1 | #!/usr/bin/env python2
|
2 | """osh/word_compile.py.
|
3 |
|
4 | These functions are called after parsing, but don't depend on any runtime
|
5 | values.
|
6 | """
|
7 | from _devbuild.gen.id_kind_asdl import Id, Id_t, Id_str
|
8 | from _devbuild.gen.syntax_asdl import (
|
9 | Token,
|
10 | CharCode,
|
11 | word_part_e,
|
12 | word_part_t,
|
13 | )
|
14 | from data_lang import j8
|
15 | from frontend import consts
|
16 | from frontend import lexer
|
17 | from mycpp import mylib
|
18 | from mycpp.mylib import log, switch
|
19 |
|
20 | from typing import List, Optional, cast
|
21 |
|
22 |
|
23 | def EvalCharLiteralForRegex(tok):
|
24 | # type: (Token) -> CharCode
|
25 | """For regex char classes.
|
26 |
|
27 | Similar logic as below.
|
28 | """
|
29 | id_ = tok.id
|
30 | value = lexer.TokenVal(tok)
|
31 |
|
32 | with switch(id_) as case:
|
33 | if case(Id.Char_UBraced):
|
34 | s = lexer.TokenSlice(tok, 3, -1) # \u{123}
|
35 | i = int(s, 16)
|
36 | return CharCode(tok, i, True) # u_braced
|
37 |
|
38 | elif case(Id.Char_OneChar): # \'
|
39 | # value[1] -> mylib.ByteAt()
|
40 | one_char_str = consts.LookupCharC(value[1])
|
41 | return CharCode(tok, ord(one_char_str), False)
|
42 |
|
43 | elif case(Id.Char_Hex):
|
44 | s = lexer.TokenSliceLeft(tok, 2)
|
45 | i = int(s, 16)
|
46 | return CharCode(tok, i, False)
|
47 |
|
48 | elif case(Id.Lit_Chars, Id.Expr_Name, Id.Expr_DecInt):
|
49 | # Id.Lit_Chars: Token in single quoted string ['a'] is Id.Lit_Chars
|
50 | # Id.Expr_Name: [a-z] is ['a'-'Z'], and [a z] is ['a' 'Z']
|
51 | # Id.Expr_DecInt: [0-9] is ['0'-'9'], and [0 9] is ['0' '9']
|
52 |
|
53 | assert len(value) == 1, tok
|
54 | # value[0] -> mylib.ByteAt()
|
55 | return CharCode(tok, ord(value[0]), False)
|
56 |
|
57 | else:
|
58 | raise AssertionError(tok)
|
59 |
|
60 |
|
61 | def EvalCStringToken(id_, value):
|
62 | # type: (Id_t, str) -> Optional[str]
|
63 | """This function is shared between echo -e and $''.
|
64 |
|
65 | $'' could use it at compile time, much like brace expansion in braces.py.
|
66 | """
|
67 | if id_ in (Id.Lit_Chars, Id.Lit_CharsWithoutPrefix, Id.Unknown_Backslash,
|
68 | Id.Char_AsciiControl):
|
69 | # shopt -u parse_backslash detects Unknown_Backslash at PARSE time in YSH.
|
70 |
|
71 | # Char_AsciiControl is allowed in YSH code, for newlines in u''
|
72 | # strings, just like r'' has
|
73 | # TODO: could allow ONLY newline?
|
74 | return value
|
75 |
|
76 | # single quotes in the middle of a triple quoted string
|
77 | elif id_ == Id.Right_SingleQuote:
|
78 | return value
|
79 |
|
80 | elif id_ == Id.Char_OneChar:
|
81 | c = value[1]
|
82 | return consts.LookupCharC(c)
|
83 |
|
84 | elif id_ == Id.Char_Stop: # \c returns a special sentinel
|
85 | return None
|
86 |
|
87 | elif id_ in (Id.Char_Octal3, Id.Char_Octal4):
|
88 | if id_ == Id.Char_Octal3: # $'\377' (disallowed at parse time in YSH)
|
89 | s = value[1:]
|
90 | else: # echo -e '\0377'
|
91 | s = value[2:]
|
92 |
|
93 | i = int(s, 8)
|
94 | if i >= 256:
|
95 | i = i % 256
|
96 | # NOTE: This is for strict mode
|
97 | #raise AssertionError('Out of range')
|
98 | return chr(i)
|
99 |
|
100 | elif id_ in (Id.Char_Hex, Id.Char_YHex):
|
101 | s = value[2:]
|
102 | i = int(s, 16)
|
103 | return chr(i)
|
104 |
|
105 | elif id_ in (Id.Char_Unicode4, Id.Char_Unicode8):
|
106 | s = value[2:]
|
107 | i = int(s, 16)
|
108 | #util.log('i = %d', i)
|
109 | return j8.Utf8Encode(i)
|
110 |
|
111 | elif id_ == Id.Char_UBraced:
|
112 | s = value[3:-1] # \u{123}
|
113 | i = int(s, 16)
|
114 | return j8.Utf8Encode(i)
|
115 |
|
116 | else:
|
117 | raise AssertionError(Id_str(id_))
|
118 |
|
119 |
|
120 | def EvalSingleQuoted2(id_, tokens):
|
121 | # type: (Id_t, List[Token]) -> str
|
122 | """ Done at parse time """
|
123 | if id_ in (Id.Left_SingleQuote, Id.Left_RSingleQuote, Id.Left_TSingleQuote,
|
124 | Id.Left_RTSingleQuote):
|
125 | strs = [lexer.TokenVal(t) for t in tokens]
|
126 |
|
127 | elif id_ in (Id.Left_DollarSingleQuote, Id.Left_USingleQuote,
|
128 | Id.Left_BSingleQuote, Id.Left_UTSingleQuote,
|
129 | Id.Left_BTSingleQuote):
|
130 | if 0:
|
131 | for t in tokens:
|
132 | print('T %s' % t)
|
133 |
|
134 | strs = [EvalCStringToken(t.id, lexer.TokenVal(t)) for t in tokens]
|
135 |
|
136 | else:
|
137 | raise AssertionError(id_)
|
138 | return ''.join(strs)
|
139 |
|
140 |
|
141 | def _TokenConsistsOf(tok, byte_set):
|
142 | # type: (Token, str) -> bool
|
143 | start = tok.col
|
144 | end = tok.col + tok.length
|
145 | for i in xrange(start, end):
|
146 | b = mylib.ByteAt(tok.line.content, i)
|
147 | if not mylib.ByteInSet(b, byte_set):
|
148 | return False
|
149 | return True
|
150 |
|
151 |
|
152 | def _IsLeadingSpace(tok):
|
153 | # type: (Token) -> bool
|
154 | """ Determine if the token before ''' etc. is space to trim """
|
155 | return _TokenConsistsOf(tok, ' \t')
|
156 |
|
157 |
|
158 | def _IsTrailingSpace(tok):
|
159 | # type: (Token) -> bool
|
160 | """ Determine if the space/newlines after ''' should be trimmed
|
161 |
|
162 | Like s.isspace(), without legacy \f \v and Unicode.
|
163 | """
|
164 | return _TokenConsistsOf(tok, ' \n\r\t')
|
165 |
|
166 |
|
167 | # Whitespace trimming algorithms:
|
168 | #
|
169 | # 1. Trim what's after opening ''' or """, if it's whitespace
|
170 | # 2. Determine what's before closing ''' or """ -- this is what you strip
|
171 | # 3. Strip each line by mutating the token
|
172 | # - Change the ID from Id.Lit_Chars -> Id.Lit_CharsWithoutPrefix to maintain
|
173 | # the lossless invariant
|
174 |
|
175 |
|
176 | def RemoveLeadingSpaceDQ(parts):
|
177 | # type: (List[word_part_t]) -> None
|
178 | if len(parts) <= 1: # We need at least 2 parts to strip anything
|
179 | return
|
180 |
|
181 | # The first token may have a newline
|
182 | UP_first = parts[0]
|
183 | if UP_first.tag() == word_part_e.Literal:
|
184 | first = cast(Token, UP_first)
|
185 | #log('T %s', first_part)
|
186 | if _IsTrailingSpace(first):
|
187 | # Remove the first part. TODO: This could be expensive if there are many
|
188 | # lines.
|
189 | parts.pop(0)
|
190 |
|
191 | UP_last = parts[-1]
|
192 | to_strip = None # type: Optional[str]
|
193 | if UP_last.tag() == word_part_e.Literal:
|
194 | last = cast(Token, UP_last)
|
195 | if _IsLeadingSpace(last):
|
196 | to_strip = lexer.TokenVal(last)
|
197 | parts.pop() # Remove the last part
|
198 |
|
199 | if to_strip is None:
|
200 | return
|
201 |
|
202 | n = len(to_strip)
|
203 | for part in parts:
|
204 | if part.tag() != word_part_e.Literal:
|
205 | line_ended = False
|
206 | continue
|
207 |
|
208 | lit_tok = cast(Token, part)
|
209 |
|
210 | if lit_tok.col == 0 and lexer.TokenStartsWith(lit_tok, to_strip):
|
211 | # TODO: Lexer should not populate this!
|
212 | assert lit_tok.tval is None, lit_tok.tval
|
213 |
|
214 | lit_tok.col = n
|
215 | lit_tok.length -= n
|
216 | #log('n = %d, %s', n, lit_tok)
|
217 |
|
218 | assert lit_tok.id == Id.Lit_Chars, lit_tok
|
219 | # --tool lossless-cat has a special case for this
|
220 | lit_tok.id = Id.Lit_CharsWithoutPrefix
|
221 |
|
222 |
|
223 | def RemoveLeadingSpaceSQ(tokens):
|
224 | # type: (List[Token]) -> None
|
225 | """Strip leading whitespace from tokens.
|
226 |
|
227 | May return original list unmodified, or a new list.
|
228 |
|
229 | Must respect lossless invariant - see test/lossless/multiline-str.sh
|
230 |
|
231 | For now we create NEW Id.Ignored_LeadingSpace tokens, and are NOT in the
|
232 | arena.
|
233 |
|
234 | Quirk to make more consistent:
|
235 | In $''' and r''' and ''', we have Lit_Chars \n
|
236 | In u''' and b''', we have Char_AsciiControl \n
|
237 | """
|
238 | if 0:
|
239 | log('--')
|
240 | for tok in tokens:
|
241 | log('tok %s', tok)
|
242 | log('--')
|
243 |
|
244 | if len(tokens) <= 1: # We need at least 2 parts to strip anything
|
245 | return
|
246 |
|
247 | # var x = ''' # strip initial newline/whitespace
|
248 | # x
|
249 | # '''
|
250 | first = tokens[0]
|
251 | if first.id in (Id.Lit_Chars, Id.Char_AsciiControl):
|
252 | if _IsTrailingSpace(first):
|
253 | tokens.pop(0) # Remove the first part
|
254 |
|
255 | # Figure out what to strip, based on last token
|
256 | last = tokens[-1]
|
257 | to_strip = None # type: Optional[str]
|
258 | if last.id in (Id.Lit_Chars, Id.Char_AsciiControl):
|
259 | if _IsLeadingSpace(last):
|
260 | to_strip = lexer.TokenVal(last)
|
261 | tokens.pop() # Remove the last part
|
262 |
|
263 | if to_strip is None:
|
264 | return
|
265 |
|
266 | #log('SQ Stripping %r', to_strip)
|
267 | n = len(to_strip)
|
268 |
|
269 | #log('--')
|
270 | for tok in tokens: # line_ended reset on every iteration
|
271 | #log('tok %s', tok)
|
272 | # Strip leading space on tokens that begin lines, by bumping start col
|
273 | if tok.col == 0 and lexer.TokenStartsWith(tok, to_strip):
|
274 | tok.col = n
|
275 | tok.length -= n
|
276 |
|
277 | assert tok.id == Id.Lit_Chars, tok
|
278 | # --tool lossless-cat has a special case for this
|
279 | tok.id = Id.Lit_CharsWithoutPrefix
|
280 |
|
281 | #log('STRIP tok %s', tok)
|