]>
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17 | # (outdated, API calls described there do not work anymore, but useful to understand some basic | |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 | ||
21689d7a | 92 | def __init__(self, text, comment_flag, array_size, first_line, first_col, func): |
3896b0ea | 93 | assert first_line > 0, 'Wrong line number' |
21689d7a | 94 | assert comment_flag is None or comment_flag == '!' or comment_flag in [ '!', '||', '->' ] |
06ccae0f | 95 | self.lines = [ text ] |
21689d7a | 96 | self.comment_flag = comment_flag |
06ccae0f | 97 | self.array_size = array_size |
98 | self.first_line = first_line | |
99 | self.first_col = first_col | |
100 | self.func = func | |
101 | ||
21689d7a | 102 | def is_transient(self): |
103 | return self.comment_flag == '!' | |
104 | ||
105 | def is_dontsplit(self): | |
106 | return self.comment_flag == '||' | |
107 | ||
108 | def is_ptr(self): | |
109 | return self.comment_flag == '->' | |
110 | ||
06ccae0f | 111 | def has_comment(self, line): |
112 | return line == self.first_line | |
113 | ||
114 | def __str__(self): | |
115 | ||
21689d7a | 116 | if self.is_transient(): |
06ccae0f | 117 | tt = '!transient! ' |
21689d7a | 118 | elif self.is_dontsplit(): |
119 | tt = '!dontsplit! ' | |
120 | elif self.is_ptr(): | |
121 | tt = '!ptr! ' | |
06ccae0f | 122 | else: |
123 | tt = '' | |
124 | ||
125 | if self.array_size is not None: | |
126 | ars = '[%s] ' % self.array_size | |
127 | else: | |
128 | ars = '' | |
129 | ||
130 | return "<MemberComment for %s: [%d,%d] %s%s%s>" % (self.func, self.first_line, self.first_col, tt, ars, self.lines[0]) | |
131 | ||
132 | ||
133 | ## A dummy comment that removes comment lines. | |
134 | class RemoveComment(Comment): | |
135 | ||
136 | def __init__(self, first_line, last_line): | |
3896b0ea | 137 | assert first_line > 0 and last_line >= first_line, 'Wrong line numbers' |
06ccae0f | 138 | self.first_line = first_line |
139 | self.last_line = last_line | |
140 | self.func = '<remove>' | |
141 | ||
142 | def __str__(self): | |
143 | return "<RemoveComment: [%d,%d]>" % (self.first_line, self.last_line) | |
144 | ||
145 | ||
146 | ## Parses method comments. | |
f329fa92 | 147 | # |
06ccae0f | 148 | # @param cursor Current libclang parser cursor |
149 | # @param comments Array of comments: new ones will be appended there | |
150 | def comment_method(cursor, comments): | |
151 | ||
152 | # we are looking for the following structure: method -> compound statement -> comment, i.e. we | |
153 | # need to extract the first comment in the compound statement composing the method | |
154 | ||
155 | in_compound_stmt = False | |
156 | expect_comment = False | |
157 | emit_comment = False | |
158 | ||
159 | comment = [] | |
160 | comment_function = cursor.spelling or cursor.displayname | |
161 | comment_line_start = -1 | |
162 | comment_line_end = -1 | |
163 | comment_col_start = -1 | |
164 | comment_col_end = -1 | |
165 | comment_indent = -1 | |
166 | ||
167 | for token in cursor.get_tokens(): | |
168 | ||
169 | if token.cursor.kind == clang.cindex.CursorKind.COMPOUND_STMT: | |
170 | if not in_compound_stmt: | |
171 | in_compound_stmt = True | |
172 | expect_comment = True | |
173 | comment_line_end = -1 | |
174 | else: | |
175 | if in_compound_stmt: | |
176 | in_compound_stmt = False | |
177 | emit_comment = True | |
178 | ||
179 | # tkind = str(token.kind)[str(token.kind).index('.')+1:] | |
180 | # ckind = str(token.cursor.kind)[str(token.cursor.kind).index('.')+1:] | |
181 | ||
182 | if in_compound_stmt: | |
183 | ||
184 | if expect_comment: | |
185 | ||
186 | extent = token.extent | |
187 | line_start = extent.start.line | |
188 | line_end = extent.end.line | |
189 | ||
190 | if token.kind == clang.cindex.TokenKind.PUNCTUATION and token.spelling == '{': | |
191 | pass | |
192 | ||
193 | 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)): | |
194 | comment_line_end = line_end | |
195 | comment_col_end = extent.end.column | |
196 | ||
197 | if comment_indent == -1 or (extent.start.column-1) < comment_indent: | |
198 | comment_indent = extent.start.column-1 | |
199 | ||
200 | if comment_line_start == -1: | |
201 | comment_line_start = line_start | |
202 | comment_col_start = extent.start.column | |
203 | comment.extend( token.spelling.split('\n') ) | |
204 | ||
205 | # multiline comments are parsed in one go, therefore don't expect subsequent comments | |
206 | if line_end - line_start > 0: | |
207 | emit_comment = True | |
208 | expect_comment = False | |
209 | ||
210 | else: | |
211 | emit_comment = True | |
212 | expect_comment = False | |
213 | ||
214 | if emit_comment: | |
215 | ||
6f0e3bf3 | 216 | if comment_line_start > 0: |
06ccae0f | 217 | |
ee1793c7 | 218 | comment = refactor_comment( comment, infilename=str(cursor.location.file) ) |
6f0e3bf3 | 219 | |
220 | if len(comment) > 0: | |
221 | logging.debug("Comment found for function %s" % Colt(comment_function).magenta()) | |
222 | comments.append( Comment(comment, comment_line_start, comment_col_start, comment_line_end, comment_col_end, comment_indent, comment_function) ) | |
223 | else: | |
5dccb084 | 224 | logging.debug('Empty comment found for function %s: collapsing' % Colt(comment_function).magenta()) |
225 | comments.append( Comment([''], comment_line_start, comment_col_start, comment_line_end, comment_col_end, comment_indent, comment_function) ) | |
226 | #comments.append(RemoveComment(comment_line_start, comment_line_end)) | |
6f0e3bf3 | 227 | |
228 | else: | |
229 | logging.warning('No comment found for function %s' % Colt(comment_function).magenta()) | |
06ccae0f | 230 | |
231 | comment = [] | |
232 | comment_line_start = -1 | |
233 | comment_line_end = -1 | |
234 | comment_col_start = -1 | |
235 | comment_col_end = -1 | |
236 | comment_indent = -1 | |
237 | ||
238 | emit_comment = False | |
239 | break | |
240 | ||
241 | ||
242 | ## Parses comments to class data members. | |
243 | # | |
244 | # @param cursor Current libclang parser cursor | |
245 | # @param comments Array of comments: new ones will be appended there | |
246 | def comment_datamember(cursor, comments): | |
247 | ||
248 | # Note: libclang 3.5 seems to have problems parsing a certain type of FIELD_DECL, so we revert | |
249 | # to a partial manual parsing. When parsing fails, the cursor's "extent" is not set properly, | |
250 | # returning a line range 0-0. We therefore make the not-so-absurd assumption that the datamember | |
251 | # definition is fully on one line, and we take the line number from cursor.location. | |
252 | ||
253 | line_num = cursor.location.line | |
254 | raw = None | |
255 | prev = None | |
256 | found = False | |
257 | ||
21689d7a | 258 | # Huge overkill: current line saved in "raw", previous in "prev" |
06ccae0f | 259 | with open(str(cursor.location.file)) as fp: |
260 | cur_line = 0 | |
261 | for raw in fp: | |
262 | cur_line = cur_line + 1 | |
263 | if cur_line == line_num: | |
264 | found = True | |
265 | break | |
266 | prev = raw | |
267 | ||
268 | assert found, 'A line that should exist was not found in file' % cursor.location.file | |
269 | ||
21689d7a | 270 | recomm = r'(//(!|\|\||->)|///?)(\[([0-9,]+)\])?<?\s*(.*?)\s*$' |
271 | recomm_prevline = r'^\s*///\s*(.*?)\s*$' | |
06ccae0f | 272 | |
273 | mcomm = re.search(recomm, raw) | |
274 | if mcomm: | |
54203c62 | 275 | # If it does not match, we do not have a comment |
06ccae0f | 276 | member_name = cursor.spelling; |
21689d7a | 277 | comment_flag = mcomm.group(2) |
06ccae0f | 278 | array_size = mcomm.group(4) |
279 | text = mcomm.group(5) | |
280 | ||
281 | col_num = mcomm.start()+1; | |
282 | ||
283 | if array_size is not None and prev is not None: | |
284 | # ROOT arrays with comments already converted to Doxygen have the member description on the | |
285 | # previous line | |
21689d7a | 286 | mcomm_prevline = re.search(recomm_prevline, prev) |
287 | if mcomm_prevline: | |
288 | text = mcomm_prevline.group(1) | |
06ccae0f | 289 | comments.append(RemoveComment(line_num-1, line_num-1)) |
290 | ||
291 | logging.debug('Comment found for member %s' % Colt(member_name).magenta()) | |
292 | ||
293 | comments.append( MemberComment( | |
294 | text, | |
21689d7a | 295 | comment_flag, |
06ccae0f | 296 | array_size, |
297 | line_num, | |
298 | col_num, | |
299 | member_name )) | |
300 | ||
06ccae0f | 301 | |
302 | ## Parses class description (beginning of file). | |
303 | # | |
304 | # The clang parser does not work in this case so we do it manually, but it is very simple: we keep | |
305 | # the first consecutive sequence of single-line comments (//) we find - provided that it occurs | |
306 | # before any other comment found so far in the file (the comments array is inspected to ensure | |
307 | # this). | |
f329fa92 | 308 | # |
06ccae0f | 309 | # Multi-line comments (/* ... */) are not considered as they are commonly used to display |
310 | # copyright notice. | |
f329fa92 | 311 | # |
06ccae0f | 312 | # @param filename Name of the current file |
313 | # @param comments Array of comments: new ones will be appended there | |
314 | def comment_classdesc(filename, comments): | |
315 | ||
316 | recomm = r'^\s*///?(\s*.*?)\s*/*\s*$' | |
317 | ||
ccf0fca4 | 318 | reclass_doxy = r'(?i)^\s*\\(class|file):?\s*([^.]*)' |
06ccae0f | 319 | class_name_doxy = None |
320 | ||
321 | reauthor = r'(?i)^\s*\\?authors?:?\s*(.*?)\s*(,?\s*([0-9./-]+))?\s*$' | |
322 | redate = r'(?i)^\s*\\?date:?\s*([0-9./-]+)\s*$' | |
323 | author = None | |
324 | date = None | |
325 | ||
326 | comment_lines = [] | |
327 | ||
328 | start_line = -1 | |
329 | end_line = -1 | |
330 | ||
331 | line_num = 0 | |
332 | ||
ccf0fca4 | 333 | is_macro = filename.endswith('.C') |
334 | ||
06ccae0f | 335 | with open(filename, 'r') as fp: |
336 | ||
337 | for raw in fp: | |
338 | ||
339 | line_num = line_num + 1 | |
340 | ||
424eef90 | 341 | if raw.strip() == '' and start_line > 0: |
06ccae0f | 342 | # Skip empty lines |
06ccae0f | 343 | continue |
344 | ||
345 | stripped = strip_html(raw) | |
346 | mcomm = re.search(recomm, stripped) | |
347 | if mcomm: | |
348 | ||
424eef90 | 349 | if start_line == -1: |
06ccae0f | 350 | |
351 | # First line. Check that we do not overlap with other comments | |
352 | comment_overlaps = False | |
353 | for c in comments: | |
354 | if c.has_comment(line_num): | |
355 | comment_overlaps = True | |
356 | break | |
357 | ||
358 | if comment_overlaps: | |
359 | # No need to look for other comments | |
360 | break | |
361 | ||
362 | start_line = line_num | |
363 | ||
dde83b2b | 364 | end_line = line_num |
06ccae0f | 365 | append = True |
366 | ||
367 | mclass_doxy = re.search(reclass_doxy, mcomm.group(1)) | |
368 | if mclass_doxy: | |
ccf0fca4 | 369 | class_name_doxy = mclass_doxy.group(2) |
06ccae0f | 370 | append = False |
371 | else: | |
372 | mauthor = re.search(reauthor, mcomm.group(1)) | |
373 | if mauthor: | |
374 | author = mauthor.group(1) | |
375 | if date is None: | |
376 | # Date specified in the standalone \date field has priority | |
9b8614a7 | 377 | date = mauthor.group(3) |
06ccae0f | 378 | append = False |
379 | else: | |
380 | mdate = re.search(redate, mcomm.group(1)) | |
381 | if mdate: | |
382 | date = mdate.group(1) | |
383 | append = False | |
384 | ||
385 | if append: | |
386 | comment_lines.append( mcomm.group(1) ) | |
387 | ||
388 | else: | |
424eef90 | 389 | if start_line > 0: |
06ccae0f | 390 | break |
391 | ||
392 | if class_name_doxy is None: | |
393 | ||
394 | # No \class specified: guess it from file name | |
395 | reclass = r'^(.*/)?(.*?)(\..*)?$' | |
396 | mclass = re.search( reclass, filename ) | |
397 | if mclass: | |
398 | class_name_doxy = mclass.group(2) | |
399 | else: | |
400 | assert False, 'Regexp unable to extract classname from file' | |
401 | ||
424eef90 | 402 | if start_line > 0: |
06ccae0f | 403 | |
ccf0fca4 | 404 | # Prepend \class or \file specifier (and an empty line) |
405 | if is_macro: | |
406 | comment_lines[:0] = [ '\\file ' + class_name_doxy + '.C' ] | |
407 | else: | |
408 | comment_lines[:0] = [ '\\class ' + class_name_doxy ] | |
424eef90 | 409 | comment_lines.append('') |
06ccae0f | 410 | |
84039997 | 411 | # Append author and date if they exist |
424eef90 | 412 | if author is not None: |
413 | comment_lines.append( '\\author ' + author ) | |
06ccae0f | 414 | |
424eef90 | 415 | if date is not None: |
416 | comment_lines.append( '\\date ' + date ) | |
417 | ||
ee1793c7 | 418 | comment_lines = refactor_comment(comment_lines, do_strip_html=False, infilename=filename) |
424eef90 | 419 | logging.debug('Comment found for class %s' % Colt(class_name_doxy).magenta()) |
420 | comments.append(Comment( | |
421 | comment_lines, | |
422 | start_line, 1, end_line, 1, | |
423 | 0, class_name_doxy | |
424 | )) | |
425 | ||
426 | else: | |
427 | ||
428 | logging.warning('No comment found for class %s' % Colt(class_name_doxy).magenta()) | |
06ccae0f | 429 | |
430 | ||
431 | ## Traverse the AST recursively starting from the current cursor. | |
432 | # | |
433 | # @param cursor A Clang parser cursor | |
434 | # @param filename Name of the current file | |
435 | # @param comments Array of comments: new ones will be appended there | |
436 | # @param recursion Current recursion depth | |
84039997 | 437 | # @param in_func True if we are inside a function or method |
438 | def traverse_ast(cursor, filename, comments, recursion=0, in_func=False): | |
06ccae0f | 439 | |
440 | # libclang traverses included files as well: we do not want this behavior | |
441 | if cursor.location.file is not None and str(cursor.location.file) != filename: | |
442 | logging.debug("Skipping processing of included %s" % cursor.location.file) | |
443 | return | |
444 | ||
445 | text = cursor.spelling or cursor.displayname | |
446 | kind = str(cursor.kind)[str(cursor.kind).index('.')+1:] | |
447 | ||
ccf0fca4 | 448 | is_macro = filename.endswith('.C') |
449 | ||
06ccae0f | 450 | indent = '' |
451 | for i in range(0, recursion): | |
452 | indent = indent + ' ' | |
453 | ||
2d3fd2ec | 454 | if cursor.kind in [ clang.cindex.CursorKind.CXX_METHOD, clang.cindex.CursorKind.CONSTRUCTOR, |
e215c6c6 | 455 | clang.cindex.CursorKind.DESTRUCTOR, clang.cindex.CursorKind.FUNCTION_DECL ]: |
06ccae0f | 456 | |
457 | # cursor ran into a C++ method | |
458 | logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) ) | |
459 | comment_method(cursor, comments) | |
84039997 | 460 | in_func = True |
06ccae0f | 461 | |
84039997 | 462 | elif not is_macro and not in_func and \ |
463 | cursor.kind in [ clang.cindex.CursorKind.FIELD_DECL, clang.cindex.CursorKind.VAR_DECL ]: | |
06ccae0f | 464 | |
465 | # cursor ran into a data member declaration | |
466 | logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) ) | |
467 | comment_datamember(cursor, comments) | |
468 | ||
469 | else: | |
470 | ||
471 | logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, kind, text) ) | |
472 | ||
473 | for child_cursor in cursor.get_children(): | |
84039997 | 474 | traverse_ast(child_cursor, filename, comments, recursion+1, in_func) |
06ccae0f | 475 | |
476 | if recursion == 0: | |
477 | comment_classdesc(filename, comments) | |
478 | ||
479 | ||
480 | ## Strip some HTML tags from the given string. Returns clean string. | |
481 | # | |
482 | # @param s Input string | |
483 | def strip_html(s): | |
798167cb | 484 | rehtml = r'(?i)</?(P|BR)/?>' |
06ccae0f | 485 | return re.sub(rehtml, '', s) |
486 | ||
487 | ||
488 | ## Remove garbage from comments and convert special tags from THtml to Doxygen. | |
489 | # | |
490 | # @param comment An array containing the lines of the original comment | |
ee1793c7 | 491 | def refactor_comment(comment, do_strip_html=True, infilename=None): |
06ccae0f | 492 | |
493 | recomm = r'^(/{2,}|/\*)? ?(\s*.*?)\s*((/{2,})?\s*|\*/)$' | |
494 | regarbage = r'^(?i)\s*([\s*=-_#]+|(Begin|End)_Html)\s*$' | |
495 | ||
03a6b7aa | 496 | # Support for LaTeX blocks spanning on multiple lines |
497 | relatex = r'(?i)^((.*?)\s+)?(BEGIN|END)_LATEX([.,;:\s]+.*)?$' | |
498 | in_latex = False | |
499 | latex_block = False | |
500 | ||
501 | # Support for LaTeX blocks on a single line | |
9db428b7 | 502 | reinline_latex = r'(?i)(.*)BEGIN_LATEX\s+(.*?)\s+END_LATEX(.*)$' |
503 | ||
35b193b4 | 504 | # Match <pre> (to turn it into the ~~~ Markdown syntax) |
505 | reblock = r'(?i)^(\s*)</?PRE>\s*$' | |
506 | ||
ee1793c7 | 507 | # Macro blocks for pictures generation |
508 | in_macro = False | |
509 | current_macro = [] | |
510 | remacro = r'(?i)^\s*(BEGIN|END)_MACRO(\((.*?)\))?\s*$' | |
511 | ||
06ccae0f | 512 | new_comment = [] |
513 | insert_blank = False | |
514 | wait_first_non_blank = True | |
515 | for line_comment in comment: | |
516 | ||
ee1793c7 | 517 | # Check if we are in a macro block |
518 | mmacro = re.search(remacro, line_comment) | |
519 | if mmacro: | |
520 | if in_macro: | |
521 | in_macro = False | |
522 | ||
523 | # Dump macro | |
524 | outimg = write_macro(infilename, current_macro) + '.png' | |
525 | current_macro = [] | |
526 | ||
527 | # Insert image | |
528 | new_comment.append( '![Picture from ROOT macro](%s)' % (outimg) ) | |
529 | ||
530 | logging.debug( 'Found macro for generating image %s' % Colt(outimg).magenta() ) | |
531 | ||
532 | else: | |
533 | in_macro = True | |
534 | ||
535 | continue | |
536 | elif in_macro: | |
537 | current_macro.append( line_comment ) | |
538 | continue | |
539 | ||
06ccae0f | 540 | # Strip some HTML tags |
541 | if do_strip_html: | |
542 | line_comment = strip_html(line_comment) | |
543 | ||
544 | mcomm = re.search( recomm, line_comment ) | |
545 | if mcomm: | |
546 | new_line_comment = mcomm.group(2) | |
547 | mgarbage = re.search( regarbage, new_line_comment ) | |
548 | ||
549 | if new_line_comment == '' or mgarbage is not None: | |
550 | insert_blank = True | |
551 | else: | |
552 | if insert_blank and not wait_first_non_blank: | |
553 | new_comment.append('') | |
554 | insert_blank = False | |
555 | wait_first_non_blank = False | |
9db428b7 | 556 | |
557 | # Postprocessing: LaTeX formulas in ROOT format | |
558 | # Marked by BEGIN_LATEX ... END_LATEX and they use # in place of \ | |
559 | # There can be several ROOT LaTeX forumlas per line | |
560 | while True: | |
561 | minline_latex = re.search( reinline_latex, new_line_comment ) | |
562 | if minline_latex: | |
563 | new_line_comment = '%s\\f$%s\\f$%s' % \ | |
564 | ( minline_latex.group(1), minline_latex.group(2).replace('#', '\\'), | |
565 | minline_latex.group(3) ) | |
566 | else: | |
567 | break | |
568 | ||
03a6b7aa | 569 | # ROOT LaTeX: do we have a Begin/End_LaTeX block? |
570 | # Note: the presence of LaTeX "closures" does not exclude the possibility to have a begin | |
571 | # block here left without a corresponding ending block | |
572 | mlatex = re.search( relatex, new_line_comment ) | |
573 | if mlatex: | |
574 | ||
575 | # before and after parts have been already stripped | |
576 | l_before = mlatex.group(2) | |
577 | l_after = mlatex.group(4) | |
578 | is_begin = mlatex.group(3).upper() == 'BEGIN' # if not, END | |
579 | ||
580 | if l_before is None: | |
581 | l_before = '' | |
582 | if l_after is None: | |
583 | l_after = '' | |
584 | ||
585 | if is_begin: | |
586 | ||
587 | # Begin of LaTeX part | |
588 | ||
589 | in_latex = True | |
590 | if l_before == '' and l_after == '': | |
591 | ||
592 | # Opening tag alone: mark the beginning of a block: \f[ ... \f] | |
593 | latex_block = True | |
594 | new_comment.append( '\\f[' ) | |
595 | ||
596 | else: | |
597 | # Mark the beginning of inline: \f$ ... \f$ | |
598 | latex_block = False | |
599 | new_comment.append( | |
600 | '%s \\f$%s' % ( l_before, l_after.replace('#', '\\') ) | |
601 | ) | |
602 | ||
603 | else: | |
604 | ||
605 | # End of LaTeX part | |
606 | in_latex = False | |
607 | ||
608 | if latex_block: | |
609 | ||
610 | # Closing a LaTeX block | |
611 | if l_before != '': | |
612 | new_comment.append( l_before.replace('#', '\\') ) | |
613 | new_comment.append( '\\f]' ) | |
614 | if l_after != '': | |
615 | new_comment.append( l_after ) | |
616 | ||
617 | else: | |
618 | ||
619 | # Closing a LaTeX inline | |
620 | new_comment.append( | |
621 | '%s\\f$%s' % ( l_before.replace('#', '\\'), l_after ) | |
622 | ) | |
623 | ||
624 | # Prevent appending lines (we have already done that) | |
625 | new_line_comment = None | |
626 | ||
35b193b4 | 627 | # If we are not in a LaTeX block, look for <pre> tags and transform them into Doxygen code |
628 | # blocks (using ~~~ ... ~~~). Only <pre> tags on a single line are supported | |
629 | if new_line_comment is not None and not in_latex: | |
630 | ||
631 | mblock = re.search( reblock, new_line_comment ) | |
632 | if mblock: | |
633 | new_comment.append( mblock.group(1)+'~~~' ) | |
634 | new_line_comment = None | |
635 | ||
03a6b7aa | 636 | if new_line_comment is not None: |
637 | if in_latex: | |
638 | new_line_comment = new_line_comment.replace('#', '\\') | |
639 | new_comment.append( new_line_comment ) | |
06ccae0f | 640 | |
641 | else: | |
642 | assert False, 'Comment regexp does not match' | |
643 | ||
644 | return new_comment | |
645 | ||
646 | ||
ee1793c7 | 647 | ## Dumps an image-generating macro to the correct place. Returns a string with the image path, |
648 | # without the extension. | |
649 | # | |
650 | # @param infilename File name of the source file | |
651 | # @param macro_lines Array of macro lines | |
652 | def write_macro(infilename, macro_lines): | |
653 | ||
654 | # Calculate hash | |
655 | digh = hashlib.sha1() | |
656 | for l in macro_lines: | |
657 | digh.update(l) | |
658 | digh.update('\n') | |
659 | short_digest = digh.hexdigest()[0:7] | |
660 | ||
661 | outdir = '%s/imgdoc' % os.path.dirname(infilename) | |
662 | outprefix = '%s/%s_%s' % ( | |
663 | outdir, | |
664 | os.path.basename(infilename).replace('.', '_'), | |
665 | short_digest | |
666 | ) | |
667 | outmacro = '%s.C' % outprefix | |
668 | ||
669 | # Make directory | |
670 | if not os.path.isdir(outdir): | |
671 | # do not catch: let everything die on error | |
672 | logging.debug('Creating directory %s' % Colt(outdir).magenta()) | |
673 | os.mkdir(outdir) | |
674 | ||
675 | # Create file (do not catch errors either) | |
676 | with open(outmacro, 'w') as omfp: | |
677 | logging.debug('Writing macro %s' % Colt(outmacro).magenta()) | |
678 | for l in macro_lines: | |
679 | omfp.write(l) | |
680 | omfp.write('\n') | |
681 | ||
682 | return outprefix | |
683 | ||
684 | ||
06ccae0f | 685 | ## Rewrites all comments from the given file handler. |
686 | # | |
687 | # @param fhin The file handler to read from | |
688 | # @param fhout The file handler to write to | |
689 | # @param comments Array of comments | |
690 | def rewrite_comments(fhin, fhout, comments): | |
691 | ||
692 | line_num = 0 | |
693 | in_comment = False | |
694 | skip_empty = False | |
695 | comm = None | |
696 | prev_comm = None | |
697 | ||
698 | rindent = r'^(\s*)' | |
699 | ||
c0094f3c | 700 | def dump_comment_block(cmt): |
fc82f9c7 | 701 | text_indent = '' |
702 | for i in range(0, cmt.indent): | |
703 | text_indent = text_indent + ' ' | |
704 | ||
705 | for lc in cmt.lines: | |
706 | fhout.write('%s///' % text_indent ) | |
707 | lc = lc.rstrip() | |
708 | if len(lc) != 0: | |
709 | fhout.write(' ') | |
710 | fhout.write(lc) | |
711 | fhout.write('\n') | |
712 | ||
713 | # Empty new line at the end of the comment | |
714 | fhout.write('\n') | |
c0094f3c | 715 | |
716 | ||
06ccae0f | 717 | for line in fhin: |
718 | ||
719 | line_num = line_num + 1 | |
720 | ||
721 | # Find current comment | |
722 | prev_comm = comm | |
723 | comm = None | |
724 | for c in comments: | |
725 | if c.has_comment(line_num): | |
726 | comm = c | |
727 | ||
728 | if comm: | |
729 | ||
730 | if isinstance(comm, MemberComment): | |
c0094f3c | 731 | |
732 | # end comment block | |
733 | if in_comment: | |
734 | dump_comment_block(prev_comm) | |
735 | in_comment = False | |
736 | ||
06ccae0f | 737 | non_comment = line[ 0:comm.first_col-1 ] |
738 | ||
21689d7a | 739 | if comm.array_size is not None or comm.is_dontsplit() or comm.is_ptr(): |
06ccae0f | 740 | |
21689d7a | 741 | # This is a special case: comment will be split in two lines: one before the comment for |
742 | # Doxygen as "member description", and the other right after the comment on the same line | |
743 | # to be parsed by ROOT's C++ parser | |
744 | ||
745 | # Keep indent on the generated line of comment before member definition | |
06ccae0f | 746 | mindent = re.search(rindent, line) |
21689d7a | 747 | |
748 | # Get correct comment flag, if any | |
749 | if comm.comment_flag is not None: | |
750 | cflag = comm.comment_flag | |
751 | else: | |
752 | cflag = '' | |
753 | ||
754 | # Get correct array size, if any | |
755 | if comm.array_size is not None: | |
756 | asize = '[%s]' % comm.array_size | |
06ccae0f | 757 | else: |
21689d7a | 758 | asize = '' |
06ccae0f | 759 | |
21689d7a | 760 | # Write on two lines |
761 | fhout.write('%s/// %s\n%s//%s%s\n' % ( | |
06ccae0f | 762 | mindent.group(1), |
763 | comm.lines[0], | |
764 | non_comment, | |
21689d7a | 765 | cflag, |
766 | asize | |
06ccae0f | 767 | )) |
768 | ||
769 | else: | |
770 | ||
21689d7a | 771 | # Single-line comments with the "transient" flag can be kept on one line in a way that |
772 | # they are correctly interpreted by both ROOT and Doxygen | |
773 | ||
774 | if comm.is_transient(): | |
06ccae0f | 775 | tt = '!' |
776 | else: | |
777 | tt = '/' | |
778 | ||
779 | fhout.write('%s//%s< %s\n' % ( | |
780 | non_comment, | |
781 | tt, | |
782 | comm.lines[0] | |
783 | )) | |
784 | ||
785 | elif isinstance(comm, RemoveComment): | |
c0094f3c | 786 | # End comment block and skip this line |
787 | if in_comment: | |
788 | dump_comment_block(prev_comm) | |
789 | in_comment = False | |
06ccae0f | 790 | |
791 | elif prev_comm is None: | |
c0094f3c | 792 | |
06ccae0f | 793 | # Beginning of a new comment block of type Comment |
794 | in_comment = True | |
795 | ||
796 | # Extract the non-comment part and print it if it exists | |
797 | non_comment = line[ 0:comm.first_col-1 ].rstrip() | |
798 | if non_comment != '': | |
799 | fhout.write( non_comment + '\n' ) | |
800 | ||
801 | else: | |
802 | ||
803 | if in_comment: | |
804 | ||
805 | # We have just exited a comment block of type Comment | |
c0094f3c | 806 | dump_comment_block(prev_comm) |
06ccae0f | 807 | in_comment = False |
06ccae0f | 808 | skip_empty = True |
809 | ||
810 | line_out = line.rstrip('\n') | |
811 | if skip_empty: | |
812 | skip_empty = False | |
813 | if line_out.strip() != '': | |
814 | fhout.write( line_out + '\n' ) | |
815 | else: | |
816 | fhout.write( line_out + '\n' ) | |
817 | ||
f329fa92 | 818 | |
819 | ## The main function. | |
820 | # | |
06ccae0f | 821 | # Return value is the executable's return value. |
f329fa92 | 822 | def main(argv): |
823 | ||
06ccae0f | 824 | # Setup logging on stderr |
825 | log_level = logging.INFO | |
826 | logging.basicConfig( | |
827 | level=log_level, | |
828 | format='%(levelname)-8s %(funcName)-20s %(message)s', | |
829 | stream=sys.stderr | |
830 | ) | |
f329fa92 | 831 | |
06ccae0f | 832 | # Parse command-line options |
833 | output_on_stdout = False | |
7afec95e | 834 | include_flags = [] |
06ccae0f | 835 | try: |
7afec95e | 836 | opts, args = getopt.getopt( argv, 'odI:', [ 'debug=', 'stdout' ] ) |
06ccae0f | 837 | for o, a in opts: |
838 | if o == '--debug': | |
839 | log_level = getattr( logging, a.upper(), None ) | |
840 | if not isinstance(log_level, int): | |
841 | raise getopt.GetoptError('log level must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL') | |
842 | elif o == '-d': | |
843 | log_level = logging.DEBUG | |
844 | elif o == '-o' or o == '--stdout': | |
06ccae0f | 845 | output_on_stdout = True |
7afec95e | 846 | elif o == '-I': |
847 | if os.path.isdir(a): | |
848 | include_flags.extend( [ '-I', a ] ) | |
849 | else: | |
850 | logging.fatal('Include directory not found: %s' % Colt(a).magenta()) | |
851 | return 2 | |
06ccae0f | 852 | else: |
853 | assert False, 'Unhandled argument' | |
854 | except getopt.GetoptError as e: | |
855 | logging.fatal('Invalid arguments: %s' % e) | |
856 | return 1 | |
f329fa92 | 857 | |
06ccae0f | 858 | logging.getLogger('').setLevel(log_level) |
f329fa92 | 859 | |
06ccae0f | 860 | # Attempt to load libclang from a list of known locations |
861 | libclang_locations = [ | |
862 | '/usr/lib/llvm-3.5/lib/libclang.so.1', | |
863 | '/usr/lib/libclang.so', | |
864 | '/Library/Developer/CommandLineTools/usr/lib/libclang.dylib' | |
865 | ] | |
866 | libclang_found = False | |
f329fa92 | 867 | |
06ccae0f | 868 | for lib in libclang_locations: |
869 | if os.path.isfile(lib): | |
870 | clang.cindex.Config.set_library_file(lib) | |
871 | libclang_found = True | |
872 | break | |
f329fa92 | 873 | |
06ccae0f | 874 | if not libclang_found: |
875 | logging.fatal('Cannot find libclang') | |
876 | return 1 | |
877 | ||
878 | # Loop over all files | |
879 | for fn in args: | |
880 | ||
881 | logging.info('Input file: %s' % Colt(fn).magenta()) | |
882 | index = clang.cindex.Index.create() | |
7afec95e | 883 | clang_args = [ '-x', 'c++' ] |
884 | clang_args.extend( include_flags ) | |
885 | translation_unit = index.parse(fn, args=clang_args) | |
06ccae0f | 886 | |
887 | comments = [] | |
888 | traverse_ast( translation_unit.cursor, fn, comments ) | |
889 | for c in comments: | |
890 | ||
891 | logging.debug("Comment found for entity %s:" % Colt(c.func).magenta()) | |
f329fa92 | 892 | |
06ccae0f | 893 | if isinstance(c, MemberComment): |
894 | ||
21689d7a | 895 | if c.is_transient(): |
896 | flag_text = Colt('transient ').yellow() | |
897 | elif c.is_dontsplit(): | |
898 | flag_text = Colt('dontsplit ').yellow() | |
899 | elif c.is_ptr(): | |
900 | flag_text = Colt('ptr ').yellow() | |
06ccae0f | 901 | else: |
21689d7a | 902 | flag_text = '' |
06ccae0f | 903 | |
904 | if c.array_size is not None: | |
905 | array_text = Colt('arraysize=%s ' % c.array_size).yellow() | |
906 | else: | |
907 | array_text = '' | |
908 | ||
909 | logging.debug( | |
910 | "%s %s%s{%s}" % ( \ | |
911 | Colt("[%d,%d]" % (c.first_line, c.first_col)).green(), | |
21689d7a | 912 | flag_text, |
06ccae0f | 913 | array_text, |
914 | Colt(c.lines[0]).cyan() | |
915 | )) | |
916 | ||
917 | elif isinstance(c, RemoveComment): | |
918 | ||
919 | logging.debug( Colt('[%d,%d]' % (c.first_line, c.last_line)).green() ) | |
920 | ||
921 | else: | |
922 | for l in c.lines: | |
923 | logging.debug( | |
924 | Colt("[%d,%d:%d,%d] " % (c.first_line, c.first_col, c.last_line, c.last_col)).green() + | |
925 | "{%s}" % Colt(l).cyan() | |
926 | ) | |
f329fa92 | 927 | |
928 | try: | |
06ccae0f | 929 | |
930 | if output_on_stdout: | |
931 | with open(fn, 'r') as fhin: | |
932 | rewrite_comments( fhin, sys.stdout, comments ) | |
933 | else: | |
934 | fn_back = fn + '.thtml2doxy_backup' | |
935 | os.rename( fn, fn_back ) | |
936 | ||
937 | with open(fn_back, 'r') as fhin, open(fn, 'w') as fhout: | |
938 | rewrite_comments( fhin, fhout, comments ) | |
939 | ||
940 | os.remove( fn_back ) | |
941 | logging.info("File %s converted to Doxygen: check differences before committing!" % Colt(fn).magenta()) | |
942 | except (IOError,OSError) as e: | |
943 | logging.error('File operation failed: %s' % e) | |
f329fa92 | 944 | |
945 | return 0 | |
946 | ||
06ccae0f | 947 | |
f329fa92 | 948 | if __name__ == '__main__': |
06ccae0f | 949 | sys.exit( main( sys.argv[1:] ) ) |