contrib/redox.py

   1 #!/usr/bin/python
   2 #
   3 #  Copyright (c) 2008 The Tor Project, Inc.
   4 #  See LICENSE for licensing information.
   5 #
   6 # Hi!
   7 # I'm redox.py, the Tor redocumentation tool!
   8 # I am a horrible hack!
   9 # I read the output of doxygen from stderr, and add missing DOCDOC comments
  10 #   to tell you where documentation should go!
  11 # To use me, edit the stuff below...
  12 #  ...and run 'make doxygen 2>doxygen.stderr' ...
  13 #  ...and run ./contrib/redox.py < doxygen.stderr !
  14 # I'll make a bunch of new files by adding missing DOCDOC comments to your
  15 #    source.  Those files will have names like ./src/common/util.c.newdoc.
  16 # You will want to look over the changes by hand before checking them in.
  17 #
  18 # So, here's your workflow:
  19 #
  20 # 0. Make sure you're running a bourne shell for the redirects below.
  21 # 1. make doxygen 1>doxygen.stdout 2>doxygen.stderr.
  22 # 2. grep Warning doxygen.stderr | grep -v 'is not documented' | less
  23 #      [This will tell you about all the bogus doxygen output you have]
  24 # 3. python ./contrib/redox.py <doxygen.stderr
  25 #      [This will make lots of .newdoc files with DOCDOC comments for
  26 #       whatever was missing documentation.]
  27 # 4. Look over those .newdoc files, and see which docdoc comments you
  28 #     want to merge into the main file.  If it's all good, just run
  29 #     "mv fname.c.newdoc fname.c".  Otherwise, you'll need to merge
  30 #     the parts you like by hand.
  31
  32 # Which files should we ignore warning from?  Mostly, these are external
  33 # files that we've snarfed in from somebody else, whose C we do no intend
  34 # to document for them.
  35 SKIP_FILES = [ "OpenBSD_malloc_Linux.c",
  36                "eventdns.c",
  37                "eventdns.h",
  38                "strlcat.c",
  39                "strlcpy.c",
  40                "aes.c",
  41                "aes.h" ]
  42
  43 # What names of things never need javadoc
  44 SKIP_NAME_PATTERNS = [ r'^.*_c_id$',
  45                        r'^.*_H_ID$' ]
  46
  47 # Which types of things should get DOCDOC comments added if they are
  48 # missing documentation?  Recognized types are in KINDS below.
  49 #ADD_DOCDOCS_TO_TYPES = [ 'function', 'type', 'typedef' ]
  50 ADD_DOCDOCS_TO_TYPES = [ 'variable' ]
  51
  52 # ====================
  53 # The rest of this should not need hacking.
  54
  55 import re
  56 import sys
  57
  58 KINDS = [ "type", "field", "typedef", "define", "function", "variable" ]
  59
  60 NODOC_LINE_RE = re.compile(r'^([^:]+):(\d+): (\w+): (.*) is not documented\.$')
  61
  62 THING_RE = re.compile(r'^Member ([a-zA-Z0-9_]+).*\((typedef|define|function|variable)\) of (file|class) ')
  63
  64 SKIP_NAMES = [re.compile(s) for s in SKIP_NAME_PATTERNS]
  65
  66 def parsething(thing):
  67     """I figure out what 'foobar baz in quux quum is not documented' means,
  68        and return: the name of the foobar, and the kind of the foobar.
  69     """
  70     if thing.startswith("Compound "):
  71         tp, name = "type", thing.split()[1]
  72     else:
  73         m = THING_RE.match(thing)
  74         if not m:
  75             print thing
  76             return None, None
  77         else:
  78             name, tp, parent = m.groups()
  79             if parent == 'class':
  80                 if tp == 'variable' or tp == 'function':
  81                     tp = 'field'
  82
  83     return name, tp
  84
  85 def read():
  86     """I snarf doxygen stderr from stdin, and parse all the "foo has no
  87        documentation messages.  I return a map from filename to lists
  88        of tuples of (alleged line number, name of thing, kind of thing)
  89     """
  90     errs = {}
  91     for line in sys.stdin:
  92         m = NODOC_LINE_RE.match(line)
  93         if m:
  94             file, line, tp, thing = m.groups()
  95             assert tp == 'Warning'
  96             name, kind = parsething(thing)
  97             errs.setdefault(file, []).append((int(line), name, kind))
  98
  99     return errs
 100
 101 def findline(lines, lineno, ident):
 102     """Given a list of all the lines in the file (adjusted so 1-indexing works),
 103        a line number that ident is alledgedly on, and ident, I figure out
 104        the line where ident was really declared."""
 105     for lineno in xrange(lineno, 0, -1):
 106         if ident in lines[lineno]:
 107             return lineno
 108
 109     return None
 110
 111 FUNC_PAT = re.compile(r"^[A-Za-z0-9_]+\(")
 112
 113 def hascomment(lines, lineno, kind):
 114     """I return true if it looks like there's already a good comment about
 115        the thing on lineno of lines of type kind. """
 116     if "*/" in lines[lineno-1]:
 117         return True
 118     if kind == 'function' and FUNC_PAT.match(lines[lineno]):
 119         if "*/" in lines[lineno-2]:
 120             return True
 121     return False
 122
 123 def hasdocdoc(lines, lineno, kind):
 124     """I return true if it looks like there's already a docdoc comment about
 125        the thing on lineno of lines of type kind."""
 126     if "DOCDOC" in lines[lineno] or "DOCDOC" in lines[lineno-1]:
 127         return True
 128     if kind == 'function' and FUNC_PAT.match(lines[lineno]):
 129         if "DOCDOC" in lines[lineno-2]:
 130             return True
 131     return False
 132
 133 def checkf(fn, errs):
 134     """I go through the output of read() for a single file, and build a list
 135        of tuples of things that want DOCDOC comments.  Each tuple has:
 136        the line number where the comment goes; the kind of thing; its name.
 137     """
 138     for skip in SKIP_FILES:
 139         if fn.endswith(skip):
 140             print "Skipping",fn
 141             return
 142
 143     comments = []
 144     lines = [ None ]
 145     try:
 146         lines.extend( open(fn, 'r').readlines() )
 147     except IOError:
 148         return
 149
 150     for line, name, kind in errs:
 151         if any(pat.match(name) for pat in SKIP_NAMES):
 152             continue
 153
 154         if kind not in ADD_DOCDOCS_TO_TYPES:
 155             continue
 156
 157         ln = findline(lines, line, name)
 158         if ln == None:
 159             print "Couldn't find the definition of %s allegedly on %s of %s"%(
 160                 name, line, fn)
 161         else:
 162             if hasdocdoc(lines, line, kind):
 163 #                print "Has a DOCDOC"
 164 #                print fn, line, name, kind
 165 #                print "\t",lines[line-2],
 166 #                print "\t",lines[line-1],
 167 #                print "\t",lines[line],
 168 #                print "-------"
 169                 pass
 170             else:
 171                 if kind == 'function' and FUNC_PAT.match(lines[ln]):
 172                     ln = ln - 1
 173
 174                 comments.append((ln, kind, name))
 175
 176     return comments
 177
 178 def applyComments(fn, entries):
 179     """I apply lots of comments to the file in fn, making a new .newdoc file.
 180     """
 181     N = 0
 182
 183     lines = [ None ]
 184     try:
 185         lines.extend( open(fn, 'r').readlines() )
 186     except IOError:
 187         return
 188
 189     # Process the comments in reverse order by line number, so that
 190     # the line numbers for the ones we haven't added yet remain valid
 191     # until we add them.  Standard trick.
 192     entries.sort()
 193     entries.reverse()
 194
 195     for ln, kind, name in entries:
 196
 197         lines.insert(ln, "/* DOCDOC %s */\n"%name)
 198         N += 1
 199
 200     outf = open(fn+".newdoc", 'w')
 201     for line in lines[1:]:
 202         outf.write(line)
 203     outf.close()
 204
 205     print "Added %s DOCDOCs to %s" %(N, fn)
 206
 207 e = read()
 208
 209 for fn, errs in e.iteritems():
 210     comments = checkf(fn, errs)
 211     if comments:
 212         applyComments(fn, comments)