| 1 | ---
|
| 2 | title: YSH Expression Language (Oils Reference)
|
| 3 | all_docs_url: ..
|
| 4 | body_css_class: width40
|
| 5 | default_highlighter: oils-sh
|
| 6 | preserve_anchor_case: yes
|
| 7 | ---
|
| 8 |
|
| 9 | <div class="doc-ref-header">
|
| 10 |
|
| 11 | [Oils Reference](index.html) —
|
| 12 | Chapter **YSH Expression Language**
|
| 13 |
|
| 14 | </div>
|
| 15 |
|
| 16 | This chapter describes the YSH expression language, which includes [Egg
|
| 17 | Expressions]($xref:eggex).
|
| 18 |
|
| 19 | <div id="dense-toc">
|
| 20 | </div>
|
| 21 |
|
| 22 | ## Assignment
|
| 23 |
|
| 24 | ### assign
|
| 25 |
|
| 26 | The `=` operator is used with assignment keywords:
|
| 27 |
|
| 28 | var x = 42
|
| 29 | setvar x = 43
|
| 30 |
|
| 31 | const y = 'k'
|
| 32 |
|
| 33 | setglobal z = 'g'
|
| 34 |
|
| 35 | ### aug-assign
|
| 36 |
|
| 37 | The augmented assignment operators are:
|
| 38 |
|
| 39 | += -= *= /= **= //= %=
|
| 40 | &= |= ^= <<= >>=
|
| 41 |
|
| 42 | They are used with `setvar` and `setglobal`. For example:
|
| 43 |
|
| 44 | setvar x += 2
|
| 45 |
|
| 46 | is the same as:
|
| 47 |
|
| 48 | setvar x = x + 2
|
| 49 |
|
| 50 | Likewise, these are the same:
|
| 51 |
|
| 52 | setglobal a[i] -= 1
|
| 53 |
|
| 54 | setglobal a[i] = a[i] - 1
|
| 55 |
|
| 56 | ## Literals
|
| 57 |
|
| 58 | ### atom-literal
|
| 59 |
|
| 60 | YSH uses JavaScript-like spellings for these three "atoms":
|
| 61 |
|
| 62 | null # type Null
|
| 63 | true false # type Bool
|
| 64 |
|
| 65 | Note: to signify "no value", you may sometimes use an empty string `''`,
|
| 66 | instead of `null`.
|
| 67 |
|
| 68 | ### int-literal
|
| 69 |
|
| 70 | Examples of integer literals:
|
| 71 |
|
| 72 | var decimal = 42
|
| 73 | var big = 42_000
|
| 74 |
|
| 75 | var hex = 0x0010_ffff
|
| 76 |
|
| 77 | var octal = 0o755
|
| 78 |
|
| 79 | var binary = 0b0001_0000
|
| 80 |
|
| 81 | ### float-lit
|
| 82 |
|
| 83 | Examples of float literals:
|
| 84 |
|
| 85 | var myfloat = 3.14
|
| 86 |
|
| 87 | var f2 = -1.5e-100
|
| 88 |
|
| 89 | ### char-literal
|
| 90 |
|
| 91 | Three kinds of unquoted backslash escapes are allowed in expression mode. They
|
| 92 | match what's available in quoted J8-style strings:
|
| 93 |
|
| 94 | var backslash = \\
|
| 95 | var quotes = \' ++ \" # same as u'\'' ++ '"'
|
| 96 |
|
| 97 | var mu = \u{3bc} # same as u'\u{3bc}'
|
| 98 |
|
| 99 | var nul = \y00 # same as b'\y00'
|
| 100 |
|
| 101 | ### ysh-string
|
| 102 |
|
| 103 | YSH has single and double-quoted strings borrowed from Bourne shell, and
|
| 104 | C-style strings borrowed from J8 Notation.
|
| 105 |
|
| 106 | Double quoted strings respect `$` interpolation:
|
| 107 |
|
| 108 | var dq = "hello $world and $(hostname)"
|
| 109 |
|
| 110 | You can add a `$` before the left quote to be explicit: `$"x is $x"` rather
|
| 111 | than `"x is $x"`.
|
| 112 |
|
| 113 | Single quoted strings may be raw:
|
| 114 |
|
| 115 | var s = r'line\n' # raw string means \n is literal, NOT a newline
|
| 116 |
|
| 117 | Or *J8 strings* with backslash escapes:
|
| 118 |
|
| 119 | var s = u'line\n \u{3bc}' # unicode string means \n is a newline
|
| 120 | var s = b'line\n \u{3bc} \yff' # same thing, but also allows bytes
|
| 121 |
|
| 122 | Both `u''` and `b''` strings evaluate to the single `Str` type. The difference
|
| 123 | is that `b''` strings allow the `\yff` byte escape.
|
| 124 |
|
| 125 | #### Notes
|
| 126 |
|
| 127 | There's no way to express a single quote in raw strings. Use one of the other
|
| 128 | forms instead:
|
| 129 |
|
| 130 | var sq = "single quote: ' "
|
| 131 | var sq = u'single quote: \' '
|
| 132 |
|
| 133 | Sometimes you can omit the `r`, e.g. where there are no backslashes and thus no
|
| 134 | ambiguity:
|
| 135 |
|
| 136 | echo 'foo'
|
| 137 | echo r'foo' # same thing
|
| 138 |
|
| 139 | The `u''` and `b''` strings are called *J8 strings* because the syntax in YSH
|
| 140 | **code** matches JSON-like **data**.
|
| 141 |
|
| 142 | var strU = u'mu = \u{3bc}' # J8 string with escapes
|
| 143 | var strB = b'bytes \yff' # J8 string that can express byte strings
|
| 144 |
|
| 145 | More examples:
|
| 146 |
|
| 147 | var myRaw = r'[a-z]\n' # raw strings can be used for regexes (not
|
| 148 | # eggexes)
|
| 149 |
|
| 150 | ### triple-quoted
|
| 151 |
|
| 152 | Triple-quoted string literals have leading whitespace stripped on each line.
|
| 153 | They come in the same variants:
|
| 154 |
|
| 155 | var dq = """
|
| 156 | hello $world and $(hostname)
|
| 157 | no leading whitespace
|
| 158 | """
|
| 159 |
|
| 160 | var myRaw = r'''
|
| 161 | raw string
|
| 162 | no leading whitespace
|
| 163 | '''
|
| 164 |
|
| 165 | var strU = u'''
|
| 166 | string that happens to be unicode \u{3bc}
|
| 167 | no leading whitespace
|
| 168 | '''
|
| 169 |
|
| 170 | var strB = b'''
|
| 171 | string that happens to be bytes \u{3bc} \yff
|
| 172 | no leading whitespace
|
| 173 | '''
|
| 174 |
|
| 175 | Again, you can omit the `r` prefix if there's no backslash, because it's not
|
| 176 | ambiguous:
|
| 177 |
|
| 178 | var myRaw = '''
|
| 179 | raw string
|
| 180 | no leading whitespace
|
| 181 | '''
|
| 182 |
|
| 183 | ### str-template
|
| 184 |
|
| 185 | String templates use the same syntax as double-quoted strings:
|
| 186 |
|
| 187 | var mytemplate = ^"name = $name, age = $age"
|
| 188 |
|
| 189 | Related topics:
|
| 190 |
|
| 191 | - [Str => replace](chap-type-method.html#replace)
|
| 192 | - [ysh-string](chap-expr-lang.html#ysh-string)
|
| 193 |
|
| 194 | ### list-literal
|
| 195 |
|
| 196 | Lists have a Python-like syntax:
|
| 197 |
|
| 198 | var mylist = ['one', 'two', [42, 43]]
|
| 199 |
|
| 200 | And a shell-like syntax:
|
| 201 |
|
| 202 | var list2 = :| one two |
|
| 203 |
|
| 204 | The shell-like syntax accepts the same syntax as a simple command:
|
| 205 |
|
| 206 | ls $mystr @ARGV *.py {foo,bar}@example.com
|
| 207 |
|
| 208 | # Rather than executing ls, evaluate words into a List
|
| 209 | var cmd = :| ls $mystr @ARGV *.py {foo,bar}@example.com |
|
| 210 |
|
| 211 | ### dict-literal
|
| 212 |
|
| 213 | Dicts look like JavaScript.
|
| 214 |
|
| 215 | var d = {
|
| 216 | key1: 'value', # key can be unquoted if it looks like a var name
|
| 217 | 'key2': 42, # or quote it
|
| 218 |
|
| 219 | ['key2' ++ suffix]: 43, # bracketed expression
|
| 220 | }
|
| 221 |
|
| 222 | Omitting a value means that the corresponding key takes the value of a var of
|
| 223 | the same name:
|
| 224 |
|
| 225 | ysh$ var x = 42
|
| 226 | ysh$ var y = 43
|
| 227 |
|
| 228 | ysh$ var d = {x, y} # values omitted
|
| 229 | ysh$ = d
|
| 230 | (Dict) {x: 42, y: 43}
|
| 231 |
|
| 232 | ### range
|
| 233 |
|
| 234 | A range is a sequence of numbers that can be iterated over:
|
| 235 |
|
| 236 | for i in (0 ..< 3) {
|
| 237 | echo $i
|
| 238 | }
|
| 239 | => 0
|
| 240 | => 1
|
| 241 | => 2
|
| 242 |
|
| 243 | The `..<` syntax is for half-open ranges. `..=` is for closed ranges:
|
| 244 |
|
| 245 | for i in (0 ..= 3) {
|
| 246 | echo $i
|
| 247 | }
|
| 248 | => 0
|
| 249 | => 1
|
| 250 | => 2
|
| 251 | => 3
|
| 252 |
|
| 253 | ### block-expr
|
| 254 |
|
| 255 | In YSH expressions, we use `^()` to create a [Command][] object:
|
| 256 |
|
| 257 | var myblock = ^(echo $PWD; ls *.txt)
|
| 258 |
|
| 259 | It's more common for [Command][] objects to be created with block arguments,
|
| 260 | which are not expressions:
|
| 261 |
|
| 262 | cd /tmp {
|
| 263 | echo $PWD
|
| 264 | ls *.txt
|
| 265 | }
|
| 266 |
|
| 267 | [Command]: chap-type-method.html#Command
|
| 268 |
|
| 269 | ### expr-literal
|
| 270 |
|
| 271 | An expression literal is an object that holds an unevaluated expression:
|
| 272 |
|
| 273 | var myexpr = ^[1 + 2*3]
|
| 274 |
|
| 275 | [Expr]: chap-type-method.html#Expr
|
| 276 |
|
| 277 | ## Operators
|
| 278 |
|
| 279 | ### op-precedence
|
| 280 |
|
| 281 | YSH operator precedence is identical to Python's operator precedence.
|
| 282 |
|
| 283 | New operators:
|
| 284 |
|
| 285 | - `++` has the same precedence as `+`
|
| 286 | - `->` and `=>` have the same precedence as `.`
|
| 287 |
|
| 288 | <!-- TODO: show grammar -->
|
| 289 |
|
| 290 |
|
| 291 | <h3 id="concat">concat <code>++</code></h3>
|
| 292 |
|
| 293 | The concatenation operator works on `Str` objects:
|
| 294 |
|
| 295 | ysh$ var s = 'hello'
|
| 296 | ysh$ var t = s ++ ' world'
|
| 297 |
|
| 298 | ysh$ = t
|
| 299 | (Str) "hello world"
|
| 300 |
|
| 301 | and `List` objects:
|
| 302 |
|
| 303 | ysh$ var L = ['one', 'two']
|
| 304 | ysh$ var M = L ++ ['three', '4']
|
| 305 |
|
| 306 | ysh$ = M
|
| 307 | (List) ["one", "two", "three", "4"]
|
| 308 |
|
| 309 | String interpolation can be nicer than `++`:
|
| 310 |
|
| 311 | var t2 = "${s} world" # same as t
|
| 312 |
|
| 313 | Likewise, splicing lists can be nicer:
|
| 314 |
|
| 315 | var M2 = :| @L three 4 | # same as M
|
| 316 |
|
| 317 | ### ysh-equals
|
| 318 |
|
| 319 | YSH has strict equality:
|
| 320 |
|
| 321 | a === b # Python-like, without type conversion
|
| 322 | a !== b # negated
|
| 323 |
|
| 324 | And type converting equality:
|
| 325 |
|
| 326 | '3' ~== 3 # True, type conversion
|
| 327 |
|
| 328 | The `~==` operator expects a string as the left operand.
|
| 329 |
|
| 330 | ---
|
| 331 |
|
| 332 | Note that:
|
| 333 |
|
| 334 | - `3 === 3.0` is false because integers and floats are different types, and
|
| 335 | there is no type conversion.
|
| 336 | - `3 ~== 3.0` is an error, because the left operand isn't a string.
|
| 337 |
|
| 338 | You may want to use explicit `int()` and `float()` to convert numbers, and then
|
| 339 | compare them.
|
| 340 |
|
| 341 | ---
|
| 342 |
|
| 343 | Compare objects for identity with `is`:
|
| 344 |
|
| 345 | ysh$ var d = {}
|
| 346 | ysh$ var e = d
|
| 347 |
|
| 348 | ysh$ = d is d
|
| 349 | (Bool) true
|
| 350 |
|
| 351 | ysh$ = d is {other: 'dict'}
|
| 352 | (Bool) false
|
| 353 |
|
| 354 | To negate `is`, use `is not` (like Python:
|
| 355 |
|
| 356 | ysh$ d is not {other: 'dict'}
|
| 357 | (Bool) true
|
| 358 |
|
| 359 | ### ysh-in
|
| 360 |
|
| 361 | The `in` operator tests if a key is in a dictionary:
|
| 362 |
|
| 363 | var d = {k: 42}
|
| 364 | if ('k' in d) {
|
| 365 | echo yes
|
| 366 | } # => yes
|
| 367 |
|
| 368 | Unlike Python, `in` doesn't work on `Str` and `List` instances. This because
|
| 369 | those operations take linear time rather than constant time (O(n) rather than
|
| 370 | O(1)).
|
| 371 |
|
| 372 | TODO: Use `includes() / contains()` methods instead.
|
| 373 |
|
| 374 | ### ysh-compare
|
| 375 |
|
| 376 | The comparison operators apply to integers or floats:
|
| 377 |
|
| 378 | 4 < 4 # => false
|
| 379 | 4 <= 4 # => true
|
| 380 |
|
| 381 | 5.0 > 5.0 # => false
|
| 382 | 5.0 >= 5.0 # => true
|
| 383 |
|
| 384 | Example in context:
|
| 385 |
|
| 386 | if (x < 0) {
|
| 387 | echo 'x is negative'
|
| 388 | }
|
| 389 |
|
| 390 | ### ysh-logical
|
| 391 |
|
| 392 | The logical operators take boolean operands, and are spelled like Python:
|
| 393 |
|
| 394 | not
|
| 395 | and or
|
| 396 |
|
| 397 | Note that they are distinct from `! && ||`, which are part of the [command
|
| 398 | language](chap-cmd-lang.html).
|
| 399 |
|
| 400 | ### ysh-arith
|
| 401 |
|
| 402 | YSH supports most of the arithmetic operators from Python. Notably, `/` and `%`
|
| 403 | differ from Python as [they round toward zero, not negative
|
| 404 | infinity](https://www.oilshell.org/blog/2024/03/release-0.21.0.html#integers-dont-do-whatever-python-or-c-does).
|
| 405 |
|
| 406 | Use `+ - *` for `Int` or `Float` addition, subtraction and multiplication. If
|
| 407 | any of the operands are `Float`s, then the output will also be a `Float`.
|
| 408 |
|
| 409 | Use `/` and `//` for `Float` division and `Int` division, respectively. `/`
|
| 410 | will _always_ result in a `Float`, meanwhile `//` will _always_ result in an
|
| 411 | `Int`.
|
| 412 |
|
| 413 | = 1 / 2 # => (Float) 0.5
|
| 414 | = 1 // 2 # => (Int) 0
|
| 415 |
|
| 416 | Use `%` to compute the _remainder_ of integer division. The left operand must
|
| 417 | be an `Int` and the right a _positive_ `Int`.
|
| 418 |
|
| 419 | = 1 % 2 # -> (Int) 1
|
| 420 | = -4 % 2 # -> (Int) 0
|
| 421 |
|
| 422 | Use `**` for exponentiation. The left operand must be an `Int` and the right a
|
| 423 | _positive_ `Int`.
|
| 424 |
|
| 425 | All arithmetic operators may coerce either of their operands from strings to a
|
| 426 | number, provided those strings are formatted as numbers.
|
| 427 |
|
| 428 | = 10 + '1' # => (Int) 11
|
| 429 |
|
| 430 | Operators like `+ - * /` will coerce strings to _either_ an `Int` or `Float`.
|
| 431 | However, operators like `// ** %` and bit shifts will coerce strings _only_ to
|
| 432 | an `Int`.
|
| 433 |
|
| 434 | = '1.14' + '2' # => (Float) 3.14
|
| 435 | = '1.14' % '2' # Type Error: Left operand is a Str
|
| 436 |
|
| 437 | ### ysh-bitwise
|
| 438 |
|
| 439 | Bitwise operators are like Python and C:
|
| 440 |
|
| 441 | ~ # unary complement
|
| 442 |
|
| 443 | & | ^ # binary and, or, xor
|
| 444 |
|
| 445 | >> << # bit shift
|
| 446 |
|
| 447 | ### ysh-ternary
|
| 448 |
|
| 449 | The ternary operator is borrowed from Python:
|
| 450 |
|
| 451 | display = 'yes' if len(s) else 'empty'
|
| 452 |
|
| 453 | ### ysh-index
|
| 454 |
|
| 455 | `Str` objects can be indexed by byte:
|
| 456 |
|
| 457 | ysh$ var s = 'cat'
|
| 458 | ysh$ = mystr[1]
|
| 459 | (Str) 'a'
|
| 460 |
|
| 461 | ysh$ = mystr[-1] # index from the end
|
| 462 | (Str) 't'
|
| 463 |
|
| 464 | `List` objects:
|
| 465 |
|
| 466 | ysh$ var mylist = [1, 2, 3]
|
| 467 | ysh$ = mylist[2]
|
| 468 | (Int) 3
|
| 469 |
|
| 470 | `Dict` objects are indexed by string key:
|
| 471 |
|
| 472 | ysh$ var mydict = {'key': 42}
|
| 473 | ysh$ = mydict['key']
|
| 474 | (Int) 42
|
| 475 |
|
| 476 | ### ysh-attr
|
| 477 |
|
| 478 | The `.` operator looks up values on either `Dict` or `Obj` instances.
|
| 479 |
|
| 480 | On dicts, it looks for the value associated with a key. That is, the
|
| 481 | expression `mydict.key` is short for `mydict['key']` (like JavaScript, but
|
| 482 | unlike Python.)
|
| 483 |
|
| 484 | ---
|
| 485 |
|
| 486 | On objects, the expression `obj.x` looks for attributes, with a special rule
|
| 487 | for bound methods. The rules are:
|
| 488 |
|
| 489 | 1. Search the properties of `obj` for a field named `x`.
|
| 490 | - If it exists, return the value literally. (It can be of any type: `Func`, `Int`,
|
| 491 | `Str`, ...)
|
| 492 | 2. Search up the prototype chain for a field named `x`.
|
| 493 | - If it exists, and is **not** a `Func`, return the value literally.
|
| 494 | - If it **is** a `Func`, return **bound method**, which is an (object,
|
| 495 | function) pair.
|
| 496 |
|
| 497 | Later, when the bound method is called, the object is passed as the first
|
| 498 | argument to the function (`self`), making it a method call. This is how a
|
| 499 | method has access to the object's properties.
|
| 500 |
|
| 501 | Example of first rule:
|
| 502 |
|
| 503 | func Free(i) {
|
| 504 | return (i + 1)
|
| 505 | }
|
| 506 | var module = Object(null, {Free})
|
| 507 | echo $[module.Free(42)] # => 43
|
| 508 |
|
| 509 | Example of second rule:
|
| 510 |
|
| 511 | func method(self, i) {
|
| 512 | return (self.n + i)
|
| 513 | }
|
| 514 | var methods = Object(null, {method})
|
| 515 | var obj = Object(methods, {n: 10})
|
| 516 | echo $[obj.method(42)] # => 52
|
| 517 |
|
| 518 | ### ysh-slice
|
| 519 |
|
| 520 | Slicing gives you a subsequence of a `Str` or `List`, as in Python.
|
| 521 |
|
| 522 | Negative indices are relative to the end.
|
| 523 |
|
| 524 | String example:
|
| 525 |
|
| 526 | $ var s = 'spam eggs'
|
| 527 | $ pp (s[1:-1])
|
| 528 | (Str) "pam egg"
|
| 529 |
|
| 530 | $ echo "x $[s[2:]]"
|
| 531 | x am eggs
|
| 532 |
|
| 533 | List example:
|
| 534 |
|
| 535 | $ var foods = ['ale', 'bean', 'corn']
|
| 536 | $ pp (foods[-2:])
|
| 537 | (List) ["bean","corn"]
|
| 538 |
|
| 539 | $ write -- @[foods[:2]]
|
| 540 | ale
|
| 541 | bean
|
| 542 |
|
| 543 | ### func-call
|
| 544 |
|
| 545 | A function call expression looks like Python:
|
| 546 |
|
| 547 | ysh$ = f('s', 't', named=42)
|
| 548 |
|
| 549 | A semicolon `;` can be used after positional args and before named args, but
|
| 550 | isn't always required:
|
| 551 |
|
| 552 | ysh$ = f('s', 't'; named=42)
|
| 553 |
|
| 554 | In these cases, the `;` is necessary:
|
| 555 |
|
| 556 | ysh$ = f(...args; ...kwargs)
|
| 557 |
|
| 558 | ysh$ = f(42, 43; ...kwargs)
|
| 559 |
|
| 560 | ### thin-arrow
|
| 561 |
|
| 562 | The thin arrow is for mutating methods:
|
| 563 |
|
| 564 | var mylist = ['bar']
|
| 565 | call mylist->pop()
|
| 566 |
|
| 567 | var mydict = {name: 'foo'}
|
| 568 | call mydict->erase('name')
|
| 569 |
|
| 570 | On `Obj` instances, `obj->mymethod` looks up the prototype chain for a function
|
| 571 | named `M/mymethod`. The `M/` prefix signals mutation.
|
| 572 |
|
| 573 | Example:
|
| 574 |
|
| 575 | func inc(self, n) {
|
| 576 | setvar self.i += n
|
| 577 | }
|
| 578 | var Counter_methods = Object(null, {'M/inc': inc})
|
| 579 | var c = Object(Counter_methods, {i: 0})
|
| 580 |
|
| 581 | call c->inc(5)
|
| 582 | echo $[c.i] # => 5
|
| 583 |
|
| 584 | It does **not** look in the properties of an object.
|
| 585 |
|
| 586 | ### fat-arrow
|
| 587 |
|
| 588 | The fat arrow is for transforming methods:
|
| 589 |
|
| 590 | if (s => startsWith('prefix')) {
|
| 591 | echo 'yes'
|
| 592 | }
|
| 593 |
|
| 594 | If the method lookup on `s` fails, it looks for free functions. This means it
|
| 595 | can be used for "chaining" transformations:
|
| 596 |
|
| 597 | var x = myFunc() => list() => join()
|
| 598 |
|
| 599 | ### match-ops
|
| 600 |
|
| 601 | YSH has four pattern matching operators: `~ !~ ~~ !~~`.
|
| 602 |
|
| 603 | Does string match an **eggex**?
|
| 604 |
|
| 605 | var filename = 'x42.py'
|
| 606 | if (filename ~ / d+ /) {
|
| 607 | echo 'number'
|
| 608 | }
|
| 609 |
|
| 610 | Does a string match a POSIX regular expression (ERE syntax)?
|
| 611 |
|
| 612 | if (filename ~ '[[:digit:]]+') {
|
| 613 | echo 'number'
|
| 614 | }
|
| 615 |
|
| 616 | Negate the result with the `!~` operator:
|
| 617 |
|
| 618 | if (filename !~ /space/ ) {
|
| 619 | echo 'no space'
|
| 620 | }
|
| 621 |
|
| 622 | if (filename !~ '[[:space:]]' ) {
|
| 623 | echo 'no space'
|
| 624 | }
|
| 625 |
|
| 626 | Does a string match a **glob**?
|
| 627 |
|
| 628 | if (filename ~~ '*.py') {
|
| 629 | echo 'Python'
|
| 630 | }
|
| 631 |
|
| 632 | if (filename !~~ '*.py') {
|
| 633 | echo 'not Python'
|
| 634 | }
|
| 635 |
|
| 636 | Take care not to confuse glob patterns and regular expressions.
|
| 637 |
|
| 638 | - Related doc: [YSH Regex API](../ysh-regex-api.html)
|
| 639 |
|
| 640 | ## Eggex
|
| 641 |
|
| 642 | ### re-literal
|
| 643 |
|
| 644 | An eggex literal looks like this:
|
| 645 |
|
| 646 | / expression ; flags ; translation preference /
|
| 647 |
|
| 648 | The flags and translation preference are both optional.
|
| 649 |
|
| 650 | Examples:
|
| 651 |
|
| 652 | var pat = / d+ / # => [[:digit:]]+
|
| 653 |
|
| 654 | You can specify flags passed to libc `regcomp()`:
|
| 655 |
|
| 656 | var pat = / d+ ; reg_icase reg_newline /
|
| 657 |
|
| 658 | You can specify a translation preference after a second semi-colon:
|
| 659 |
|
| 660 | var pat = / d+ ; ; ERE /
|
| 661 |
|
| 662 | Right now the translation preference does nothing. It could be used to
|
| 663 | translate eggex to PCRE or Python syntax.
|
| 664 |
|
| 665 | - Related doc: [Egg Expressions](../eggex.html)
|
| 666 |
|
| 667 | ### re-primitive
|
| 668 |
|
| 669 | There are two kinds of eggex primitives.
|
| 670 |
|
| 671 | "Zero-width assertions" match a position rather than a character:
|
| 672 |
|
| 673 | %start # translates to ^
|
| 674 | %end # translates to $
|
| 675 |
|
| 676 | Literal characters appear within **single** quotes:
|
| 677 |
|
| 678 | 'oh *really*' # translates to regex-escaped string
|
| 679 |
|
| 680 | Double-quoted strings are **not** eggex primitives. Instead, you can use
|
| 681 | splicing of strings:
|
| 682 |
|
| 683 | var dq = "hi $name"
|
| 684 | var eggex = / @dq /
|
| 685 |
|
| 686 | ### class-literal
|
| 687 |
|
| 688 | An eggex character class literal specifies a set. It can have individual
|
| 689 | characters and ranges:
|
| 690 |
|
| 691 | [ 'x' 'y' 'z' a-f A-F 0-9 ] # 3 chars, 3 ranges
|
| 692 |
|
| 693 | Omit quotes on ASCII characters:
|
| 694 |
|
| 695 | [ x y z ] # avoid typing 'x' 'y' 'z'
|
| 696 |
|
| 697 | Sets of characters can be written as strings
|
| 698 |
|
| 699 | [ 'xyz' ] # any of 3 chars, not a sequence of 3 chars
|
| 700 |
|
| 701 | Backslash escapes are respected:
|
| 702 |
|
| 703 | [ \\ \' \" \0 ]
|
| 704 | [ \xFF \u{3bc} ]
|
| 705 |
|
| 706 | (Note that we don't use `\yFF`, as in J8 strings.)
|
| 707 |
|
| 708 | Splicing:
|
| 709 |
|
| 710 | [ @str_var ]
|
| 711 |
|
| 712 | Negation always uses `!`
|
| 713 |
|
| 714 | ![ a-f A-F 'xyz' @str_var ]
|
| 715 |
|
| 716 | ### named-class
|
| 717 |
|
| 718 | Perl-like shortcuts for sets of characters:
|
| 719 |
|
| 720 | [ dot ] # => .
|
| 721 | [ digit ] # => [[:digit:]]
|
| 722 | [ space ] # => [[:space:]]
|
| 723 | [ word ] # => [[:alpha:]][[:digit:]]_
|
| 724 |
|
| 725 | Abbreviations:
|
| 726 |
|
| 727 | [ d s w ] # Same as [ digit space word ]
|
| 728 |
|
| 729 | Valid POSIX classes:
|
| 730 |
|
| 731 | alnum cntrl lower space
|
| 732 | alpha digit print upper
|
| 733 | blank graph punct xdigit
|
| 734 |
|
| 735 | Negated:
|
| 736 |
|
| 737 | !digit !space !word
|
| 738 | !d !s !w
|
| 739 | !alnum # etc.
|
| 740 |
|
| 741 | ### re-repeat
|
| 742 |
|
| 743 | Eggex repetition looks like POSIX syntax:
|
| 744 |
|
| 745 | / 'a'? / # zero or one
|
| 746 | / 'a'* / # zero or more
|
| 747 | / 'a'+ / # one or more
|
| 748 |
|
| 749 | Counted repetitions:
|
| 750 |
|
| 751 | / 'a'{3} / # exactly 3 repetitions
|
| 752 | / 'a'{2,4} / # between 2 to 4 repetitions
|
| 753 |
|
| 754 | ### re-compound
|
| 755 |
|
| 756 | Sequence expressions with a space:
|
| 757 |
|
| 758 | / word digit digit / # Matches 3 characters in sequence
|
| 759 | # Examples: a42, b51
|
| 760 |
|
| 761 | (Compare `/ [ word digit ] /`, which is a set matching 1 character.)
|
| 762 |
|
| 763 | Alternation with `|`:
|
| 764 |
|
| 765 | / word | digit / # Matches 'a' OR '9', for example
|
| 766 |
|
| 767 | Grouping with parentheses:
|
| 768 |
|
| 769 | / (word digit) | \\ / # Matches a9 or \
|
| 770 |
|
| 771 | ### re-capture
|
| 772 |
|
| 773 | To retrieve a substring of a string that matches an Eggex, use a "capture
|
| 774 | group" like `<capture ...>`.
|
| 775 |
|
| 776 | Here's an eggex with a **positional** capture:
|
| 777 |
|
| 778 | var pat = / 'hi ' <capture d+> / # access with _group(1)
|
| 779 | # or Match => _group(1)
|
| 780 |
|
| 781 | Captures can be **named**:
|
| 782 |
|
| 783 | <capture d+ as month> # access with _group('month')
|
| 784 | # or Match => group('month')
|
| 785 |
|
| 786 | Captures can also have a type **conversion func**:
|
| 787 |
|
| 788 | <capture d+ : int> # _group(1) returns Int
|
| 789 |
|
| 790 | <capture d+ as month: int> # _group('month') returns Int
|
| 791 |
|
| 792 | Related docs and help topics:
|
| 793 |
|
| 794 | - [YSH Regex API](../ysh-regex-api.html)
|
| 795 | - [`_group()`](chap-builtin-func.html#_group)
|
| 796 | - [`Match => group()`](chap-type-method.html#group)
|
| 797 |
|
| 798 | ### re-splice
|
| 799 |
|
| 800 | To build an eggex out of smaller expressions, you can **splice** eggexes
|
| 801 | together:
|
| 802 |
|
| 803 | var D = / [0-9][0-9] /
|
| 804 | var time = / @D ':' @D / # [0-9][0-9]:[0-9][0-9]
|
| 805 |
|
| 806 | If the variable begins with a capital letter, you can omit `@`:
|
| 807 |
|
| 808 | var ip = / D ':' D /
|
| 809 |
|
| 810 | You can also splice a string:
|
| 811 |
|
| 812 | var greeting = 'hi'
|
| 813 | var pat = / @greeting ' world' / # hi world
|
| 814 |
|
| 815 | Splicing is **not** string concatenation; it works on eggex subtrees.
|
| 816 |
|
| 817 | ### re-flags
|
| 818 |
|
| 819 | Valid ERE flags, which are passed to libc's `regcomp()`:
|
| 820 |
|
| 821 | - `reg_icase` aka `i` - ignore case
|
| 822 | - `reg_newline` - 4 matching changes related to newlines
|
| 823 |
|
| 824 | See `man regcomp`.
|
| 825 |
|
| 826 | ### re-multiline
|
| 827 |
|
| 828 | Multi-line eggexes aren't yet implemented. Splicing makes it less necessary:
|
| 829 |
|
| 830 | var Name = / <capture [a-z]+ as name> /
|
| 831 | var Num = / <capture d+ as num> /
|
| 832 | var Space = / <capture s+ as space> /
|
| 833 |
|
| 834 | # For variables named like CapWords, splicing @Name doesn't require @
|
| 835 | var lexer = / Name | Num | Space /
|