Show More
@@ -1,798 +1,799 b'' | |||
|
1 | 1 | """Input transformer machinery to support IPython special syntax. |
|
2 | 2 | |
|
3 | 3 | This includes the machinery to recognise and transform ``%magic`` commands, |
|
4 | 4 | ``!system`` commands, ``help?`` querying, prompt stripping, and so forth. |
|
5 | 5 | |
|
6 | 6 | Added: IPython 7.0. Replaces inputsplitter and inputtransformer which were |
|
7 | 7 | deprecated in 7.0. |
|
8 | 8 | """ |
|
9 | 9 | |
|
10 | 10 | # Copyright (c) IPython Development Team. |
|
11 | 11 | # Distributed under the terms of the Modified BSD License. |
|
12 | 12 | |
|
13 | 13 | import ast |
|
14 | 14 | import sys |
|
15 | 15 | from codeop import CommandCompiler, Compile |
|
16 | 16 | import re |
|
17 | 17 | import tokenize |
|
18 | 18 | from typing import List, Tuple, Optional, Any |
|
19 | 19 | import warnings |
|
20 | 20 | |
|
21 | 21 | _indent_re = re.compile(r'^[ \t]+') |
|
22 | 22 | |
|
23 | 23 | def leading_empty_lines(lines): |
|
24 | 24 | """Remove leading empty lines |
|
25 | 25 | |
|
26 | 26 | If the leading lines are empty or contain only whitespace, they will be |
|
27 | 27 | removed. |
|
28 | 28 | """ |
|
29 | 29 | if not lines: |
|
30 | 30 | return lines |
|
31 | 31 | for i, line in enumerate(lines): |
|
32 | 32 | if line and not line.isspace(): |
|
33 | 33 | return lines[i:] |
|
34 | 34 | return lines |
|
35 | 35 | |
|
36 | 36 | def leading_indent(lines): |
|
37 | 37 | """Remove leading indentation. |
|
38 | 38 | |
|
39 | 39 | If the first line starts with a spaces or tabs, the same whitespace will be |
|
40 | 40 | removed from each following line in the cell. |
|
41 | 41 | """ |
|
42 | 42 | if not lines: |
|
43 | 43 | return lines |
|
44 | 44 | m = _indent_re.match(lines[0]) |
|
45 | 45 | if not m: |
|
46 | 46 | return lines |
|
47 | 47 | space = m.group(0) |
|
48 | 48 | n = len(space) |
|
49 | 49 | return [l[n:] if l.startswith(space) else l |
|
50 | 50 | for l in lines] |
|
51 | 51 | |
|
52 | 52 | class PromptStripper: |
|
53 | 53 | """Remove matching input prompts from a block of input. |
|
54 | 54 | |
|
55 | 55 | Parameters |
|
56 | 56 | ---------- |
|
57 | 57 | prompt_re : regular expression |
|
58 | 58 | A regular expression matching any input prompt (including continuation, |
|
59 | 59 | e.g. ``...``) |
|
60 | 60 | initial_re : regular expression, optional |
|
61 | 61 | A regular expression matching only the initial prompt, but not continuation. |
|
62 | 62 | If no initial expression is given, prompt_re will be used everywhere. |
|
63 | 63 | Used mainly for plain Python prompts (``>>>``), where the continuation prompt |
|
64 | 64 | ``...`` is a valid Python expression in Python 3, so shouldn't be stripped. |
|
65 | 65 | |
|
66 | 66 | Notes |
|
67 | 67 | ----- |
|
68 | 68 | |
|
69 | 69 | If initial_re and prompt_re differ, |
|
70 | 70 | only initial_re will be tested against the first line. |
|
71 | 71 | If any prompt is found on the first two lines, |
|
72 | 72 | prompts will be stripped from the rest of the block. |
|
73 | 73 | """ |
|
74 | 74 | def __init__(self, prompt_re, initial_re=None): |
|
75 | 75 | self.prompt_re = prompt_re |
|
76 | 76 | self.initial_re = initial_re or prompt_re |
|
77 | 77 | |
|
78 | 78 | def _strip(self, lines): |
|
79 | 79 | return [self.prompt_re.sub('', l, count=1) for l in lines] |
|
80 | 80 | |
|
81 | 81 | def __call__(self, lines): |
|
82 | 82 | if not lines: |
|
83 | 83 | return lines |
|
84 | 84 | if self.initial_re.match(lines[0]) or \ |
|
85 | 85 | (len(lines) > 1 and self.prompt_re.match(lines[1])): |
|
86 | 86 | return self._strip(lines) |
|
87 | 87 | return lines |
|
88 | 88 | |
|
89 | 89 | classic_prompt = PromptStripper( |
|
90 | 90 | prompt_re=re.compile(r'^(>>>|\.\.\.)( |$)'), |
|
91 | 91 | initial_re=re.compile(r'^>>>( |$)') |
|
92 | 92 | ) |
|
93 | 93 | |
|
94 | 94 | ipython_prompt = PromptStripper( |
|
95 | 95 | re.compile( |
|
96 | 96 | r""" |
|
97 | 97 | ^( # Match from the beginning of a line, either: |
|
98 | 98 | |
|
99 | 99 | # 1. First-line prompt: |
|
100 | 100 | ((\[nav\]|\[ins\])?\ )? # Vi editing mode prompt, if it's there |
|
101 | 101 | In\ # The 'In' of the prompt, with a space |
|
102 | 102 | \[\d+\]: # Command index, as displayed in the prompt |
|
103 | 103 | \ # With a mandatory trailing space |
|
104 | 104 | |
|
105 | 105 | | # ... or ... |
|
106 | 106 | |
|
107 | 107 | # 2. The three dots of the multiline prompt |
|
108 | 108 | \s* # All leading whitespace characters |
|
109 | 109 | \.{3,}: # The three (or more) dots |
|
110 | 110 | \ ? # With an optional trailing space |
|
111 | 111 | |
|
112 | 112 | ) |
|
113 | 113 | """, |
|
114 | 114 | re.VERBOSE, |
|
115 | 115 | ) |
|
116 | 116 | ) |
|
117 | 117 | |
|
118 | 118 | |
|
119 | 119 | def cell_magic(lines): |
|
120 | 120 | if not lines or not lines[0].startswith('%%'): |
|
121 | 121 | return lines |
|
122 | 122 | if re.match(r'%%\w+\?', lines[0]): |
|
123 | 123 | # This case will be handled by help_end |
|
124 | 124 | return lines |
|
125 | 125 | magic_name, _, first_line = lines[0][2:].rstrip().partition(' ') |
|
126 | 126 | body = ''.join(lines[1:]) |
|
127 | 127 | return ['get_ipython().run_cell_magic(%r, %r, %r)\n' |
|
128 | 128 | % (magic_name, first_line, body)] |
|
129 | 129 | |
|
130 | 130 | |
|
131 | 131 | def _find_assign_op(token_line) -> Optional[int]: |
|
132 | 132 | """Get the index of the first assignment in the line ('=' not inside brackets) |
|
133 | 133 | |
|
134 | 134 | Note: We don't try to support multiple special assignment (a = b = %foo) |
|
135 | 135 | """ |
|
136 | 136 | paren_level = 0 |
|
137 | 137 | for i, ti in enumerate(token_line): |
|
138 | 138 | s = ti.string |
|
139 | 139 | if s == '=' and paren_level == 0: |
|
140 | 140 | return i |
|
141 | 141 | if s in {'(','[','{'}: |
|
142 | 142 | paren_level += 1 |
|
143 | 143 | elif s in {')', ']', '}'}: |
|
144 | 144 | if paren_level > 0: |
|
145 | 145 | paren_level -= 1 |
|
146 | 146 | return None |
|
147 | 147 | |
|
148 | 148 | def find_end_of_continued_line(lines, start_line: int): |
|
149 | 149 | """Find the last line of a line explicitly extended using backslashes. |
|
150 | 150 | |
|
151 | 151 | Uses 0-indexed line numbers. |
|
152 | 152 | """ |
|
153 | 153 | end_line = start_line |
|
154 | 154 | while lines[end_line].endswith('\\\n'): |
|
155 | 155 | end_line += 1 |
|
156 | 156 | if end_line >= len(lines): |
|
157 | 157 | break |
|
158 | 158 | return end_line |
|
159 | 159 | |
|
160 | 160 | def assemble_continued_line(lines, start: Tuple[int, int], end_line: int): |
|
161 | 161 | r"""Assemble a single line from multiple continued line pieces |
|
162 | 162 | |
|
163 | 163 | Continued lines are lines ending in ``\``, and the line following the last |
|
164 | 164 | ``\`` in the block. |
|
165 | 165 | |
|
166 | 166 | For example, this code continues over multiple lines:: |
|
167 | 167 | |
|
168 | 168 | if (assign_ix is not None) \ |
|
169 | 169 | and (len(line) >= assign_ix + 2) \ |
|
170 | 170 | and (line[assign_ix+1].string == '%') \ |
|
171 | 171 | and (line[assign_ix+2].type == tokenize.NAME): |
|
172 | 172 | |
|
173 | 173 | This statement contains four continued line pieces. |
|
174 | 174 | Assembling these pieces into a single line would give:: |
|
175 | 175 | |
|
176 | 176 | if (assign_ix is not None) and (len(line) >= assign_ix + 2) and (line[... |
|
177 | 177 | |
|
178 | 178 | This uses 0-indexed line numbers. *start* is (lineno, colno). |
|
179 | 179 | |
|
180 | 180 | Used to allow ``%magic`` and ``!system`` commands to be continued over |
|
181 | 181 | multiple lines. |
|
182 | 182 | """ |
|
183 | 183 | parts = [lines[start[0]][start[1]:]] + lines[start[0]+1:end_line+1] |
|
184 | 184 | return ' '.join([p.rstrip()[:-1] for p in parts[:-1]] # Strip backslash+newline |
|
185 | 185 | + [parts[-1].rstrip()]) # Strip newline from last line |
|
186 | 186 | |
|
187 | 187 | class TokenTransformBase: |
|
188 | 188 | """Base class for transformations which examine tokens. |
|
189 | 189 | |
|
190 | 190 | Special syntax should not be transformed when it occurs inside strings or |
|
191 | 191 | comments. This is hard to reliably avoid with regexes. The solution is to |
|
192 | 192 | tokenise the code as Python, and recognise the special syntax in the tokens. |
|
193 | 193 | |
|
194 | 194 | IPython's special syntax is not valid Python syntax, so tokenising may go |
|
195 | 195 | wrong after the special syntax starts. These classes therefore find and |
|
196 | 196 | transform *one* instance of special syntax at a time into regular Python |
|
197 | 197 | syntax. After each transformation, tokens are regenerated to find the next |
|
198 | 198 | piece of special syntax. |
|
199 | 199 | |
|
200 | 200 | Subclasses need to implement one class method (find) |
|
201 | 201 | and one regular method (transform). |
|
202 | 202 | |
|
203 | 203 | The priority attribute can select which transformation to apply if multiple |
|
204 | 204 | transformers match in the same place. Lower numbers have higher priority. |
|
205 | 205 | This allows "%magic?" to be turned into a help call rather than a magic call. |
|
206 | 206 | """ |
|
207 | 207 | # Lower numbers -> higher priority (for matches in the same location) |
|
208 | 208 | priority = 10 |
|
209 | 209 | |
|
210 | 210 | def sortby(self): |
|
211 | 211 | return self.start_line, self.start_col, self.priority |
|
212 | 212 | |
|
213 | 213 | def __init__(self, start): |
|
214 | 214 | self.start_line = start[0] - 1 # Shift from 1-index to 0-index |
|
215 | 215 | self.start_col = start[1] |
|
216 | 216 | |
|
217 | 217 | @classmethod |
|
218 | 218 | def find(cls, tokens_by_line): |
|
219 | 219 | """Find one instance of special syntax in the provided tokens. |
|
220 | 220 | |
|
221 | 221 | Tokens are grouped into logical lines for convenience, |
|
222 | 222 | so it is easy to e.g. look at the first token of each line. |
|
223 | 223 | *tokens_by_line* is a list of lists of tokenize.TokenInfo objects. |
|
224 | 224 | |
|
225 | 225 | This should return an instance of its class, pointing to the start |
|
226 | 226 | position it has found, or None if it found no match. |
|
227 | 227 | """ |
|
228 | 228 | raise NotImplementedError |
|
229 | 229 | |
|
230 | 230 | def transform(self, lines: List[str]): |
|
231 | 231 | """Transform one instance of special syntax found by ``find()`` |
|
232 | 232 | |
|
233 | 233 | Takes a list of strings representing physical lines, |
|
234 | 234 | returns a similar list of transformed lines. |
|
235 | 235 | """ |
|
236 | 236 | raise NotImplementedError |
|
237 | 237 | |
|
238 | 238 | class MagicAssign(TokenTransformBase): |
|
239 | 239 | """Transformer for assignments from magics (a = %foo)""" |
|
240 | 240 | @classmethod |
|
241 | 241 | def find(cls, tokens_by_line): |
|
242 | 242 | """Find the first magic assignment (a = %foo) in the cell. |
|
243 | 243 | """ |
|
244 | 244 | for line in tokens_by_line: |
|
245 | 245 | assign_ix = _find_assign_op(line) |
|
246 | 246 | if (assign_ix is not None) \ |
|
247 | 247 | and (len(line) >= assign_ix + 2) \ |
|
248 | 248 | and (line[assign_ix+1].string == '%') \ |
|
249 | 249 | and (line[assign_ix+2].type == tokenize.NAME): |
|
250 | 250 | return cls(line[assign_ix+1].start) |
|
251 | 251 | |
|
252 | 252 | def transform(self, lines: List[str]): |
|
253 | 253 | """Transform a magic assignment found by the ``find()`` classmethod. |
|
254 | 254 | """ |
|
255 | 255 | start_line, start_col = self.start_line, self.start_col |
|
256 | 256 | lhs = lines[start_line][:start_col] |
|
257 | 257 | end_line = find_end_of_continued_line(lines, start_line) |
|
258 | 258 | rhs = assemble_continued_line(lines, (start_line, start_col), end_line) |
|
259 | 259 | assert rhs.startswith('%'), rhs |
|
260 | 260 | magic_name, _, args = rhs[1:].partition(' ') |
|
261 | 261 | |
|
262 | 262 | lines_before = lines[:start_line] |
|
263 | 263 | call = "get_ipython().run_line_magic({!r}, {!r})".format(magic_name, args) |
|
264 | 264 | new_line = lhs + call + '\n' |
|
265 | 265 | lines_after = lines[end_line+1:] |
|
266 | 266 | |
|
267 | 267 | return lines_before + [new_line] + lines_after |
|
268 | 268 | |
|
269 | 269 | |
|
270 | 270 | class SystemAssign(TokenTransformBase): |
|
271 | 271 | """Transformer for assignments from system commands (a = !foo)""" |
|
272 | 272 | @classmethod |
|
273 | 273 | def find(cls, tokens_by_line): |
|
274 | 274 | """Find the first system assignment (a = !foo) in the cell. |
|
275 | 275 | """ |
|
276 | 276 | for line in tokens_by_line: |
|
277 | 277 | assign_ix = _find_assign_op(line) |
|
278 | 278 | if (assign_ix is not None) \ |
|
279 | 279 | and not line[assign_ix].line.strip().startswith('=') \ |
|
280 | 280 | and (len(line) >= assign_ix + 2) \ |
|
281 | 281 | and (line[assign_ix + 1].type == tokenize.ERRORTOKEN): |
|
282 | 282 | ix = assign_ix + 1 |
|
283 | 283 | |
|
284 | 284 | while ix < len(line) and line[ix].type == tokenize.ERRORTOKEN: |
|
285 | 285 | if line[ix].string == '!': |
|
286 | 286 | return cls(line[ix].start) |
|
287 | 287 | elif not line[ix].string.isspace(): |
|
288 | 288 | break |
|
289 | 289 | ix += 1 |
|
290 | 290 | |
|
291 | 291 | def transform(self, lines: List[str]): |
|
292 | 292 | """Transform a system assignment found by the ``find()`` classmethod. |
|
293 | 293 | """ |
|
294 | 294 | start_line, start_col = self.start_line, self.start_col |
|
295 | 295 | |
|
296 | 296 | lhs = lines[start_line][:start_col] |
|
297 | 297 | end_line = find_end_of_continued_line(lines, start_line) |
|
298 | 298 | rhs = assemble_continued_line(lines, (start_line, start_col), end_line) |
|
299 | 299 | assert rhs.startswith('!'), rhs |
|
300 | 300 | cmd = rhs[1:] |
|
301 | 301 | |
|
302 | 302 | lines_before = lines[:start_line] |
|
303 | 303 | call = "get_ipython().getoutput({!r})".format(cmd) |
|
304 | 304 | new_line = lhs + call + '\n' |
|
305 | 305 | lines_after = lines[end_line + 1:] |
|
306 | 306 | |
|
307 | 307 | return lines_before + [new_line] + lines_after |
|
308 | 308 | |
|
309 | 309 | # The escape sequences that define the syntax transformations IPython will |
|
310 | 310 | # apply to user input. These can NOT be just changed here: many regular |
|
311 | 311 | # expressions and other parts of the code may use their hardcoded values, and |
|
312 | 312 | # for all intents and purposes they constitute the 'IPython syntax', so they |
|
313 | 313 | # should be considered fixed. |
|
314 | 314 | |
|
315 | 315 | ESC_SHELL = '!' # Send line to underlying system shell |
|
316 | 316 | ESC_SH_CAP = '!!' # Send line to system shell and capture output |
|
317 | 317 | ESC_HELP = '?' # Find information about object |
|
318 | 318 | ESC_HELP2 = '??' # Find extra-detailed information about object |
|
319 | 319 | ESC_MAGIC = '%' # Call magic function |
|
320 | 320 | ESC_MAGIC2 = '%%' # Call cell-magic function |
|
321 | 321 | ESC_QUOTE = ',' # Split args on whitespace, quote each as string and call |
|
322 | 322 | ESC_QUOTE2 = ';' # Quote all args as a single string, call |
|
323 | 323 | ESC_PAREN = '/' # Call first argument with rest of line as arguments |
|
324 | 324 | |
|
325 | 325 | ESCAPE_SINGLES = {'!', '?', '%', ',', ';', '/'} |
|
326 | 326 | ESCAPE_DOUBLES = {'!!', '??'} # %% (cell magic) is handled separately |
|
327 | 327 | |
|
328 | 328 | def _make_help_call(target, esc, next_input=None): |
|
329 | 329 | """Prepares a pinfo(2)/psearch call from a target name and the escape |
|
330 | 330 | (i.e. ? or ??)""" |
|
331 | 331 | method = 'pinfo2' if esc == '??' \ |
|
332 | 332 | else 'psearch' if '*' in target \ |
|
333 | 333 | else 'pinfo' |
|
334 | 334 | arg = " ".join([method, target]) |
|
335 | 335 | #Prepare arguments for get_ipython().run_line_magic(magic_name, magic_args) |
|
336 | 336 | t_magic_name, _, t_magic_arg_s = arg.partition(' ') |
|
337 | 337 | t_magic_name = t_magic_name.lstrip(ESC_MAGIC) |
|
338 | 338 | if next_input is None: |
|
339 | 339 | return 'get_ipython().run_line_magic(%r, %r)' % (t_magic_name, t_magic_arg_s) |
|
340 | 340 | else: |
|
341 | 341 | return 'get_ipython().set_next_input(%r);get_ipython().run_line_magic(%r, %r)' % \ |
|
342 | 342 | (next_input, t_magic_name, t_magic_arg_s) |
|
343 | 343 | |
|
344 | 344 | def _tr_help(content): |
|
345 | 345 | """Translate lines escaped with: ? |
|
346 | 346 | |
|
347 | 347 | A naked help line should fire the intro help screen (shell.show_usage()) |
|
348 | 348 | """ |
|
349 | 349 | if not content: |
|
350 | 350 | return 'get_ipython().show_usage()' |
|
351 | 351 | |
|
352 | 352 | return _make_help_call(content, '?') |
|
353 | 353 | |
|
354 | 354 | def _tr_help2(content): |
|
355 | 355 | """Translate lines escaped with: ?? |
|
356 | 356 | |
|
357 | 357 | A naked help line should fire the intro help screen (shell.show_usage()) |
|
358 | 358 | """ |
|
359 | 359 | if not content: |
|
360 | 360 | return 'get_ipython().show_usage()' |
|
361 | 361 | |
|
362 | 362 | return _make_help_call(content, '??') |
|
363 | 363 | |
|
364 | 364 | def _tr_magic(content): |
|
365 | 365 | "Translate lines escaped with a percent sign: %" |
|
366 | 366 | name, _, args = content.partition(' ') |
|
367 | 367 | return 'get_ipython().run_line_magic(%r, %r)' % (name, args) |
|
368 | 368 | |
|
369 | 369 | def _tr_quote(content): |
|
370 | 370 | "Translate lines escaped with a comma: ," |
|
371 | 371 | name, _, args = content.partition(' ') |
|
372 | 372 | return '%s("%s")' % (name, '", "'.join(args.split()) ) |
|
373 | 373 | |
|
374 | 374 | def _tr_quote2(content): |
|
375 | 375 | "Translate lines escaped with a semicolon: ;" |
|
376 | 376 | name, _, args = content.partition(' ') |
|
377 | 377 | return '%s("%s")' % (name, args) |
|
378 | 378 | |
|
379 | 379 | def _tr_paren(content): |
|
380 | 380 | "Translate lines escaped with a slash: /" |
|
381 | 381 | name, _, args = content.partition(' ') |
|
382 | 382 | return '%s(%s)' % (name, ", ".join(args.split())) |
|
383 | 383 | |
|
384 | 384 | tr = { ESC_SHELL : 'get_ipython().system({!r})'.format, |
|
385 | 385 | ESC_SH_CAP : 'get_ipython().getoutput({!r})'.format, |
|
386 | 386 | ESC_HELP : _tr_help, |
|
387 | 387 | ESC_HELP2 : _tr_help2, |
|
388 | 388 | ESC_MAGIC : _tr_magic, |
|
389 | 389 | ESC_QUOTE : _tr_quote, |
|
390 | 390 | ESC_QUOTE2 : _tr_quote2, |
|
391 | 391 | ESC_PAREN : _tr_paren } |
|
392 | 392 | |
|
393 | 393 | class EscapedCommand(TokenTransformBase): |
|
394 | 394 | """Transformer for escaped commands like %foo, !foo, or /foo""" |
|
395 | 395 | @classmethod |
|
396 | 396 | def find(cls, tokens_by_line): |
|
397 | 397 | """Find the first escaped command (%foo, !foo, etc.) in the cell. |
|
398 | 398 | """ |
|
399 | 399 | for line in tokens_by_line: |
|
400 | 400 | if not line: |
|
401 | 401 | continue |
|
402 | 402 | ix = 0 |
|
403 | 403 | ll = len(line) |
|
404 | 404 | while ll > ix and line[ix].type in {tokenize.INDENT, tokenize.DEDENT}: |
|
405 | 405 | ix += 1 |
|
406 | 406 | if ix >= ll: |
|
407 | 407 | continue |
|
408 | 408 | if line[ix].string in ESCAPE_SINGLES: |
|
409 | 409 | return cls(line[ix].start) |
|
410 | 410 | |
|
411 | 411 | def transform(self, lines): |
|
412 | 412 | """Transform an escaped line found by the ``find()`` classmethod. |
|
413 | 413 | """ |
|
414 | 414 | start_line, start_col = self.start_line, self.start_col |
|
415 | 415 | |
|
416 | 416 | indent = lines[start_line][:start_col] |
|
417 | 417 | end_line = find_end_of_continued_line(lines, start_line) |
|
418 | 418 | line = assemble_continued_line(lines, (start_line, start_col), end_line) |
|
419 | 419 | |
|
420 | 420 | if len(line) > 1 and line[:2] in ESCAPE_DOUBLES: |
|
421 | 421 | escape, content = line[:2], line[2:] |
|
422 | 422 | else: |
|
423 | 423 | escape, content = line[:1], line[1:] |
|
424 | 424 | |
|
425 | 425 | if escape in tr: |
|
426 | 426 | call = tr[escape](content) |
|
427 | 427 | else: |
|
428 | 428 | call = '' |
|
429 | 429 | |
|
430 | 430 | lines_before = lines[:start_line] |
|
431 | 431 | new_line = indent + call + '\n' |
|
432 | 432 | lines_after = lines[end_line + 1:] |
|
433 | 433 | |
|
434 | 434 | return lines_before + [new_line] + lines_after |
|
435 | 435 | |
|
436 | 436 | _help_end_re = re.compile(r"""(%{0,2} |
|
437 | 437 | (?!\d)[\w*]+ # Variable name |
|
438 | 438 | (\.(?!\d)[\w*]+)* # .etc.etc |
|
439 | 439 | ) |
|
440 | 440 | (\?\??)$ # ? or ?? |
|
441 | 441 | """, |
|
442 | 442 | re.VERBOSE) |
|
443 | 443 | |
|
444 | 444 | class HelpEnd(TokenTransformBase): |
|
445 | 445 | """Transformer for help syntax: obj? and obj??""" |
|
446 | 446 | # This needs to be higher priority (lower number) than EscapedCommand so |
|
447 | 447 | # that inspecting magics (%foo?) works. |
|
448 | 448 | priority = 5 |
|
449 | 449 | |
|
450 | 450 | def __init__(self, start, q_locn): |
|
451 | 451 | super().__init__(start) |
|
452 | 452 | self.q_line = q_locn[0] - 1 # Shift from 1-indexed to 0-indexed |
|
453 | 453 | self.q_col = q_locn[1] |
|
454 | 454 | |
|
455 | 455 | @classmethod |
|
456 | 456 | def find(cls, tokens_by_line): |
|
457 | 457 | """Find the first help command (foo?) in the cell. |
|
458 | 458 | """ |
|
459 | 459 | for line in tokens_by_line: |
|
460 | 460 | # Last token is NEWLINE; look at last but one |
|
461 | 461 | if len(line) > 2 and line[-2].string == '?': |
|
462 | 462 | # Find the first token that's not INDENT/DEDENT |
|
463 | 463 | ix = 0 |
|
464 | 464 | while line[ix].type in {tokenize.INDENT, tokenize.DEDENT}: |
|
465 | 465 | ix += 1 |
|
466 | 466 | return cls(line[ix].start, line[-2].start) |
|
467 | 467 | |
|
468 | 468 | def transform(self, lines): |
|
469 | 469 | """Transform a help command found by the ``find()`` classmethod. |
|
470 | 470 | """ |
|
471 | 471 | piece = ''.join(lines[self.start_line:self.q_line+1]) |
|
472 | 472 | indent, content = piece[:self.start_col], piece[self.start_col:] |
|
473 | 473 | lines_before = lines[:self.start_line] |
|
474 | 474 | lines_after = lines[self.q_line + 1:] |
|
475 | 475 | |
|
476 | 476 | m = _help_end_re.search(content) |
|
477 | 477 | if not m: |
|
478 | 478 | raise SyntaxError(content) |
|
479 | 479 | assert m is not None, content |
|
480 | 480 | target = m.group(1) |
|
481 | 481 | esc = m.group(3) |
|
482 | 482 | |
|
483 | 483 | # If we're mid-command, put it back on the next prompt for the user. |
|
484 | 484 | next_input = None |
|
485 | 485 | if (not lines_before) and (not lines_after) \ |
|
486 | 486 | and content.strip() != m.group(0): |
|
487 | 487 | next_input = content.rstrip('?\n') |
|
488 | 488 | |
|
489 | 489 | call = _make_help_call(target, esc, next_input=next_input) |
|
490 | 490 | new_line = indent + call + '\n' |
|
491 | 491 | |
|
492 | 492 | return lines_before + [new_line] + lines_after |
|
493 | 493 | |
|
494 | 494 | def make_tokens_by_line(lines:List[str]): |
|
495 | 495 | """Tokenize a series of lines and group tokens by line. |
|
496 | 496 | |
|
497 | 497 | The tokens for a multiline Python string or expression are grouped as one |
|
498 | 498 | line. All lines except the last lines should keep their line ending ('\\n', |
|
499 | 499 | '\\r\\n') for this to properly work. Use `.splitlines(keeplineending=True)` |
|
500 | 500 | for example when passing block of text to this function. |
|
501 | 501 | |
|
502 | 502 | """ |
|
503 | 503 | # NL tokens are used inside multiline expressions, but also after blank |
|
504 | 504 | # lines or comments. This is intentional - see https://bugs.python.org/issue17061 |
|
505 | 505 | # We want to group the former case together but split the latter, so we |
|
506 | 506 | # track parentheses level, similar to the internals of tokenize. |
|
507 | 507 | |
|
508 | 508 | # reexported from token on 3.7+ |
|
509 | 509 | NEWLINE, NL = tokenize.NEWLINE, tokenize.NL # type: ignore |
|
510 | 510 | tokens_by_line:List[List[Any]] = [[]] |
|
511 | 511 | if len(lines) > 1 and not lines[0].endswith(('\n', '\r', '\r\n', '\x0b', '\x0c')): |
|
512 | 512 | warnings.warn("`make_tokens_by_line` received a list of lines which do not have lineending markers ('\\n', '\\r', '\\r\\n', '\\x0b', '\\x0c'), behavior will be unspecified") |
|
513 | 513 | parenlev = 0 |
|
514 | 514 | try: |
|
515 | 515 | for token in tokenize.generate_tokens(iter(lines).__next__): |
|
516 | 516 | tokens_by_line[-1].append(token) |
|
517 | 517 | if (token.type == NEWLINE) \ |
|
518 | 518 | or ((token.type == NL) and (parenlev <= 0)): |
|
519 | 519 | tokens_by_line.append([]) |
|
520 | 520 | elif token.string in {'(', '[', '{'}: |
|
521 | 521 | parenlev += 1 |
|
522 | 522 | elif token.string in {')', ']', '}'}: |
|
523 | 523 | if parenlev > 0: |
|
524 | 524 | parenlev -= 1 |
|
525 | 525 | except tokenize.TokenError: |
|
526 | 526 | # Input ended in a multiline string or expression. That's OK for us. |
|
527 | 527 | pass |
|
528 | 528 | |
|
529 | 529 | |
|
530 | 530 | if not tokens_by_line[-1]: |
|
531 | 531 | tokens_by_line.pop() |
|
532 | 532 | |
|
533 | 533 | |
|
534 | 534 | return tokens_by_line |
|
535 | 535 | |
|
536 | 536 | |
|
537 | 537 | def has_sunken_brackets(tokens: List[tokenize.TokenInfo]): |
|
538 | 538 | """Check if the depth of brackets in the list of tokens drops below 0""" |
|
539 | 539 | parenlev = 0 |
|
540 | 540 | for token in tokens: |
|
541 | 541 | if token.string in {"(", "[", "{"}: |
|
542 | 542 | parenlev += 1 |
|
543 | 543 | elif token.string in {")", "]", "}"}: |
|
544 | 544 | parenlev -= 1 |
|
545 | 545 | if parenlev < 0: |
|
546 | 546 | return True |
|
547 | 547 | return False |
|
548 | 548 | |
|
549 | 549 | |
|
550 | 550 | def show_linewise_tokens(s: str): |
|
551 | 551 | """For investigation and debugging""" |
|
552 | 552 | if not s.endswith('\n'): |
|
553 | 553 | s += '\n' |
|
554 | 554 | lines = s.splitlines(keepends=True) |
|
555 | 555 | for line in make_tokens_by_line(lines): |
|
556 | 556 | print("Line -------") |
|
557 | 557 | for tokinfo in line: |
|
558 | 558 | print(" ", tokinfo) |
|
559 | 559 | |
|
560 | 560 | # Arbitrary limit to prevent getting stuck in infinite loops |
|
561 | 561 | TRANSFORM_LOOP_LIMIT = 500 |
|
562 | 562 | |
|
563 | 563 | class TransformerManager: |
|
564 | 564 | """Applies various transformations to a cell or code block. |
|
565 | 565 | |
|
566 | 566 | The key methods for external use are ``transform_cell()`` |
|
567 | 567 | and ``check_complete()``. |
|
568 | 568 | """ |
|
569 | 569 | def __init__(self): |
|
570 | 570 | self.cleanup_transforms = [ |
|
571 | 571 | leading_empty_lines, |
|
572 | 572 | leading_indent, |
|
573 | 573 | classic_prompt, |
|
574 | 574 | ipython_prompt, |
|
575 | 575 | ] |
|
576 | 576 | self.line_transforms = [ |
|
577 | 577 | cell_magic, |
|
578 | 578 | ] |
|
579 | 579 | self.token_transformers = [ |
|
580 | 580 | MagicAssign, |
|
581 | 581 | SystemAssign, |
|
582 | 582 | EscapedCommand, |
|
583 | 583 | HelpEnd, |
|
584 | 584 | ] |
|
585 | 585 | |
|
586 | 586 | def do_one_token_transform(self, lines): |
|
587 | 587 | """Find and run the transform earliest in the code. |
|
588 | 588 | |
|
589 | 589 | Returns (changed, lines). |
|
590 | 590 | |
|
591 | 591 | This method is called repeatedly until changed is False, indicating |
|
592 | 592 | that all available transformations are complete. |
|
593 | 593 | |
|
594 | 594 | The tokens following IPython special syntax might not be valid, so |
|
595 | 595 | the transformed code is retokenised every time to identify the next |
|
596 | 596 | piece of special syntax. Hopefully long code cells are mostly valid |
|
597 | 597 | Python, not using lots of IPython special syntax, so this shouldn't be |
|
598 | 598 | a performance issue. |
|
599 | 599 | """ |
|
600 | 600 | tokens_by_line = make_tokens_by_line(lines) |
|
601 | 601 | candidates = [] |
|
602 | 602 | for transformer_cls in self.token_transformers: |
|
603 | 603 | transformer = transformer_cls.find(tokens_by_line) |
|
604 | 604 | if transformer: |
|
605 | 605 | candidates.append(transformer) |
|
606 | 606 | |
|
607 | 607 | if not candidates: |
|
608 | 608 | # Nothing to transform |
|
609 | 609 | return False, lines |
|
610 | 610 | ordered_transformers = sorted(candidates, key=TokenTransformBase.sortby) |
|
611 | 611 | for transformer in ordered_transformers: |
|
612 | 612 | try: |
|
613 | 613 | return True, transformer.transform(lines) |
|
614 | 614 | except SyntaxError: |
|
615 | 615 | pass |
|
616 | 616 | return False, lines |
|
617 | 617 | |
|
618 | 618 | def do_token_transforms(self, lines): |
|
619 | 619 | for _ in range(TRANSFORM_LOOP_LIMIT): |
|
620 | 620 | changed, lines = self.do_one_token_transform(lines) |
|
621 | 621 | if not changed: |
|
622 | 622 | return lines |
|
623 | 623 | |
|
624 | 624 | raise RuntimeError("Input transformation still changing after " |
|
625 | 625 | "%d iterations. Aborting." % TRANSFORM_LOOP_LIMIT) |
|
626 | 626 | |
|
627 | 627 | def transform_cell(self, cell: str) -> str: |
|
628 | 628 | """Transforms a cell of input code""" |
|
629 | 629 | if not cell.endswith('\n'): |
|
630 | 630 | cell += '\n' # Ensure the cell has a trailing newline |
|
631 | 631 | lines = cell.splitlines(keepends=True) |
|
632 | 632 | for transform in self.cleanup_transforms + self.line_transforms: |
|
633 | 633 | lines = transform(lines) |
|
634 | 634 | |
|
635 | 635 | lines = self.do_token_transforms(lines) |
|
636 | 636 | return ''.join(lines) |
|
637 | 637 | |
|
638 | 638 | def check_complete(self, cell: str): |
|
639 | 639 | """Return whether a block of code is ready to execute, or should be continued |
|
640 | 640 | |
|
641 | 641 | Parameters |
|
642 | 642 | ---------- |
|
643 | 643 | source : string |
|
644 | 644 | Python input code, which can be multiline. |
|
645 | 645 | |
|
646 | 646 | Returns |
|
647 | 647 | ------- |
|
648 | 648 | status : str |
|
649 | 649 | One of 'complete', 'incomplete', or 'invalid' if source is not a |
|
650 | 650 | prefix of valid code. |
|
651 | 651 | indent_spaces : int or None |
|
652 | 652 | The number of spaces by which to indent the next line of code. If |
|
653 | 653 | status is not 'incomplete', this is None. |
|
654 | 654 | """ |
|
655 | 655 | # Remember if the lines ends in a new line. |
|
656 | 656 | ends_with_newline = False |
|
657 | 657 | for character in reversed(cell): |
|
658 | 658 | if character == '\n': |
|
659 | 659 | ends_with_newline = True |
|
660 | 660 | break |
|
661 | 661 | elif character.strip(): |
|
662 | 662 | break |
|
663 | 663 | else: |
|
664 | 664 | continue |
|
665 | 665 | |
|
666 | 666 | if not ends_with_newline: |
|
667 | 667 | # Append an newline for consistent tokenization |
|
668 | 668 | # See https://bugs.python.org/issue33899 |
|
669 | 669 | cell += '\n' |
|
670 | 670 | |
|
671 | 671 | lines = cell.splitlines(keepends=True) |
|
672 | 672 | |
|
673 | 673 | if not lines: |
|
674 | 674 | return 'complete', None |
|
675 | 675 | |
|
676 | 676 | if lines[-1].endswith('\\'): |
|
677 | 677 | # Explicit backslash continuation |
|
678 | 678 | return 'incomplete', find_last_indent(lines) |
|
679 | 679 | |
|
680 | 680 | try: |
|
681 | 681 | for transform in self.cleanup_transforms: |
|
682 | 682 | if not getattr(transform, 'has_side_effects', False): |
|
683 | 683 | lines = transform(lines) |
|
684 | 684 | except SyntaxError: |
|
685 | 685 | return 'invalid', None |
|
686 | 686 | |
|
687 | 687 | if lines[0].startswith('%%'): |
|
688 | 688 | # Special case for cell magics - completion marked by blank line |
|
689 | 689 | if lines[-1].strip(): |
|
690 | 690 | return 'incomplete', find_last_indent(lines) |
|
691 | 691 | else: |
|
692 | 692 | return 'complete', None |
|
693 | 693 | |
|
694 | 694 | try: |
|
695 | 695 | for transform in self.line_transforms: |
|
696 | 696 | if not getattr(transform, 'has_side_effects', False): |
|
697 | 697 | lines = transform(lines) |
|
698 | 698 | lines = self.do_token_transforms(lines) |
|
699 | 699 | except SyntaxError: |
|
700 | 700 | return 'invalid', None |
|
701 | 701 | |
|
702 | 702 | tokens_by_line = make_tokens_by_line(lines) |
|
703 | 703 | |
|
704 | 704 | # Bail if we got one line and there are more closing parentheses than |
|
705 | 705 | # the opening ones |
|
706 | 706 | if ( |
|
707 | 707 | len(lines) == 1 |
|
708 | 708 | and tokens_by_line |
|
709 | 709 | and has_sunken_brackets(tokens_by_line[0]) |
|
710 | 710 | ): |
|
711 | 711 | return "invalid", None |
|
712 | 712 | |
|
713 | 713 | if not tokens_by_line: |
|
714 | 714 | return 'incomplete', find_last_indent(lines) |
|
715 | 715 | |
|
716 | 716 | if tokens_by_line[-1][-1].type != tokenize.ENDMARKER: |
|
717 | 717 | # We're in a multiline string or expression |
|
718 | 718 | return 'incomplete', find_last_indent(lines) |
|
719 | 719 | |
|
720 | 720 | newline_types = {tokenize.NEWLINE, tokenize.COMMENT, tokenize.ENDMARKER} # type: ignore |
|
721 | 721 | |
|
722 | 722 | # Pop the last line which only contains DEDENTs and ENDMARKER |
|
723 | 723 | last_token_line = None |
|
724 | 724 | if {t.type for t in tokens_by_line[-1]} in [ |
|
725 | 725 | {tokenize.DEDENT, tokenize.ENDMARKER}, |
|
726 | 726 | {tokenize.ENDMARKER} |
|
727 | 727 | ] and len(tokens_by_line) > 1: |
|
728 | 728 | last_token_line = tokens_by_line.pop() |
|
729 | 729 | |
|
730 | 730 | while tokens_by_line[-1] and tokens_by_line[-1][-1].type in newline_types: |
|
731 | 731 | tokens_by_line[-1].pop() |
|
732 | 732 | |
|
733 | 733 | if not tokens_by_line[-1]: |
|
734 | 734 | return 'incomplete', find_last_indent(lines) |
|
735 | 735 | |
|
736 | 736 | if tokens_by_line[-1][-1].string == ':': |
|
737 | 737 | # The last line starts a block (e.g. 'if foo:') |
|
738 | 738 | ix = 0 |
|
739 | 739 | while tokens_by_line[-1][ix].type in {tokenize.INDENT, tokenize.DEDENT}: |
|
740 | 740 | ix += 1 |
|
741 | 741 | |
|
742 | 742 | indent = tokens_by_line[-1][ix].start[1] |
|
743 | 743 | return 'incomplete', indent + 4 |
|
744 | 744 | |
|
745 | 745 | if tokens_by_line[-1][0].line.endswith('\\'): |
|
746 | 746 | return 'incomplete', None |
|
747 | 747 | |
|
748 | 748 | # At this point, our checks think the code is complete (or invalid). |
|
749 | 749 | # We'll use codeop.compile_command to check this with the real parser |
|
750 | 750 | try: |
|
751 | 751 | with warnings.catch_warnings(): |
|
752 | 752 | warnings.simplefilter('error', SyntaxWarning) |
|
753 | 753 | res = compile_command(''.join(lines), symbol='exec') |
|
754 | 754 | except (SyntaxError, OverflowError, ValueError, TypeError, |
|
755 | 755 | MemoryError, SyntaxWarning): |
|
756 | 756 | return 'invalid', None |
|
757 | 757 | else: |
|
758 | 758 | if res is None: |
|
759 | 759 | return 'incomplete', find_last_indent(lines) |
|
760 | 760 | |
|
761 | 761 | if last_token_line and last_token_line[0].type == tokenize.DEDENT: |
|
762 | 762 | if ends_with_newline: |
|
763 | 763 | return 'complete', None |
|
764 | 764 | return 'incomplete', find_last_indent(lines) |
|
765 | 765 | |
|
766 | 766 | # If there's a blank line at the end, assume we're ready to execute |
|
767 | 767 | if not lines[-1].strip(): |
|
768 | 768 | return 'complete', None |
|
769 | 769 | |
|
770 | 770 | return 'complete', None |
|
771 | 771 | |
|
772 | 772 | |
|
773 | 773 | def find_last_indent(lines): |
|
774 | 774 | m = _indent_re.match(lines[-1]) |
|
775 | 775 | if not m: |
|
776 | 776 | return 0 |
|
777 | 777 | return len(m.group(0).replace('\t', ' '*4)) |
|
778 | 778 | |
|
779 | 779 | |
|
780 | 780 | class MaybeAsyncCompile(Compile): |
|
781 | 781 | def __init__(self, extra_flags=0): |
|
782 | 782 | super().__init__() |
|
783 | 783 | self.flags |= extra_flags |
|
784 | 784 | |
|
785 | __call__ = compile | |
|
785 | def __call__(self, *args, **kwds): | |
|
786 | return compile(*args, **kwds) | |
|
786 | 787 | |
|
787 | 788 | |
|
788 | 789 | class MaybeAsyncCommandCompiler(CommandCompiler): |
|
789 | 790 | def __init__(self, extra_flags=0): |
|
790 | 791 | self.compiler = MaybeAsyncCompile(extra_flags=extra_flags) |
|
791 | 792 | |
|
792 | 793 | |
|
793 | 794 | if (sys.version_info.major, sys.version_info.minor) >= (3, 8): |
|
794 | 795 | _extra_flags = ast.PyCF_ALLOW_TOP_LEVEL_AWAIT |
|
795 | 796 | else: |
|
796 | 797 | _extra_flags = ast.PyCF_ONLY_AST |
|
797 | 798 | |
|
798 | 799 | compile_command = MaybeAsyncCommandCompiler(extra_flags=_extra_flags) |
General Comments 0
You need to be logged in to leave comments.
Login now