import re
import logging
import getopt
+import hashlib
import clang.cindex
## A data member comment.
class MemberComment:
- def __init__(self, text, is_transient, array_size, first_line, first_col, func):
+ def __init__(self, text, comment_flag, array_size, first_line, first_col, func):
assert first_line > 0, 'Wrong line number'
+ assert comment_flag is None or comment_flag == '!' or comment_flag in [ '!', '||', '->' ]
self.lines = [ text ]
- self.is_transient = is_transient
+ self.comment_flag = comment_flag
self.array_size = array_size
self.first_line = first_line
self.first_col = first_col
self.func = func
+ def is_transient(self):
+ return self.comment_flag == '!'
+
+ def is_dontsplit(self):
+ return self.comment_flag == '||'
+
+ def is_ptr(self):
+ return self.comment_flag == '->'
+
def has_comment(self, line):
return line == self.first_line
def __str__(self):
- if self.is_transient:
+ if self.is_transient():
tt = '!transient! '
+ elif self.is_dontsplit():
+ tt = '!dontsplit! '
+ elif self.is_ptr():
+ tt = '!ptr! '
else:
tt = ''
if comment_line_start > 0:
- comment = refactor_comment( comment )
+ comment = refactor_comment( comment, infilename=str(cursor.location.file) )
if len(comment) > 0:
logging.debug("Comment found for function %s" % Colt(comment_function).magenta())
prev = None
found = False
- # Huge overkill
+ # Huge overkill: current line saved in "raw", previous in "prev"
with open(str(cursor.location.file)) as fp:
cur_line = 0
for raw in fp:
assert found, 'A line that should exist was not found in file' % cursor.location.file
- recomm = r'(//(!)|///?)(\[(.*?)\])?<?\s*(.*?)\s*$'
- recomm_doxyary = r'^\s*///\s*(.*?)\s*$'
+ recomm = r'(//(!|\|\||->)|///?)(\[([0-9,]+)\])?<?\s*(.*?)\s*$'
+ recomm_prevline = r'^\s*///\s*(.*?)\s*$'
mcomm = re.search(recomm, raw)
if mcomm:
# If it does not match, we do not have a comment
member_name = cursor.spelling;
- is_transient = mcomm.group(2) is not None
+ comment_flag = mcomm.group(2)
array_size = mcomm.group(4)
text = mcomm.group(5)
if array_size is not None and prev is not None:
# ROOT arrays with comments already converted to Doxygen have the member description on the
# previous line
- mcomm_doxyary = re.search(recomm_doxyary, prev)
- if mcomm_doxyary:
- text = mcomm_doxyary.group(1)
+ mcomm_prevline = re.search(recomm_prevline, prev)
+ if mcomm_prevline:
+ text = mcomm_prevline.group(1)
comments.append(RemoveComment(line_num-1, line_num-1))
logging.debug('Comment found for member %s' % Colt(member_name).magenta())
comments.append( MemberComment(
text,
- is_transient,
+ comment_flag,
array_size,
line_num,
col_num,
recomm = r'^\s*///?(\s*.*?)\s*/*\s*$'
- reclass_doxy = r'(?i)^\s*\\class:?\s*(.*?)\s*$'
+ reclass_doxy = r'(?i)^\s*\\(class|file):?\s*([^.]*)'
class_name_doxy = None
reauthor = r'(?i)^\s*\\?authors?:?\s*(.*?)\s*(,?\s*([0-9./-]+))?\s*$'
line_num = 0
+ is_macro = filename.endswith('.C')
+
with open(filename, 'r') as fp:
for raw in fp:
if raw.strip() == '' and start_line > 0:
# Skip empty lines
- end_line = line_num - 1
continue
stripped = strip_html(raw)
start_line = line_num
+ end_line = line_num
append = True
mclass_doxy = re.search(reclass_doxy, mcomm.group(1))
if mclass_doxy:
- class_name_doxy = mclass_doxy.group(1)
+ class_name_doxy = mclass_doxy.group(2)
append = False
else:
mauthor = re.search(reauthor, mcomm.group(1))
author = mauthor.group(1)
if date is None:
# Date specified in the standalone \date field has priority
- date = mauthor.group(2)
+ date = mauthor.group(3)
append = False
else:
mdate = re.search(redate, mcomm.group(1))
else:
if start_line > 0:
- # End of our comment
- if end_line == -1:
- end_line = line_num - 1
break
if class_name_doxy is None:
if start_line > 0:
- # Prepend \class specifier (and an empty line)
- comment_lines[:0] = [ '\\class ' + class_name_doxy ]
+ # Prepend \class or \file specifier (and an empty line)
+ if is_macro:
+ comment_lines[:0] = [ '\\file ' + class_name_doxy + '.C' ]
+ else:
+ comment_lines[:0] = [ '\\class ' + class_name_doxy ]
# Append author and date if they exist
comment_lines.append('')
if date is not None:
comment_lines.append( '\\date ' + date )
- comment_lines = refactor_comment(comment_lines, do_strip_html=False)
+ comment_lines = refactor_comment(comment_lines, do_strip_html=False, infilename=filename)
logging.debug('Comment found for class %s' % Colt(class_name_doxy).magenta())
comments.append(Comment(
comment_lines,
text = cursor.spelling or cursor.displayname
kind = str(cursor.kind)[str(cursor.kind).index('.')+1:]
+ is_macro = filename.endswith('.C')
+
indent = ''
for i in range(0, recursion):
indent = indent + ' '
- if cursor.kind == clang.cindex.CursorKind.CXX_METHOD or cursor.kind == clang.cindex.CursorKind.CONSTRUCTOR or cursor.kind == clang.cindex.CursorKind.DESTRUCTOR:
+ if cursor.kind in [ clang.cindex.CursorKind.CXX_METHOD, clang.cindex.CursorKind.CONSTRUCTOR,
+ clang.cindex.CursorKind.DESTRUCTOR, clang.cindex.CursorKind.FUNCTION_DECL ]:
# cursor ran into a C++ method
logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
comment_method(cursor, comments)
- elif cursor.kind == clang.cindex.CursorKind.FIELD_DECL:
+ elif not is_macro and cursor.kind in [ clang.cindex.CursorKind.FIELD_DECL, clang.cindex.CursorKind.VAR_DECL ]:
# cursor ran into a data member declaration
logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
#
# @param s Input string
def strip_html(s):
- rehtml = r'(?i)</?(P|H[0-9]|BR)/?>'
+ rehtml = r'(?i)</?(P|BR)/?>'
return re.sub(rehtml, '', s)
## Remove garbage from comments and convert special tags from THtml to Doxygen.
#
# @param comment An array containing the lines of the original comment
-def refactor_comment(comment, do_strip_html=True):
+def refactor_comment(comment, do_strip_html=True, infilename=None):
recomm = r'^(/{2,}|/\*)? ?(\s*.*?)\s*((/{2,})?\s*|\*/)$'
regarbage = r'^(?i)\s*([\s*=-_#]+|(Begin|End)_Html)\s*$'
+ # Support for LaTeX blocks spanning on multiple lines
+ relatex = r'(?i)^((.*?)\s+)?(BEGIN|END)_LATEX([.,;:\s]+.*)?$'
+ in_latex = False
+ latex_block = False
+
+ # Support for LaTeX blocks on a single line
+ reinline_latex = r'(?i)(.*)BEGIN_LATEX\s+(.*?)\s+END_LATEX(.*)$'
+
+ # Match <pre> (to turn it into the ~~~ Markdown syntax)
+ reblock = r'(?i)^(\s*)</?PRE>\s*$'
+
+ # Macro blocks for pictures generation
+ in_macro = False
+ current_macro = []
+ remacro = r'(?i)^\s*(BEGIN|END)_MACRO(\((.*?)\))?\s*$'
+
new_comment = []
insert_blank = False
wait_first_non_blank = True
for line_comment in comment:
+ # Check if we are in a macro block
+ mmacro = re.search(remacro, line_comment)
+ if mmacro:
+ if in_macro:
+ in_macro = False
+
+ # Dump macro
+ outimg = write_macro(infilename, current_macro) + '.png'
+ current_macro = []
+
+ # Insert image
+ new_comment.append( '' % (outimg) )
+
+ logging.debug( 'Found macro for generating image %s' % Colt(outimg).magenta() )
+
+ else:
+ in_macro = True
+
+ continue
+ elif in_macro:
+ current_macro.append( line_comment )
+ continue
+
# Strip some HTML tags
if do_strip_html:
line_comment = strip_html(line_comment)
new_comment.append('')
insert_blank = False
wait_first_non_blank = False
- new_comment.append( new_line_comment )
+
+ # Postprocessing: LaTeX formulas in ROOT format
+ # Marked by BEGIN_LATEX ... END_LATEX and they use # in place of \
+ # There can be several ROOT LaTeX forumlas per line
+ while True:
+ minline_latex = re.search( reinline_latex, new_line_comment )
+ if minline_latex:
+ new_line_comment = '%s\\f$%s\\f$%s' % \
+ ( minline_latex.group(1), minline_latex.group(2).replace('#', '\\'),
+ minline_latex.group(3) )
+ else:
+ break
+
+ # ROOT LaTeX: do we have a Begin/End_LaTeX block?
+ # Note: the presence of LaTeX "closures" does not exclude the possibility to have a begin
+ # block here left without a corresponding ending block
+ mlatex = re.search( relatex, new_line_comment )
+ if mlatex:
+
+ # before and after parts have been already stripped
+ l_before = mlatex.group(2)
+ l_after = mlatex.group(4)
+ is_begin = mlatex.group(3).upper() == 'BEGIN' # if not, END
+
+ if l_before is None:
+ l_before = ''
+ if l_after is None:
+ l_after = ''
+
+ if is_begin:
+
+ # Begin of LaTeX part
+
+ in_latex = True
+ if l_before == '' and l_after == '':
+
+ # Opening tag alone: mark the beginning of a block: \f[ ... \f]
+ latex_block = True
+ new_comment.append( '\\f[' )
+
+ else:
+ # Mark the beginning of inline: \f$ ... \f$
+ latex_block = False
+ new_comment.append(
+ '%s \\f$%s' % ( l_before, l_after.replace('#', '\\') )
+ )
+
+ else:
+
+ # End of LaTeX part
+ in_latex = False
+
+ if latex_block:
+
+ # Closing a LaTeX block
+ if l_before != '':
+ new_comment.append( l_before.replace('#', '\\') )
+ new_comment.append( '\\f]' )
+ if l_after != '':
+ new_comment.append( l_after )
+
+ else:
+
+ # Closing a LaTeX inline
+ new_comment.append(
+ '%s\\f$%s' % ( l_before.replace('#', '\\'), l_after )
+ )
+
+ # Prevent appending lines (we have already done that)
+ new_line_comment = None
+
+ # If we are not in a LaTeX block, look for <pre> tags and transform them into Doxygen code
+ # blocks (using ~~~ ... ~~~). Only <pre> tags on a single line are supported
+ if new_line_comment is not None and not in_latex:
+
+ mblock = re.search( reblock, new_line_comment )
+ if mblock:
+ new_comment.append( mblock.group(1)+'~~~' )
+ new_line_comment = None
+
+ if new_line_comment is not None:
+ if in_latex:
+ new_line_comment = new_line_comment.replace('#', '\\')
+ new_comment.append( new_line_comment )
else:
assert False, 'Comment regexp does not match'
return new_comment
+## Dumps an image-generating macro to the correct place. Returns a string with the image path,
+# without the extension.
+#
+# @param infilename File name of the source file
+# @param macro_lines Array of macro lines
+def write_macro(infilename, macro_lines):
+
+ # Calculate hash
+ digh = hashlib.sha1()
+ for l in macro_lines:
+ digh.update(l)
+ digh.update('\n')
+ short_digest = digh.hexdigest()[0:7]
+
+ outdir = '%s/imgdoc' % os.path.dirname(infilename)
+ outprefix = '%s/%s_%s' % (
+ outdir,
+ os.path.basename(infilename).replace('.', '_'),
+ short_digest
+ )
+ outmacro = '%s.C' % outprefix
+
+ # Make directory
+ if not os.path.isdir(outdir):
+ # do not catch: let everything die on error
+ logging.debug('Creating directory %s' % Colt(outdir).magenta())
+ os.mkdir(outdir)
+
+ # Create file (do not catch errors either)
+ with open(outmacro, 'w') as omfp:
+ logging.debug('Writing macro %s' % Colt(outmacro).magenta())
+ for l in macro_lines:
+ omfp.write(l)
+ omfp.write('\n')
+
+ return outprefix
+
+
## Rewrites all comments from the given file handler.
#
# @param fhin The file handler to read from
rindent = r'^(\s*)'
+
+ def dump_comment_block(cmt):
+ text_indent = ''
+ for i in range(0, cmt.indent):
+ text_indent = text_indent + ' '
+
+ for lc in cmt.lines:
+ fhout.write( "%s/// %s\n" % (text_indent, lc) );
+ fhout.write('\n')
+
+
for line in fhin:
line_num = line_num + 1
if comm:
if isinstance(comm, MemberComment):
+
+ # end comment block
+ if in_comment:
+ dump_comment_block(prev_comm)
+ in_comment = False
+
non_comment = line[ 0:comm.first_col-1 ]
- if comm.array_size is not None:
+ if comm.array_size is not None or comm.is_dontsplit() or comm.is_ptr():
+ # This is a special case: comment will be split in two lines: one before the comment for
+ # Doxygen as "member description", and the other right after the comment on the same line
+ # to be parsed by ROOT's C++ parser
+
+ # Keep indent on the generated line of comment before member definition
mindent = re.search(rindent, line)
- if comm.is_transient:
- tt = '!'
+
+ # Get correct comment flag, if any
+ if comm.comment_flag is not None:
+ cflag = comm.comment_flag
else:
- tt = ''
+ cflag = ''
- # Special case: we need multiple lines not to confuse ROOT's C++ parser
- fhout.write('%s/// %s\n%s//%s[%s]\n' % (
+ # Get correct array size, if any
+ if comm.array_size is not None:
+ asize = '[%s]' % comm.array_size
+ else:
+ asize = ''
+
+ # Write on two lines
+ fhout.write('%s/// %s\n%s//%s%s\n' % (
mindent.group(1),
comm.lines[0],
non_comment,
- tt,
- comm.array_size
+ cflag,
+ asize
))
else:
- if comm.is_transient:
+ # Single-line comments with the "transient" flag can be kept on one line in a way that
+ # they are correctly interpreted by both ROOT and Doxygen
+
+ if comm.is_transient():
tt = '!'
else:
tt = '/'
))
elif isinstance(comm, RemoveComment):
- # Do nothing: just skip line
- pass
+ # End comment block and skip this line
+ if in_comment:
+ dump_comment_block(prev_comm)
+ in_comment = False
elif prev_comm is None:
+
# Beginning of a new comment block of type Comment
in_comment = True
if in_comment:
# We have just exited a comment block of type Comment
+ dump_comment_block(prev_comm)
in_comment = False
-
- # Dump revamped comment, if applicable
- text_indent = ''
- for i in range(0,prev_comm.indent):
- text_indent = text_indent + ' '
-
- for lc in prev_comm.lines:
- fhout.write( "%s/// %s\n" % (text_indent, lc) );
- fhout.write('\n')
skip_empty = True
line_out = line.rstrip('\n')
# Parse command-line options
output_on_stdout = False
+ include_flags = []
try:
- opts, args = getopt.getopt( argv, 'od', [ 'debug=', 'stdout' ] )
+ opts, args = getopt.getopt( argv, 'odI:', [ 'debug=', 'stdout' ] )
for o, a in opts:
if o == '--debug':
log_level = getattr( logging, a.upper(), None )
elif o == '-d':
log_level = logging.DEBUG
elif o == '-o' or o == '--stdout':
- logging.debug('Output on stdout instead of replacing original files')
output_on_stdout = True
+ elif o == '-I':
+ if os.path.isdir(a):
+ include_flags.extend( [ '-I', a ] )
+ else:
+ logging.fatal('Include directory not found: %s' % Colt(a).magenta())
+ return 2
else:
assert False, 'Unhandled argument'
except getopt.GetoptError as e:
logging.info('Input file: %s' % Colt(fn).magenta())
index = clang.cindex.Index.create()
- translation_unit = index.parse(fn, args=['-x', 'c++'])
+ clang_args = [ '-x', 'c++' ]
+ clang_args.extend( include_flags )
+ translation_unit = index.parse(fn, args=clang_args)
comments = []
traverse_ast( translation_unit.cursor, fn, comments )
if isinstance(c, MemberComment):
- if c.is_transient:
- transient_text = Colt('transient ').yellow()
+ if c.is_transient():
+ flag_text = Colt('transient ').yellow()
+ elif c.is_dontsplit():
+ flag_text = Colt('dontsplit ').yellow()
+ elif c.is_ptr():
+ flag_text = Colt('ptr ').yellow()
else:
- transient_text = ''
+ flag_text = ''
if c.array_size is not None:
array_text = Colt('arraysize=%s ' % c.array_size).yellow()
logging.debug(
"%s %s%s{%s}" % ( \
Colt("[%d,%d]" % (c.first_line, c.first_col)).green(),
- transient_text,
+ flag_text,
array_text,
Colt(c.lines[0]).cyan()
))
git://git.uio.no / u / mrichter / AliRoot.git / blobdiff