OILS / tools / ysh_ify.py View on Github | oilshell.org

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