', Id.Arith_DStar), # exponentiation
|
| 1048 | C('++', Id.Arith_DPlus), # Option for string/list concatenation
|
| 1049 |
|
| 1050 | C('<', Id.Arith_Less),
|
| 1051 | C('>', Id.Arith_Great),
|
| 1052 | C('<=', Id.Arith_LessEqual),
|
| 1053 | C('>=', Id.Arith_GreatEqual),
|
| 1054 | C('===', Id.Expr_TEqual),
|
| 1055 | C('!==', Id.Expr_NotDEqual),
|
| 1056 |
|
| 1057 | C('==', Id.Unknown_DEqual), # user must choose === or ~==
|
| 1058 |
|
| 1059 | # Bitwise operators
|
| 1060 | C('&', Id.Arith_Amp),
|
| 1061 | C('|', Id.Arith_Pipe),
|
| 1062 | C('>>', Id.Arith_DGreat),
|
| 1063 | C('<<', Id.Arith_DLess), # Doesn't Java also have <<< ?
|
| 1064 |
|
| 1065 | # Bitwise complement, as well as infix pattern matching
|
| 1066 | C('~', Id.Arith_Tilde),
|
| 1067 | C('!~', Id.Expr_NotTilde),
|
| 1068 | C('~~', Id.Expr_DTilde),
|
| 1069 | C('!~~', Id.Expr_NotDTilde),
|
| 1070 |
|
| 1071 | # Left out for now:
|
| 1072 | # ++ -- -- needed for loops, awk?
|
| 1073 | # ! && || -- needed for find dialect
|
| 1074 | # = += etc.
|
| 1075 |
|
| 1076 | C('=', Id.Arith_Equal),
|
| 1077 |
|
| 1078 | C('+=', Id.Arith_PlusEqual),
|
| 1079 | C('-=', Id.Arith_MinusEqual),
|
| 1080 | C('*=', Id.Arith_StarEqual),
|
| 1081 | C('/=', Id.Arith_SlashEqual),
|
| 1082 | C('%=', Id.Arith_PercentEqual),
|
| 1083 |
|
| 1084 | C('>>=', Id.Arith_DGreatEqual),
|
| 1085 | C('<<=', Id.Arith_DLessEqual),
|
| 1086 | C('&=', Id.Arith_AmpEqual),
|
| 1087 | C('|=', Id.Arith_PipeEqual),
|
| 1088 | C('^=', Id.Arith_CaretEqual), # Exponentiation
|
| 1089 |
|
| 1090 | # Augmented assignment that YSH has, but sh and OSH don't have
|
| 1091 | C('**=', Id.Expr_DStarEqual),
|
| 1092 | C('//=', Id.Expr_DSlashEqual),
|
| 1093 |
|
| 1094 | #
|
| 1095 | # Expr
|
| 1096 | #
|
| 1097 |
|
| 1098 | C('!', Id.Expr_Bang), # For eggex negation
|
| 1099 |
|
| 1100 | C('//', Id.Expr_DSlash), # For YSH integer division
|
| 1101 | C('~==', Id.Expr_TildeDEqual), # approximate equality
|
| 1102 |
|
| 1103 | C('.', Id.Expr_Dot), # d.key is alias for d['key']
|
| 1104 | C('..', Id.Expr_DDot), # range 1..5
|
| 1105 | C('->', Id.Expr_RArrow), # s->startswith()
|
| 1106 | C('$', Id.Expr_Dollar), # legacy regex end: /d+ $/ (better written /d+ >/
|
| 1107 |
|
| 1108 | # Reserved this. Go uses it for channels, etc.
|
| 1109 | # I guess it conflicts with -4<-3, but that's OK -- spaces suffices.
|
| 1110 | C('<-', Id.Expr_Reserved),
|
| 1111 | C('=>', Id.Expr_RDArrow), # for df => filter(age > 10)
|
| 1112 | # and match (x) { 1 => "one" }
|
| 1113 | # note: other languages use |>
|
| 1114 | # R/dplyr uses %>%
|
| 1115 |
|
| 1116 | C('...', Id.Expr_Ellipsis), # f(...args) and maybe a[:, ...]
|
| 1117 |
|
| 1118 | # For multiline regex literals?
|
| 1119 | C('///', Id.Expr_Reserved),
|
| 1120 |
|
| 1121 | # Splat operators
|
| 1122 | C('@', Id.Expr_At),
|
| 1123 | # NOTE: Unused
|
| 1124 | C('@@', Id.Expr_DoubleAt),
|
| 1125 | ] + _EXPR_NEWLINE_COMMENT + _EXPR_ARITH_SHARED
|
| 1126 |
|
| 1127 | LEXER_DEF[lex_mode_e.FuncParens] = [
|
| 1128 | # () with spaces
|
| 1129 | R(r'[ \t]*\([ \t]*\)', Id.LookAhead_FuncParens),
|
| 1130 | # anything else
|
| 1131 | R(r'[^\0]', Id.Unknown_Tok)
|
| 1132 | ]
|