#! /usr/bin/env python

# Original code by Guido van Rossum; extensive changes by Sam Bayer,

# including code to check URL fragments.

"""Web tree checker.

This utility is handy to check a subweb of the world-wide web for

errors.  A subweb is specified by giving one or more ``root URLs''; a

page belongs to the subweb if one of the root URLs is an initial

prefix of it.

File URL extension:

In order to easy the checking of subwebs via the local file system,

the interpretation of ``file:'' URLs is extended to mimic the behavior

of your average HTTP daemon: if a directory pathname is given, the

file index.html in that directory is returned if it exists, otherwise

a directory listing is returned.  Now, you can point webchecker to the

document tree in the local file system of your HTTP daemon, and have

most of it checked.  In fact the default works this way if your local

web tree is located at /usr/local/etc/httpd/htdpcs (the default for

the NCSA HTTP daemon and probably others).

Report printed:

When done, it reports pages with bad links within the subweb.  When

interrupted, it reports for the pages that it has checked already.

In verbose mode, additional messages are printed during the

information gathering phase.  By default, it prints a summary of its

work status every 50 URLs (adjustable with the -r option), and it

reports errors as they are encountered.  Use the -q option to disable

this output.

Checkpoint feature:

Whether interrupted or not, it dumps its state (a Python pickle) to a

checkpoint file and the -R option allows it to restart from the

checkpoint (assuming that the pages on the subweb that were already

processed haven't changed).  Even when it has run till completion, -R

can still be useful -- it will print the reports again, and -Rq prints

the errors only.  In this case, the checkpoint file is not written

again.  The checkpoint file can be set with the -d option.

The checkpoint file is written as a Python pickle.  Remember that

Python's pickle module is currently quite slow.  Give it the time it

needs to load and save the checkpoint file.  When interrupted while

writing the checkpoint file, the old checkpoint file is not

overwritten, but all work done in the current run is lost.

Miscellaneous:

- You may find the (Tk-based) GUI version easier to use.  See wcgui.py.

- Webchecker honors the "robots.txt" convention.  Thanks to Skip

Montanaro for his robotparser.py module (included in this directory)!

The agent name is hardwired to "webchecker".  URLs that are disallowed

by the robots.txt file are reported as external URLs.

- Because the SGML parser is a bit slow, very large SGML files are

skipped.  The size limit can be set with the -m option.

- When the server or protocol does not tell us a file's type, we guess

it based on the URL's suffix.  The mimetypes.py module (also in this

directory) has a built-in table mapping most currently known suffixes,

and in addition attempts to read the mime.types configuration files in

the default locations of Netscape and the NCSA HTTP daemon.

- We follow links indicated by <A>, <FRAME> and <IMG> tags.  We also

honor the <BASE> tag.

- We now check internal NAME anchor links, as well as toplevel links.

- Checking external links is now done by default; use -x to *disable*

this feature.  External links are now checked during normal

processing.  (XXX The status of a checked link could be categorized

better.  Later...)

- If external links are not checked, you can use the -t flag to

provide specific overrides to -x.

Usage: webchecker.py [option] ... [rooturl] ...

Options:

-R        -- restart from checkpoint file

-d file   -- checkpoint filename (default %(DUMPFILE)s)

-m bytes  -- skip HTML pages larger than this size (default %(MAXPAGE)d)

-n        -- reports only, no checking (use with -R)

-q        -- quiet operation (also suppresses external links report)

-r number -- number of links processed per round (default %(ROUNDSIZE)d)

-t root   -- specify root dir which should be treated as internal (can repeat)

-v        -- verbose operation; repeating -v will increase verbosity

-x        -- don't check external links (these are often slow to check)

-a        -- don't check name anchors

Arguments:

rooturl   -- URL to start checking

             (default %(DEFROOT)s)

"""

__version__ = "$Revision: 50851 $"

import sys

import os

from types import *

import StringIO

import getopt

import pickle

import urllib

import urlparse

import sgmllib

import cgi

import mimetypes

import robotparser

# Extract real version number if necessary

if __version__[0] == '$':

    _v = __version__.split()

    if len(_v) == 3:

        __version__ = _v[1]

# Tunable parameters

DEFROOT = "file:/usr/local/etc/httpd/htdocs/"   # Default root URL

CHECKEXT = 1                            # Check external references (1 deep)

VERBOSE = 1                             # Verbosity level (0-3)

MAXPAGE = 150000                        # Ignore files bigger than this

ROUNDSIZE = 50                          # Number of links processed per round

DUMPFILE = "@webchecker.pickle"         # Pickled checkpoint

AGENTNAME = "webchecker"                # Agent name for robots.txt parser

NONAMES = 0                             # Force name anchor checking

# Global variables

def main():

    checkext = CHECKEXT

    verbose = VERBOSE

    maxpage = MAXPAGE

    roundsize = ROUNDSIZE

    dumpfile = DUMPFILE

    restart = 0

    norun = 0

    try:

        opts, args = getopt.getopt(sys.argv[1:], 'Rd:m:nqr:t:vxa')

    except getopt.error, msg:

        sys.stdout = sys.stderr

        print msg

        print __doc__%globals()

        sys.exit(2)

    # The extra_roots variable collects extra roots.

    extra_roots = []

    nonames = NONAMES

    for o, a in opts:

        if o == '-R':

            restart = 1

        if o == '-d':

            dumpfile = a

        if o == '-m':

            maxpage = int(a)

        if o == '-n':

            norun = 1

        if o == '-q':

            verbose = 0

        if o == '-r':

            roundsize = int(a)

        if o == '-t':

            extra_roots.append(a)

        if o == '-a':

            nonames = not nonames

        if o == '-v':

            verbose = verbose + 1

        if o == '-x':

            checkext = not checkext

    if verbose > 0:

        print AGENTNAME, "version", __version__

    if restart:

        c = load_pickle(dumpfile=dumpfile, verbose=verbose)

    else:

        c = Checker()

    c.setflags(checkext=checkext, verbose=verbose,

               maxpage=maxpage, roundsize=roundsize,

               nonames=nonames

               )

    if not restart and not args:

        args.append(DEFROOT)

    for arg in args:

        c.addroot(arg)

    # The -t flag is only needed if external links are not to be

    # checked. So -t values are ignored unless -x was specified.

    if not checkext:

        for root in extra_roots:

            # Make sure it's terminated by a slash,

            # so that addroot doesn't discard the last

            # directory component.

            if root[-1] != "/":

                root = root + "/"

            c.addroot(root, add_to_do = 0)

    try:

        if not norun:

            try:

                c.run()

            except KeyboardInterrupt:

                if verbose > 0:

                    print "[run interrupted]"

        try:

            c.report()

        except KeyboardInterrupt:

            if verbose > 0:

                print "[report interrupted]"

    finally:

        if c.save_pickle(dumpfile):

            if dumpfile == DUMPFILE:

                print "Use ``%s -R'' to restart." % sys.argv[0]

            else:

                print "Use ``%s -R -d %s'' to restart." % (sys.argv[0],

                                                           dumpfile)

def load_pickle(dumpfile=DUMPFILE, verbose=VERBOSE):

    if verbose > 0:

        print "Loading checkpoint from %s ..." % dumpfile

    f = open(dumpfile, "rb")

    c = pickle.load(f)

    f.close()

    if verbose > 0:

        print "Done."

        print "Root:", "\n      ".join(c.roots)

    return c

class Checker:

    checkext = CHECKEXT

    verbose = VERBOSE

    maxpage = MAXPAGE

    roundsize = ROUNDSIZE

    nonames = NONAMES

    validflags = tuple(dir())

    def __init__(self):

        self.reset()

    def setflags(self, **kw):

        for key in kw.keys():

            if key not in self.validflags:

                raise NameError, "invalid keyword argument: %s" % str(key)

        for key, value in kw.items():

            setattr(self, key, value)

    def reset(self):

        self.roots = []

        self.todo = {}

        self.done = {}

        self.bad = {}

        # Add a name table, so that the name URLs can be checked. Also

        # serves as an implicit cache for which URLs are done.

        self.name_table = {}

        self.round = 0

        # The following are not pickled:

        self.robots = {}

        self.errors = {}

        self.urlopener = MyURLopener()

        self.changed = 0

    def note(self, level, format, *args):

        if self.verbose > level:

            if args:

                format = format%args

            self.message(format)

    def message(self, format, *args):

        if args:

            format = format%args

        print format

    def __getstate__(self):

        return (self.roots, self.todo, self.done, self.bad, self.round)

    def __setstate__(self, state):

        self.reset()

        (self.roots, self.todo, self.done, self.bad, self.round) = state

        for root in self.roots:

            self.addrobot(root)

        for url in self.bad.keys():

            self.markerror(url)

    def addroot(self, root, add_to_do = 1):

        if root not in self.roots:

            troot = root

            scheme, netloc, path, params, query, fragment = \

                    urlparse.urlparse(root)

            i = path.rfind("/") + 1

            if 0 < i < len(path):

                path = path[:i]

                troot = urlparse.urlunparse((scheme, netloc, path,

                                             params, query, fragment))

            self.roots.append(troot)

            self.addrobot(root)

            if add_to_do:

                self.newlink((root, ""), ("<root>", root))

    def addrobot(self, root):

        root = urlparse.urljoin(root, "/")

        if self.robots.has_key(root): return

        url = urlparse.urljoin(root, "/robots.txt")

        self.robots[root] = rp = robotparser.RobotFileParser()

        self.note(2, "Parsing %s", url)

        rp.debug = self.verbose > 3

        rp.set_url(url)

        try:

            rp.read()

        except (OSError, IOError), msg:

            self.note(1, "I/O error parsing %s: %s", url, msg)

    def run(self):

        while self.todo:

            self.round = self.round + 1

            self.note(0, "\nRound %d (%s)\n", self.round, self.status())

            urls = self.todo.keys()

            urls.sort()

            del urls[self.roundsize:]

            for url in urls:

                self.dopage(url)

    def status(self):

        return "%d total, %d to do, %d done, %d bad" % (

            len(self.todo)+len(self.done),

            len(self.todo), len(self.done),

            len(self.bad))

    def report(self):

        self.message("")

        if not self.todo: s = "Final"

        else: s = "Interim"

        self.message("%s Report (%s)", s, self.status())

        self.report_errors()

    def report_errors(self):

        if not self.bad:

            self.message("\nNo errors")

            return

        self.message("\nError Report:")

        sources = self.errors.keys()

        sources.sort()

        for source in sources:

            triples = self.errors[source]

            self.message("")

            if len(triples) > 1:

                self.message("%d Errors in %s", len(triples), source)

            else:

                self.message("Error in %s", source)

            # Call self.format_url() instead of referring

            # to the URL directly, since the URLs in these

            # triples is now a (URL, fragment) pair. The value

            # of the "source" variable comes from the list of

            # origins, and is a URL, not a pair.

            for url, rawlink, msg in triples:

                if rawlink != self.format_url(url): s = " (%s)" % rawlink

                else: s = ""

                self.message("  HREF %s%s\n    msg %s",

                             self.format_url(url), s, msg)

    def dopage(self, url_pair):

        # All printing of URLs uses format_url(); argument changed to

        # url_pair for clarity.

        if self.verbose > 1:

            if self.verbose > 2:

                self.show("Check ", self.format_url(url_pair),

                          "  from", self.todo[url_pair])

            else:

                self.message("Check %s", self.format_url(url_pair))

        url, local_fragment = url_pair

        if local_fragment and self.nonames:

            self.markdone(url_pair)

            return

        try:

            page = self.getpage(url_pair)

        except sgmllib.SGMLParseError, msg:

            msg = self.sanitize(msg)

            self.note(0, "Error parsing %s: %s",

                          self.format_url(url_pair), msg)

            # Dont actually mark the URL as bad - it exists, just

            # we can't parse it!

            page = None

        if page:

            # Store the page which corresponds to this URL.

            self.name_table[url] = page

            # If there is a fragment in this url_pair, and it's not

            # in the list of names for the page, call setbad(), since

            # it's a missing anchor.

            if local_fragment and local_fragment not in page.getnames():

                self.setbad(url_pair, ("Missing name anchor `%s'" % local_fragment))

            for info in page.getlinkinfos():

                # getlinkinfos() now returns the fragment as well,

                # and we store that fragment here in the "todo" dictionary.

                link, rawlink, fragment = info

                # However, we don't want the fragment as the origin, since

                # the origin is logically a page.

                origin = url, rawlink

                self.newlink((link, fragment), origin)

        else:

            # If no page has been created yet, we want to

            # record that fact.

            self.name_table[url_pair[0]] = None

        self.markdone(url_pair)

    def newlink(self, url, origin):

        if self.done.has_key(url):

            self.newdonelink(url, origin)

        else:

            self.newtodolink(url, origin)

    def newdonelink(self, url, origin):

        if origin not in self.done[url]:

            self.done[url].append(origin)

        # Call self.format_url(), since the URL here

        # is now a (URL, fragment) pair.

        self.note(3, "  Done link %s", self.format_url(url))

        # Make sure that if it's bad, that the origin gets added.

        if self.bad.has_key(url):

            source, rawlink = origin

            triple = url, rawlink, self.bad[url]

            self.seterror(source, triple)

    def newtodolink(self, url, origin):

        # Call self.format_url(), since the URL here

        # is now a (URL, fragment) pair.

        if self.todo.has_key(url):

            if origin not in self.todo[url]:

                self.todo[url].append(origin)

            self.note(3, "  Seen todo link %s", self.format_url(url))

        else:

            self.todo[url] = [origin]

            self.note(3, "  New todo link %s", self.format_url(url))

    def format_url(self, url):

        link, fragment = url

        if fragment: return link + "#" + fragment

        else: return link

    def markdone(self, url):

        self.done[url] = self.todo[url]

        del self.todo[url]

        self.changed = 1

    def inroots(self, url):

        for root in self.roots:

            if url[:len(root)] == root:

                return self.isallowed(root, url)

        return 0

    def isallowed(self, root, url):

        root = urlparse.urljoin(root, "/")

        return self.robots[root].can_fetch(AGENTNAME, url)

    def getpage(self, url_pair):

        # Incoming argument name is a (URL, fragment) pair.

        # The page may have been cached in the name_table variable.

        url, fragment = url_pair

        if self.name_table.has_key(url):

            return self.name_table[url]

        scheme, path = urllib.splittype(url)

        if scheme in ('mailto', 'news', 'javascript', 'telnet'):

            self.note(1, " Not checking %s URL" % scheme)

            return None

        isint = self.inroots(url)

        # Ensure that openpage gets the URL pair to

        # print out its error message and record the error pair

        # correctly.

        if not isint:

            if not self.checkext:

                self.note(1, " Not checking ext link")

                return None

            f = self.openpage(url_pair)

            if f:

                self.safeclose(f)

            return None

        text, nurl = self.readhtml(url_pair)

        if nurl != url:

            self.note(1, " Redirected to %s", nurl)

            url = nurl

        if text:

            return Page(text, url, maxpage=self.maxpage, checker=self)

    # These next three functions take (URL, fragment) pairs as

    # arguments, so that openpage() receives the appropriate tuple to

    # record error messages.

    def readhtml(self, url_pair):

        url, fragment = url_pair

        text = None

        f, url = self.openhtml(url_pair)

        if f:

            text = f.read()

            f.close()

        return text, url