]>
Commit | Line | Data |
---|---|---|
f329fa92 | 1 | #!/usr/bin/env python |
2 | ||
06ccae0f | 3 | ## @package thtml2doxy_clang |
4 | # Translates THtml C++ comments to Doxygen using libclang as parser. | |
5 | # | |
6 | # This code relies on Python bindings for libclang: libclang's interface is pretty unstable, and | |
7 | # its Python bindings are unstable as well. | |
8 | # | |
9 | # AST (Abstract Source Tree) traversal is performed entirely using libclang used as a C++ parser, | |
10 | # instead of attempting to write a parser ourselves. | |
11 | # | |
12 | # This code (expecially AST traversal) was inspired by: | |
13 | # | |
14 | # - [Implementing a code generator with libclang](http://szelei.me/code-generator/) | |
15 | # (this refers to API calls used here) | |
16 | # - [Parsing C++ in Python with Clang](http://eli.thegreenplace.net/2011/07/03/parsing-c-in-python-with-clang) | |
17 | # (outdated, API calls described there do not work anymore, but useful to understand some basic | |
18 | # concepts) | |
19 | # | |
20 | # Usage: | |
21 | # | |
22 | # `thtml2doxy_clang [--stdout|-o] [-d] [--debug=DEBUG_LEVEL] file1 [file2 [file3...]]` | |
23 | # | |
24 | # Parameters: | |
25 | # | |
26 | # - `--stdout|-o`: output all on standard output instead of writing files in place | |
27 | # - `-d`: enable debug mode (very verbose output) | |
28 | # - `--debug=DEBUG_LEVEL`: set debug level to one of `DEBUG`, `INFO`, `WARNING`, `ERROR`, | |
29 | # `CRITICAL` | |
30 | # | |
31 | # @author Dario Berzano, CERN | |
32 | # @date 2014-12-05 | |
33 | ||
34 | ||
f329fa92 | 35 | import sys |
36 | import os | |
37 | import re | |
06ccae0f | 38 | import logging |
39 | import getopt | |
ee1793c7 | 40 | import hashlib |
06ccae0f | 41 | import clang.cindex |
42 | ||
43 | ||
44 | ## Brain-dead color output for terminal. | |
45 | class Colt(str): | |
46 | ||
47 | def red(self): | |
48 | return self.color('\033[31m') | |
49 | ||
50 | def green(self): | |
51 | return self.color('\033[32m') | |
52 | ||
53 | def yellow(self): | |
54 | return self.color('\033[33m') | |
55 | ||
56 | def blue(self): | |
57 | return self.color('\033[34m') | |
58 | ||
59 | def magenta(self): | |
60 | return self.color('\033[35m') | |
61 | ||
62 | def cyan(self): | |
63 | return self.color('\033[36m') | |
64 | ||
65 | def color(self, c): | |
66 | return c + self + '\033[m' | |
67 | ||
f329fa92 | 68 | |
06ccae0f | 69 | ## Comment. |
70 | class Comment: | |
71 | ||
72 | def __init__(self, lines, first_line, first_col, last_line, last_col, indent, func): | |
3896b0ea | 73 | assert first_line > 0 and last_line >= first_line, 'Wrong line numbers' |
06ccae0f | 74 | self.lines = lines |
75 | self.first_line = first_line | |
76 | self.first_col = first_col | |
77 | self.last_line = last_line | |
78 | self.last_col = last_col | |
79 | self.indent = indent | |
80 | self.func = func | |
81 | ||
82 | def has_comment(self, line): | |
83 | return line >= self.first_line and line <= self.last_line | |
84 | ||
85 | def __str__(self): | |
86 | return "<Comment for %s: [%d,%d:%d,%d] %s>" % (self.func, self.first_line, self.first_col, self.last_line, self.last_col, self.lines) | |
87 | ||
88 | ||
89 | ## A data member comment. | |
90 | class MemberComment: | |
91 | ||
92 | def __init__(self, text, is_transient, array_size, first_line, first_col, func): | |
3896b0ea | 93 | assert first_line > 0, 'Wrong line number' |
06ccae0f | 94 | self.lines = [ text ] |
95 | self.is_transient = is_transient | |
96 | self.array_size = array_size | |
97 | self.first_line = first_line | |
98 | self.first_col = first_col | |
99 | self.func = func | |
100 | ||
101 | def has_comment(self, line): | |
102 | return line == self.first_line | |
103 | ||
104 | def __str__(self): | |
105 | ||
106 | if self.is_transient: | |
107 | tt = '!transient! ' | |
108 | else: | |
109 | tt = '' | |
110 | ||
111 | if self.array_size is not None: | |
112 | ars = '[%s] ' % self.array_size | |
113 | else: | |
114 | ars = '' | |
115 | ||
116 | return "<MemberComment for %s: [%d,%d] %s%s%s>" % (self.func, self.first_line, self.first_col, tt, ars, self.lines[0]) | |
117 | ||
118 | ||
119 | ## A dummy comment that removes comment lines. | |
120 | class RemoveComment(Comment): | |
121 | ||
122 | def __init__(self, first_line, last_line): | |
3896b0ea | 123 | assert first_line > 0 and last_line >= first_line, 'Wrong line numbers' |
06ccae0f | 124 | self.first_line = first_line |
125 | self.last_line = last_line | |
126 | self.func = '<remove>' | |
127 | ||
128 | def __str__(self): | |
129 | return "<RemoveComment: [%d,%d]>" % (self.first_line, self.last_line) | |
130 | ||
131 | ||
132 | ## Parses method comments. | |
f329fa92 | 133 | # |
06ccae0f | 134 | # @param cursor Current libclang parser cursor |
135 | # @param comments Array of comments: new ones will be appended there | |
136 | def comment_method(cursor, comments): | |
137 | ||
138 | # we are looking for the following structure: method -> compound statement -> comment, i.e. we | |
139 | # need to extract the first comment in the compound statement composing the method | |
140 | ||
141 | in_compound_stmt = False | |
142 | expect_comment = False | |
143 | emit_comment = False | |
144 | ||
145 | comment = [] | |
146 | comment_function = cursor.spelling or cursor.displayname | |
147 | comment_line_start = -1 | |
148 | comment_line_end = -1 | |
149 | comment_col_start = -1 | |
150 | comment_col_end = -1 | |
151 | comment_indent = -1 | |
152 | ||
153 | for token in cursor.get_tokens(): | |
154 | ||
155 | if token.cursor.kind == clang.cindex.CursorKind.COMPOUND_STMT: | |
156 | if not in_compound_stmt: | |
157 | in_compound_stmt = True | |
158 | expect_comment = True | |
159 | comment_line_end = -1 | |
160 | else: | |
161 | if in_compound_stmt: | |
162 | in_compound_stmt = False | |
163 | emit_comment = True | |
164 | ||
165 | # tkind = str(token.kind)[str(token.kind).index('.')+1:] | |
166 | # ckind = str(token.cursor.kind)[str(token.cursor.kind).index('.')+1:] | |
167 | ||
168 | if in_compound_stmt: | |
169 | ||
170 | if expect_comment: | |
171 | ||
172 | extent = token.extent | |
173 | line_start = extent.start.line | |
174 | line_end = extent.end.line | |
175 | ||
176 | if token.kind == clang.cindex.TokenKind.PUNCTUATION and token.spelling == '{': | |
177 | pass | |
178 | ||
179 | elif token.kind == clang.cindex.TokenKind.COMMENT and (comment_line_end == -1 or (line_start == comment_line_end+1 and line_end-line_start == 0)): | |
180 | comment_line_end = line_end | |
181 | comment_col_end = extent.end.column | |
182 | ||
183 | if comment_indent == -1 or (extent.start.column-1) < comment_indent: | |
184 | comment_indent = extent.start.column-1 | |
185 | ||
186 | if comment_line_start == -1: | |
187 | comment_line_start = line_start | |
188 | comment_col_start = extent.start.column | |
189 | comment.extend( token.spelling.split('\n') ) | |
190 | ||
191 | # multiline comments are parsed in one go, therefore don't expect subsequent comments | |
192 | if line_end - line_start > 0: | |
193 | emit_comment = True | |
194 | expect_comment = False | |
195 | ||
196 | else: | |
197 | emit_comment = True | |
198 | expect_comment = False | |
199 | ||
200 | if emit_comment: | |
201 | ||
6f0e3bf3 | 202 | if comment_line_start > 0: |
06ccae0f | 203 | |
ee1793c7 | 204 | comment = refactor_comment( comment, infilename=str(cursor.location.file) ) |
6f0e3bf3 | 205 | |
206 | if len(comment) > 0: | |
207 | logging.debug("Comment found for function %s" % Colt(comment_function).magenta()) | |
208 | comments.append( Comment(comment, comment_line_start, comment_col_start, comment_line_end, comment_col_end, comment_indent, comment_function) ) | |
209 | else: | |
5dccb084 | 210 | logging.debug('Empty comment found for function %s: collapsing' % Colt(comment_function).magenta()) |
211 | comments.append( Comment([''], comment_line_start, comment_col_start, comment_line_end, comment_col_end, comment_indent, comment_function) ) | |
212 | #comments.append(RemoveComment(comment_line_start, comment_line_end)) | |
6f0e3bf3 | 213 | |
214 | else: | |
215 | logging.warning('No comment found for function %s' % Colt(comment_function).magenta()) | |
06ccae0f | 216 | |
217 | comment = [] | |
218 | comment_line_start = -1 | |
219 | comment_line_end = -1 | |
220 | comment_col_start = -1 | |
221 | comment_col_end = -1 | |
222 | comment_indent = -1 | |
223 | ||
224 | emit_comment = False | |
225 | break | |
226 | ||
227 | ||
228 | ## Parses comments to class data members. | |
229 | # | |
230 | # @param cursor Current libclang parser cursor | |
231 | # @param comments Array of comments: new ones will be appended there | |
232 | def comment_datamember(cursor, comments): | |
233 | ||
234 | # Note: libclang 3.5 seems to have problems parsing a certain type of FIELD_DECL, so we revert | |
235 | # to a partial manual parsing. When parsing fails, the cursor's "extent" is not set properly, | |
236 | # returning a line range 0-0. We therefore make the not-so-absurd assumption that the datamember | |
237 | # definition is fully on one line, and we take the line number from cursor.location. | |
238 | ||
239 | line_num = cursor.location.line | |
240 | raw = None | |
241 | prev = None | |
242 | found = False | |
243 | ||
244 | # Huge overkill | |
245 | with open(str(cursor.location.file)) as fp: | |
246 | cur_line = 0 | |
247 | for raw in fp: | |
248 | cur_line = cur_line + 1 | |
249 | if cur_line == line_num: | |
250 | found = True | |
251 | break | |
252 | prev = raw | |
253 | ||
254 | assert found, 'A line that should exist was not found in file' % cursor.location.file | |
255 | ||
256 | recomm = r'(//(!)|///?)(\[(.*?)\])?<?\s*(.*?)\s*$' | |
257 | recomm_doxyary = r'^\s*///\s*(.*?)\s*$' | |
258 | ||
259 | mcomm = re.search(recomm, raw) | |
260 | if mcomm: | |
54203c62 | 261 | # If it does not match, we do not have a comment |
06ccae0f | 262 | member_name = cursor.spelling; |
263 | is_transient = mcomm.group(2) is not None | |
264 | array_size = mcomm.group(4) | |
265 | text = mcomm.group(5) | |
266 | ||
267 | col_num = mcomm.start()+1; | |
268 | ||
269 | if array_size is not None and prev is not None: | |
270 | # ROOT arrays with comments already converted to Doxygen have the member description on the | |
271 | # previous line | |
272 | mcomm_doxyary = re.search(recomm_doxyary, prev) | |
273 | if mcomm_doxyary: | |
274 | text = mcomm_doxyary.group(1) | |
275 | comments.append(RemoveComment(line_num-1, line_num-1)) | |
276 | ||
277 | logging.debug('Comment found for member %s' % Colt(member_name).magenta()) | |
278 | ||
279 | comments.append( MemberComment( | |
280 | text, | |
281 | is_transient, | |
282 | array_size, | |
283 | line_num, | |
284 | col_num, | |
285 | member_name )) | |
286 | ||
06ccae0f | 287 | |
288 | ## Parses class description (beginning of file). | |
289 | # | |
290 | # The clang parser does not work in this case so we do it manually, but it is very simple: we keep | |
291 | # the first consecutive sequence of single-line comments (//) we find - provided that it occurs | |
292 | # before any other comment found so far in the file (the comments array is inspected to ensure | |
293 | # this). | |
f329fa92 | 294 | # |
06ccae0f | 295 | # Multi-line comments (/* ... */) are not considered as they are commonly used to display |
296 | # copyright notice. | |
f329fa92 | 297 | # |
06ccae0f | 298 | # @param filename Name of the current file |
299 | # @param comments Array of comments: new ones will be appended there | |
300 | def comment_classdesc(filename, comments): | |
301 | ||
302 | recomm = r'^\s*///?(\s*.*?)\s*/*\s*$' | |
303 | ||
304 | reclass_doxy = r'(?i)^\s*\\class:?\s*(.*?)\s*$' | |
305 | class_name_doxy = None | |
306 | ||
307 | reauthor = r'(?i)^\s*\\?authors?:?\s*(.*?)\s*(,?\s*([0-9./-]+))?\s*$' | |
308 | redate = r'(?i)^\s*\\?date:?\s*([0-9./-]+)\s*$' | |
309 | author = None | |
310 | date = None | |
311 | ||
312 | comment_lines = [] | |
313 | ||
314 | start_line = -1 | |
315 | end_line = -1 | |
316 | ||
317 | line_num = 0 | |
318 | ||
319 | with open(filename, 'r') as fp: | |
320 | ||
321 | for raw in fp: | |
322 | ||
323 | line_num = line_num + 1 | |
324 | ||
424eef90 | 325 | if raw.strip() == '' and start_line > 0: |
06ccae0f | 326 | # Skip empty lines |
06ccae0f | 327 | continue |
328 | ||
329 | stripped = strip_html(raw) | |
330 | mcomm = re.search(recomm, stripped) | |
331 | if mcomm: | |
332 | ||
424eef90 | 333 | if start_line == -1: |
06ccae0f | 334 | |
335 | # First line. Check that we do not overlap with other comments | |
336 | comment_overlaps = False | |
337 | for c in comments: | |
338 | if c.has_comment(line_num): | |
339 | comment_overlaps = True | |
340 | break | |
341 | ||
342 | if comment_overlaps: | |
343 | # No need to look for other comments | |
344 | break | |
345 | ||
346 | start_line = line_num | |
347 | ||
dde83b2b | 348 | end_line = line_num |
06ccae0f | 349 | append = True |
350 | ||
351 | mclass_doxy = re.search(reclass_doxy, mcomm.group(1)) | |
352 | if mclass_doxy: | |
353 | class_name_doxy = mclass_doxy.group(1) | |
354 | append = False | |
355 | else: | |
356 | mauthor = re.search(reauthor, mcomm.group(1)) | |
357 | if mauthor: | |
358 | author = mauthor.group(1) | |
359 | if date is None: | |
360 | # Date specified in the standalone \date field has priority | |
9b8614a7 | 361 | date = mauthor.group(3) |
06ccae0f | 362 | append = False |
363 | else: | |
364 | mdate = re.search(redate, mcomm.group(1)) | |
365 | if mdate: | |
366 | date = mdate.group(1) | |
367 | append = False | |
368 | ||
369 | if append: | |
370 | comment_lines.append( mcomm.group(1) ) | |
371 | ||
372 | else: | |
424eef90 | 373 | if start_line > 0: |
06ccae0f | 374 | break |
375 | ||
376 | if class_name_doxy is None: | |
377 | ||
378 | # No \class specified: guess it from file name | |
379 | reclass = r'^(.*/)?(.*?)(\..*)?$' | |
380 | mclass = re.search( reclass, filename ) | |
381 | if mclass: | |
382 | class_name_doxy = mclass.group(2) | |
383 | else: | |
384 | assert False, 'Regexp unable to extract classname from file' | |
385 | ||
424eef90 | 386 | if start_line > 0: |
06ccae0f | 387 | |
424eef90 | 388 | # Prepend \class specifier (and an empty line) |
389 | comment_lines[:0] = [ '\\class ' + class_name_doxy ] | |
06ccae0f | 390 | |
424eef90 | 391 | # Append author and date if they exist |
392 | comment_lines.append('') | |
06ccae0f | 393 | |
424eef90 | 394 | if author is not None: |
395 | comment_lines.append( '\\author ' + author ) | |
06ccae0f | 396 | |
424eef90 | 397 | if date is not None: |
398 | comment_lines.append( '\\date ' + date ) | |
399 | ||
ee1793c7 | 400 | comment_lines = refactor_comment(comment_lines, do_strip_html=False, infilename=filename) |
424eef90 | 401 | logging.debug('Comment found for class %s' % Colt(class_name_doxy).magenta()) |
402 | comments.append(Comment( | |
403 | comment_lines, | |
404 | start_line, 1, end_line, 1, | |
405 | 0, class_name_doxy | |
406 | )) | |
407 | ||
408 | else: | |
409 | ||
410 | logging.warning('No comment found for class %s' % Colt(class_name_doxy).magenta()) | |
06ccae0f | 411 | |
412 | ||
413 | ## Traverse the AST recursively starting from the current cursor. | |
414 | # | |
415 | # @param cursor A Clang parser cursor | |
416 | # @param filename Name of the current file | |
417 | # @param comments Array of comments: new ones will be appended there | |
418 | # @param recursion Current recursion depth | |
419 | def traverse_ast(cursor, filename, comments, recursion=0): | |
420 | ||
421 | # libclang traverses included files as well: we do not want this behavior | |
422 | if cursor.location.file is not None and str(cursor.location.file) != filename: | |
423 | logging.debug("Skipping processing of included %s" % cursor.location.file) | |
424 | return | |
425 | ||
426 | text = cursor.spelling or cursor.displayname | |
427 | kind = str(cursor.kind)[str(cursor.kind).index('.')+1:] | |
428 | ||
429 | indent = '' | |
430 | for i in range(0, recursion): | |
431 | indent = indent + ' ' | |
432 | ||
433 | if cursor.kind == clang.cindex.CursorKind.CXX_METHOD or cursor.kind == clang.cindex.CursorKind.CONSTRUCTOR or cursor.kind == clang.cindex.CursorKind.DESTRUCTOR: | |
434 | ||
435 | # cursor ran into a C++ method | |
436 | logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) ) | |
437 | comment_method(cursor, comments) | |
438 | ||
439 | elif cursor.kind == clang.cindex.CursorKind.FIELD_DECL: | |
440 | ||
441 | # cursor ran into a data member declaration | |
442 | logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) ) | |
443 | comment_datamember(cursor, comments) | |
444 | ||
445 | else: | |
446 | ||
447 | logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, kind, text) ) | |
448 | ||
449 | for child_cursor in cursor.get_children(): | |
450 | traverse_ast(child_cursor, filename, comments, recursion+1) | |
451 | ||
452 | if recursion == 0: | |
453 | comment_classdesc(filename, comments) | |
454 | ||
455 | ||
456 | ## Strip some HTML tags from the given string. Returns clean string. | |
457 | # | |
458 | # @param s Input string | |
459 | def strip_html(s): | |
798167cb | 460 | rehtml = r'(?i)</?(P|BR)/?>' |
06ccae0f | 461 | return re.sub(rehtml, '', s) |
462 | ||
463 | ||
464 | ## Remove garbage from comments and convert special tags from THtml to Doxygen. | |
465 | # | |
466 | # @param comment An array containing the lines of the original comment | |
ee1793c7 | 467 | def refactor_comment(comment, do_strip_html=True, infilename=None): |
06ccae0f | 468 | |
469 | recomm = r'^(/{2,}|/\*)? ?(\s*.*?)\s*((/{2,})?\s*|\*/)$' | |
470 | regarbage = r'^(?i)\s*([\s*=-_#]+|(Begin|End)_Html)\s*$' | |
471 | ||
03a6b7aa | 472 | # Support for LaTeX blocks spanning on multiple lines |
473 | relatex = r'(?i)^((.*?)\s+)?(BEGIN|END)_LATEX([.,;:\s]+.*)?$' | |
474 | in_latex = False | |
475 | latex_block = False | |
476 | ||
477 | # Support for LaTeX blocks on a single line | |
9db428b7 | 478 | reinline_latex = r'(?i)(.*)BEGIN_LATEX\s+(.*?)\s+END_LATEX(.*)$' |
479 | ||
35b193b4 | 480 | # Match <pre> (to turn it into the ~~~ Markdown syntax) |
481 | reblock = r'(?i)^(\s*)</?PRE>\s*$' | |
482 | ||
ee1793c7 | 483 | # Macro blocks for pictures generation |
484 | in_macro = False | |
485 | current_macro = [] | |
486 | remacro = r'(?i)^\s*(BEGIN|END)_MACRO(\((.*?)\))?\s*$' | |
487 | ||
06ccae0f | 488 | new_comment = [] |
489 | insert_blank = False | |
490 | wait_first_non_blank = True | |
491 | for line_comment in comment: | |
492 | ||
ee1793c7 | 493 | # Check if we are in a macro block |
494 | mmacro = re.search(remacro, line_comment) | |
495 | if mmacro: | |
496 | if in_macro: | |
497 | in_macro = False | |
498 | ||
499 | # Dump macro | |
500 | outimg = write_macro(infilename, current_macro) + '.png' | |
501 | current_macro = [] | |
502 | ||
503 | # Insert image | |
504 | new_comment.append( '![Picture from ROOT macro](%s)' % (outimg) ) | |
505 | ||
506 | logging.debug( 'Found macro for generating image %s' % Colt(outimg).magenta() ) | |
507 | ||
508 | else: | |
509 | in_macro = True | |
510 | ||
511 | continue | |
512 | elif in_macro: | |
513 | current_macro.append( line_comment ) | |
514 | continue | |
515 | ||
06ccae0f | 516 | # Strip some HTML tags |
517 | if do_strip_html: | |
518 | line_comment = strip_html(line_comment) | |
519 | ||
520 | mcomm = re.search( recomm, line_comment ) | |
521 | if mcomm: | |
522 | new_line_comment = mcomm.group(2) | |
523 | mgarbage = re.search( regarbage, new_line_comment ) | |
524 | ||
525 | if new_line_comment == '' or mgarbage is not None: | |
526 | insert_blank = True | |
527 | else: | |
528 | if insert_blank and not wait_first_non_blank: | |
529 | new_comment.append('') | |
530 | insert_blank = False | |
531 | wait_first_non_blank = False | |
9db428b7 | 532 | |
533 | # Postprocessing: LaTeX formulas in ROOT format | |
534 | # Marked by BEGIN_LATEX ... END_LATEX and they use # in place of \ | |
535 | # There can be several ROOT LaTeX forumlas per line | |
536 | while True: | |
537 | minline_latex = re.search( reinline_latex, new_line_comment ) | |
538 | if minline_latex: | |
539 | new_line_comment = '%s\\f$%s\\f$%s' % \ | |
540 | ( minline_latex.group(1), minline_latex.group(2).replace('#', '\\'), | |
541 | minline_latex.group(3) ) | |
542 | else: | |
543 | break | |
544 | ||
03a6b7aa | 545 | # ROOT LaTeX: do we have a Begin/End_LaTeX block? |
546 | # Note: the presence of LaTeX "closures" does not exclude the possibility to have a begin | |
547 | # block here left without a corresponding ending block | |
548 | mlatex = re.search( relatex, new_line_comment ) | |
549 | if mlatex: | |
550 | ||
551 | # before and after parts have been already stripped | |
552 | l_before = mlatex.group(2) | |
553 | l_after = mlatex.group(4) | |
554 | is_begin = mlatex.group(3).upper() == 'BEGIN' # if not, END | |
555 | ||
556 | if l_before is None: | |
557 | l_before = '' | |
558 | if l_after is None: | |
559 | l_after = '' | |
560 | ||
561 | if is_begin: | |
562 | ||
563 | # Begin of LaTeX part | |
564 | ||
565 | in_latex = True | |
566 | if l_before == '' and l_after == '': | |
567 | ||
568 | # Opening tag alone: mark the beginning of a block: \f[ ... \f] | |
569 | latex_block = True | |
570 | new_comment.append( '\\f[' ) | |
571 | ||
572 | else: | |
573 | # Mark the beginning of inline: \f$ ... \f$ | |
574 | latex_block = False | |
575 | new_comment.append( | |
576 | '%s \\f$%s' % ( l_before, l_after.replace('#', '\\') ) | |
577 | ) | |
578 | ||
579 | else: | |
580 | ||
581 | # End of LaTeX part | |
582 | in_latex = False | |
583 | ||
584 | if latex_block: | |
585 | ||
586 | # Closing a LaTeX block | |
587 | if l_before != '': | |
588 | new_comment.append( l_before.replace('#', '\\') ) | |
589 | new_comment.append( '\\f]' ) | |
590 | if l_after != '': | |
591 | new_comment.append( l_after ) | |
592 | ||
593 | else: | |
594 | ||
595 | # Closing a LaTeX inline | |
596 | new_comment.append( | |
597 | '%s\\f$%s' % ( l_before.replace('#', '\\'), l_after ) | |
598 | ) | |
599 | ||
600 | # Prevent appending lines (we have already done that) | |
601 | new_line_comment = None | |
602 | ||
35b193b4 | 603 | # If we are not in a LaTeX block, look for <pre> tags and transform them into Doxygen code |
604 | # blocks (using ~~~ ... ~~~). Only <pre> tags on a single line are supported | |
605 | if new_line_comment is not None and not in_latex: | |
606 | ||
607 | mblock = re.search( reblock, new_line_comment ) | |
608 | if mblock: | |
609 | new_comment.append( mblock.group(1)+'~~~' ) | |
610 | new_line_comment = None | |
611 | ||
03a6b7aa | 612 | if new_line_comment is not None: |
613 | if in_latex: | |
614 | new_line_comment = new_line_comment.replace('#', '\\') | |
615 | new_comment.append( new_line_comment ) | |
06ccae0f | 616 | |
617 | else: | |
618 | assert False, 'Comment regexp does not match' | |
619 | ||
620 | return new_comment | |
621 | ||
622 | ||
ee1793c7 | 623 | ## Dumps an image-generating macro to the correct place. Returns a string with the image path, |
624 | # without the extension. | |
625 | # | |
626 | # @param infilename File name of the source file | |
627 | # @param macro_lines Array of macro lines | |
628 | def write_macro(infilename, macro_lines): | |
629 | ||
630 | # Calculate hash | |
631 | digh = hashlib.sha1() | |
632 | for l in macro_lines: | |
633 | digh.update(l) | |
634 | digh.update('\n') | |
635 | short_digest = digh.hexdigest()[0:7] | |
636 | ||
637 | outdir = '%s/imgdoc' % os.path.dirname(infilename) | |
638 | outprefix = '%s/%s_%s' % ( | |
639 | outdir, | |
640 | os.path.basename(infilename).replace('.', '_'), | |
641 | short_digest | |
642 | ) | |
643 | outmacro = '%s.C' % outprefix | |
644 | ||
645 | # Make directory | |
646 | if not os.path.isdir(outdir): | |
647 | # do not catch: let everything die on error | |
648 | logging.debug('Creating directory %s' % Colt(outdir).magenta()) | |
649 | os.mkdir(outdir) | |
650 | ||
651 | # Create file (do not catch errors either) | |
652 | with open(outmacro, 'w') as omfp: | |
653 | logging.debug('Writing macro %s' % Colt(outmacro).magenta()) | |
654 | for l in macro_lines: | |
655 | omfp.write(l) | |
656 | omfp.write('\n') | |
657 | ||
658 | return outprefix | |
659 | ||
660 | ||
06ccae0f | 661 | ## Rewrites all comments from the given file handler. |
662 | # | |
663 | # @param fhin The file handler to read from | |
664 | # @param fhout The file handler to write to | |
665 | # @param comments Array of comments | |
666 | def rewrite_comments(fhin, fhout, comments): | |
667 | ||
668 | line_num = 0 | |
669 | in_comment = False | |
670 | skip_empty = False | |
671 | comm = None | |
672 | prev_comm = None | |
673 | ||
674 | rindent = r'^(\s*)' | |
675 | ||
676 | for line in fhin: | |
677 | ||
678 | line_num = line_num + 1 | |
679 | ||
680 | # Find current comment | |
681 | prev_comm = comm | |
682 | comm = None | |
683 | for c in comments: | |
684 | if c.has_comment(line_num): | |
685 | comm = c | |
686 | ||
687 | if comm: | |
688 | ||
689 | if isinstance(comm, MemberComment): | |
690 | non_comment = line[ 0:comm.first_col-1 ] | |
691 | ||
692 | if comm.array_size is not None: | |
693 | ||
694 | mindent = re.search(rindent, line) | |
695 | if comm.is_transient: | |
696 | tt = '!' | |
697 | else: | |
698 | tt = '' | |
699 | ||
700 | # Special case: we need multiple lines not to confuse ROOT's C++ parser | |
701 | fhout.write('%s/// %s\n%s//%s[%s]\n' % ( | |
702 | mindent.group(1), | |
703 | comm.lines[0], | |
704 | non_comment, | |
705 | tt, | |
706 | comm.array_size | |
707 | )) | |
708 | ||
709 | else: | |
710 | ||
711 | if comm.is_transient: | |
712 | tt = '!' | |
713 | else: | |
714 | tt = '/' | |
715 | ||
716 | fhout.write('%s//%s< %s\n' % ( | |
717 | non_comment, | |
718 | tt, | |
719 | comm.lines[0] | |
720 | )) | |
721 | ||
722 | elif isinstance(comm, RemoveComment): | |
723 | # Do nothing: just skip line | |
724 | pass | |
725 | ||
726 | elif prev_comm is None: | |
727 | # Beginning of a new comment block of type Comment | |
728 | in_comment = True | |
729 | ||
730 | # Extract the non-comment part and print it if it exists | |
731 | non_comment = line[ 0:comm.first_col-1 ].rstrip() | |
732 | if non_comment != '': | |
733 | fhout.write( non_comment + '\n' ) | |
734 | ||
735 | else: | |
736 | ||
737 | if in_comment: | |
738 | ||
739 | # We have just exited a comment block of type Comment | |
740 | in_comment = False | |
741 | ||
742 | # Dump revamped comment, if applicable | |
743 | text_indent = '' | |
744 | for i in range(0,prev_comm.indent): | |
745 | text_indent = text_indent + ' ' | |
746 | ||
747 | for lc in prev_comm.lines: | |
748 | fhout.write( "%s/// %s\n" % (text_indent, lc) ); | |
749 | fhout.write('\n') | |
750 | skip_empty = True | |
751 | ||
752 | line_out = line.rstrip('\n') | |
753 | if skip_empty: | |
754 | skip_empty = False | |
755 | if line_out.strip() != '': | |
756 | fhout.write( line_out + '\n' ) | |
757 | else: | |
758 | fhout.write( line_out + '\n' ) | |
759 | ||
f329fa92 | 760 | |
761 | ## The main function. | |
762 | # | |
06ccae0f | 763 | # Return value is the executable's return value. |
f329fa92 | 764 | def main(argv): |
765 | ||
06ccae0f | 766 | # Setup logging on stderr |
767 | log_level = logging.INFO | |
768 | logging.basicConfig( | |
769 | level=log_level, | |
770 | format='%(levelname)-8s %(funcName)-20s %(message)s', | |
771 | stream=sys.stderr | |
772 | ) | |
f329fa92 | 773 | |
06ccae0f | 774 | # Parse command-line options |
775 | output_on_stdout = False | |
7afec95e | 776 | include_flags = [] |
06ccae0f | 777 | try: |
7afec95e | 778 | opts, args = getopt.getopt( argv, 'odI:', [ 'debug=', 'stdout' ] ) |
06ccae0f | 779 | for o, a in opts: |
780 | if o == '--debug': | |
781 | log_level = getattr( logging, a.upper(), None ) | |
782 | if not isinstance(log_level, int): | |
783 | raise getopt.GetoptError('log level must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL') | |
784 | elif o == '-d': | |
785 | log_level = logging.DEBUG | |
786 | elif o == '-o' or o == '--stdout': | |
06ccae0f | 787 | output_on_stdout = True |
7afec95e | 788 | elif o == '-I': |
789 | if os.path.isdir(a): | |
790 | include_flags.extend( [ '-I', a ] ) | |
791 | else: | |
792 | logging.fatal('Include directory not found: %s' % Colt(a).magenta()) | |
793 | return 2 | |
06ccae0f | 794 | else: |
795 | assert False, 'Unhandled argument' | |
796 | except getopt.GetoptError as e: | |
797 | logging.fatal('Invalid arguments: %s' % e) | |
798 | return 1 | |
f329fa92 | 799 | |
06ccae0f | 800 | logging.getLogger('').setLevel(log_level) |
f329fa92 | 801 | |
06ccae0f | 802 | # Attempt to load libclang from a list of known locations |
803 | libclang_locations = [ | |
804 | '/usr/lib/llvm-3.5/lib/libclang.so.1', | |
805 | '/usr/lib/libclang.so', | |
806 | '/Library/Developer/CommandLineTools/usr/lib/libclang.dylib' | |
807 | ] | |
808 | libclang_found = False | |
f329fa92 | 809 | |
06ccae0f | 810 | for lib in libclang_locations: |
811 | if os.path.isfile(lib): | |
812 | clang.cindex.Config.set_library_file(lib) | |
813 | libclang_found = True | |
814 | break | |
f329fa92 | 815 | |
06ccae0f | 816 | if not libclang_found: |
817 | logging.fatal('Cannot find libclang') | |
818 | return 1 | |
819 | ||
820 | # Loop over all files | |
821 | for fn in args: | |
822 | ||
823 | logging.info('Input file: %s' % Colt(fn).magenta()) | |
824 | index = clang.cindex.Index.create() | |
7afec95e | 825 | clang_args = [ '-x', 'c++' ] |
826 | clang_args.extend( include_flags ) | |
827 | translation_unit = index.parse(fn, args=clang_args) | |
06ccae0f | 828 | |
829 | comments = [] | |
830 | traverse_ast( translation_unit.cursor, fn, comments ) | |
831 | for c in comments: | |
832 | ||
833 | logging.debug("Comment found for entity %s:" % Colt(c.func).magenta()) | |
f329fa92 | 834 | |
06ccae0f | 835 | if isinstance(c, MemberComment): |
836 | ||
837 | if c.is_transient: | |
838 | transient_text = Colt('transient ').yellow() | |
839 | else: | |
840 | transient_text = '' | |
841 | ||
842 | if c.array_size is not None: | |
843 | array_text = Colt('arraysize=%s ' % c.array_size).yellow() | |
844 | else: | |
845 | array_text = '' | |
846 | ||
847 | logging.debug( | |
848 | "%s %s%s{%s}" % ( \ | |
849 | Colt("[%d,%d]" % (c.first_line, c.first_col)).green(), | |
850 | transient_text, | |
851 | array_text, | |
852 | Colt(c.lines[0]).cyan() | |
853 | )) | |
854 | ||
855 | elif isinstance(c, RemoveComment): | |
856 | ||
857 | logging.debug( Colt('[%d,%d]' % (c.first_line, c.last_line)).green() ) | |
858 | ||
859 | else: | |
860 | for l in c.lines: | |
861 | logging.debug( | |
862 | Colt("[%d,%d:%d,%d] " % (c.first_line, c.first_col, c.last_line, c.last_col)).green() + | |
863 | "{%s}" % Colt(l).cyan() | |
864 | ) | |
f329fa92 | 865 | |
866 | try: | |
06ccae0f | 867 | |
868 | if output_on_stdout: | |
869 | with open(fn, 'r') as fhin: | |
870 | rewrite_comments( fhin, sys.stdout, comments ) | |
871 | else: | |
872 | fn_back = fn + '.thtml2doxy_backup' | |
873 | os.rename( fn, fn_back ) | |
874 | ||
875 | with open(fn_back, 'r') as fhin, open(fn, 'w') as fhout: | |
876 | rewrite_comments( fhin, fhout, comments ) | |
877 | ||
878 | os.remove( fn_back ) | |
879 | logging.info("File %s converted to Doxygen: check differences before committing!" % Colt(fn).magenta()) | |
880 | except (IOError,OSError) as e: | |
881 | logging.error('File operation failed: %s' % e) | |
f329fa92 | 882 | |
883 | return 0 | |
884 | ||
06ccae0f | 885 | |
f329fa92 | 886 | if __name__ == '__main__': |
06ccae0f | 887 | sys.exit( main( sys.argv[1:] ) ) |