scripts/maint/redox.py

   1 #!/usr/bin/python
   2 #
   3 #  Copyright (c) 2008-2017, 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 ./scripts/maint/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 ./scripts/maint/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                "strlcat.c",
  37                "strlcpy.c",
  38                "sha256.c",
  39                "sha256.h",
  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           "enumeration" ]
  60
  61 NODOC_LINE_RE = re.compile(r'^([^:]+):(\d+): (\w+): (.*) is not documented\.$')
  62
  63 THING_RE = re.compile(r'^Member ([a-zA-Z0-9_]+).*\((typedef|define|function|variable|enumeration|macro definition)\) of (file|class) ')
  64
  65 SKIP_NAMES = [re.compile(s) for s in SKIP_NAME_PATTERNS]
  66
  67 def parsething(thing):
  68     """I figure out what 'foobar baz in quux quum is not documented' means,
  69        and return: the name of the foobar, and the kind of the foobar.
  70     """
  71     if thing.startswith("Compound "):
  72         tp, name = "type", thing.split()[1]
  73     else:
  74         m = THING_RE.match(thing)
  75         if not m:
  76             print thing, "???? Format didn't match."
  77             return None, None
  78         else:
  79             name, tp, parent = m.groups()
  80             if parent == 'class':
  81                 if tp == 'variable' or tp == 'function':
  82                     tp = 'field'
  83
  84     return name, tp
  85
  86 def read():
  87     """I snarf doxygen stderr from stdin, and parse all the "foo has no
  88        documentation messages.  I return a map from filename to lists
  89        of tuples of (alleged line number, name of thing, kind of thing)
  90     """
  91     errs = {}
  92     for line in sys.stdin:
  93         m = NODOC_LINE_RE.match(line)
  94         if m:
  95             file, line, tp, thing = m.groups()
  96             assert tp.lower() == 'warning'
  97             name, kind = parsething(thing)
  98             errs.setdefault(file, []).append((int(line), name, kind))
  99
 100     return errs
 101
 102 def findline(lines, lineno, ident):
 103     """Given a list of all the lines in the file (adjusted so 1-indexing works),
 104        a line number that ident is allegedly on, and ident, I figure out
 105        the line where ident was really declared."""
 106     lno = lineno
 107     for lineno in xrange(lineno, 0, -1):
 108         try:
 109             if ident in lines[lineno]:
 110                 return lineno
 111         except IndexError:
 112             continue
 113
 114     return None
 115
 116 FUNC_PAT = re.compile(r"^[A-Za-z0-9_]+\(")
 117
 118 def hascomment(lines, lineno, kind):
 119     """I return true if it looks like there's already a good comment about
 120        the thing on lineno of lines of type kind. """
 121     if "*/" in lines[lineno-1]:
 122         return True
 123     if kind == 'function' and FUNC_PAT.match(lines[lineno]):
 124         if "*/" in lines[lineno-2]:
 125             return True
 126     return False
 127
 128 def hasdocdoc(lines, lineno, kind):
 129     """I return true if it looks like there's already a docdoc comment about
 130        the thing on lineno of lines of type kind."""
 131     try:
 132         if "DOCDOC" in lines[lineno]:
 133             return True
 134     except IndexError:
 135         pass
 136     try:
 137         if "DOCDOC" in lines[lineno-1]:
 138             return True
 139     except IndexError:
 140         pass
 141     if kind == 'function' and FUNC_PAT.match(lines[lineno]):
 142         if "DOCDOC" in lines[lineno-2]:
 143             return True
 144     return False
 145
 146 def checkf(fn, errs):
 147     """I go through the output of read() for a single file, and build a list
 148        of tuples of things that want DOCDOC comments.  Each tuple has:
 149        the line number where the comment goes; the kind of thing; its name.
 150     """
 151     for skip in SKIP_FILES:
 152         if fn.endswith(skip):
 153             print "Skipping",fn
 154             return
 155
 156     comments = []
 157     lines = [ None ]
 158     try:
 159         lines.extend( open(fn, 'r').readlines() )
 160     except IOError:
 161         return
 162
 163     for line, name, kind in errs:
 164         if any(pat.match(name) for pat in SKIP_NAMES):
 165             continue
 166
 167         if kind not in ADD_DOCDOCS_TO_TYPES:
 168             continue
 169
 170         ln = findline(lines, line, name)
 171         if ln == None:
 172             print "Couldn't find the definition of %s allegedly on %s of %s"%(
 173                 name, line, fn)
 174         else:
 175             if hasdocdoc(lines, line, kind):
 176 #                print "Has a DOCDOC"
 177 #                print fn, line, name, kind
 178 #                print "\t",lines[line-2],
 179 #                print "\t",lines[line-1],
 180 #                print "\t",lines[line],
 181 #                print "-------"
 182                 pass
 183             else:
 184                 if kind == 'function' and FUNC_PAT.match(lines[ln]):
 185                     ln = ln - 1
 186
 187                 comments.append((ln, kind, name))
 188
 189     return comments
 190
 191 def applyComments(fn, entries):
 192     """I apply lots of comments to the file in fn, making a new .newdoc file.
 193     """
 194     N = 0
 195
 196     lines = [ None ]
 197     try:
 198         lines.extend( open(fn, 'r').readlines() )
 199     except IOError:
 200         return
 201
 202     # Process the comments in reverse order by line number, so that
 203     # the line numbers for the ones we haven't added yet remain valid
 204     # until we add them.  Standard trick.
 205     entries.sort()
 206     entries.reverse()
 207
 208     for ln, kind, name in entries:
 209
 210         lines.insert(ln, "/* DOCDOC %s */\n"%name)
 211         N += 1
 212
 213     outf = open(fn+".newdoc", 'w')
 214     for line in lines[1:]:
 215         outf.write(line)
 216     outf.close()
 217
 218     print "Added %s DOCDOCs to %s" %(N, fn)
 219
 220 e = read()
 221
 222 for fn, errs in e.iteritems():
 223     print `(fn, errs)`
 224     comments = checkf(fn, errs)
 225     if comments:
 226         applyComments(fn, comments)