contrib/redox.py

   1 #!/usr/bin/python
   2 #
   3 #  Copyright (c) 2008-2013, 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                "sha256.c",
  41                "sha256.h",
  42                "aes.c",
  43                "aes.h" ]
  44
  45 # What names of things never need javadoc
  46 SKIP_NAME_PATTERNS = [ r'^.*_c_id$',
  47                        r'^.*_H_ID$' ]
  48
  49 # Which types of things should get DOCDOC comments added if they are
  50 # missing documentation?  Recognized types are in KINDS below.
  51 ADD_DOCDOCS_TO_TYPES = [ 'function', 'type', 'typedef' ]
  52 ADD_DOCDOCS_TO_TYPES += [ 'variable', ]
  53
  54 # ====================
  55 # The rest of this should not need hacking.
  56
  57 import re
  58 import sys
  59
  60 KINDS = [ "type", "field", "typedef", "define", "function", "variable",
  61           "enumeration" ]
  62
  63 NODOC_LINE_RE = re.compile(r'^([^:]+):(\d+): (\w+): (.*) is not documented\.$')
  64
  65 THING_RE = re.compile(r'^Member ([a-zA-Z0-9_]+).*\((typedef|define|function|variable|enumeration)\) of (file|class) ')
  66
  67 SKIP_NAMES = [re.compile(s) for s in SKIP_NAME_PATTERNS]
  68
  69 def parsething(thing):
  70     """I figure out what 'foobar baz in quux quum is not documented' means,
  71        and return: the name of the foobar, and the kind of the foobar.
  72     """
  73     if thing.startswith("Compound "):
  74         tp, name = "type", thing.split()[1]
  75     else:
  76         m = THING_RE.match(thing)
  77         if not m:
  78             print thing, "???? Format didn't match."
  79             return None, None
  80         else:
  81             name, tp, parent = m.groups()
  82             if parent == 'class':
  83                 if tp == 'variable' or tp == 'function':
  84                     tp = 'field'
  85
  86     return name, tp
  87
  88 def read():
  89     """I snarf doxygen stderr from stdin, and parse all the "foo has no
  90        documentation messages.  I return a map from filename to lists
  91        of tuples of (alleged line number, name of thing, kind of thing)
  92     """
  93     errs = {}
  94     for line in sys.stdin:
  95         m = NODOC_LINE_RE.match(line)
  96         if m:
  97             file, line, tp, thing = m.groups()
  98             assert tp.lower() == 'warning'
  99             name, kind = parsething(thing)
 100             errs.setdefault(file, []).append((int(line), name, kind))
 101
 102     return errs
 103
 104 def findline(lines, lineno, ident):
 105     """Given a list of all the lines in the file (adjusted so 1-indexing works),
 106        a line number that ident is alledgedly on, and ident, I figure out
 107        the line where ident was really declared."""
 108     for lineno in xrange(lineno, 0, -1):
 109         if ident in lines[lineno]:
 110             return lineno
 111
 112     return None
 113
 114 FUNC_PAT = re.compile(r"^[A-Za-z0-9_]+\(")
 115
 116 def hascomment(lines, lineno, kind):
 117     """I return true if it looks like there's already a good comment about
 118        the thing on lineno of lines of type kind. """
 119     if "*/" in lines[lineno-1]:
 120         return True
 121     if kind == 'function' and FUNC_PAT.match(lines[lineno]):
 122         if "*/" in lines[lineno-2]:
 123             return True
 124     return False
 125
 126 def hasdocdoc(lines, lineno, kind):
 127     """I return true if it looks like there's already a docdoc comment about
 128        the thing on lineno of lines of type kind."""
 129     if "DOCDOC" in lines[lineno] or "DOCDOC" in lines[lineno-1]:
 130         return True
 131     if kind == 'function' and FUNC_PAT.match(lines[lineno]):
 132         if "DOCDOC" in lines[lineno-2]:
 133             return True
 134     return False
 135
 136 def checkf(fn, errs):
 137     """I go through the output of read() for a single file, and build a list
 138        of tuples of things that want DOCDOC comments.  Each tuple has:
 139        the line number where the comment goes; the kind of thing; its name.
 140     """
 141     for skip in SKIP_FILES:
 142         if fn.endswith(skip):
 143             print "Skipping",fn
 144             return
 145
 146     comments = []
 147     lines = [ None ]
 148     try:
 149         lines.extend( open(fn, 'r').readlines() )
 150     except IOError:
 151         return
 152
 153     for line, name, kind in errs:
 154         if any(pat.match(name) for pat in SKIP_NAMES):
 155             continue
 156
 157         if kind not in ADD_DOCDOCS_TO_TYPES:
 158             continue
 159
 160         ln = findline(lines, line, name)
 161         if ln == None:
 162             print "Couldn't find the definition of %s allegedly on %s of %s"%(
 163                 name, line, fn)
 164         else:
 165             if hasdocdoc(lines, line, kind):
 166 #                print "Has a DOCDOC"
 167 #                print fn, line, name, kind
 168 #                print "\t",lines[line-2],
 169 #                print "\t",lines[line-1],
 170 #                print "\t",lines[line],
 171 #                print "-------"
 172                 pass
 173             else:
 174                 if kind == 'function' and FUNC_PAT.match(lines[ln]):
 175                     ln = ln - 1
 176
 177                 comments.append((ln, kind, name))
 178
 179     return comments
 180
 181 def applyComments(fn, entries):
 182     """I apply lots of comments to the file in fn, making a new .newdoc file.
 183     """
 184     N = 0
 185
 186     lines = [ None ]
 187     try:
 188         lines.extend( open(fn, 'r').readlines() )
 189     except IOError:
 190         return
 191
 192     # Process the comments in reverse order by line number, so that
 193     # the line numbers for the ones we haven't added yet remain valid
 194     # until we add them.  Standard trick.
 195     entries.sort()
 196     entries.reverse()
 197
 198     for ln, kind, name in entries:
 199
 200         lines.insert(ln, "/* DOCDOC %s */\n"%name)
 201         N += 1
 202
 203     outf = open(fn+".newdoc", 'w')
 204     for line in lines[1:]:
 205         outf.write(line)
 206     outf.close()
 207
 208     print "Added %s DOCDOCs to %s" %(N, fn)
 209
 210 e = read()
 211
 212 for fn, errs in e.iteritems():
 213     comments = checkf(fn, errs)
 214     if comments:
 215         applyComments(fn, comments)