/* Support for Robot Exclusion Standard (RES).
   Copyright (C) 2001 Free Software Foundation, Inc.

This file is part of Wget.

This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or (at
your option) any later version.

This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.  */

/* This file implements the Robot Exclusion Standard (RES).

   RES is a simple protocol that enables site admins to signalize to
   the web crawlers that certain parts of the site should not be
   accessed.  All the admin needs to do is create a "robots.txt" file
   in the web server root, and use simple commands to allow or
   disallow access to certain parts of the site.

   The first specification was written by Martijn Koster in 1994, and
   is still available at <http://www.robotstxt.org/wc/norobots.html>.
   In 1996, Martijn wrote an Internet Draft specifying an improved RES
   specification; however, that work was apparently abandoned since
   the draft has expired in 1997 and hasn't been replaced since.  The
   draft is available at
   <http://www.robotstxt.org/wc/norobots-rfc.html>.

   This file implements RES as specified by the draft.  Note that this
   only handles the "robots.txt" support.  The META tag that controls
   whether the links should be followed is handled in `html-url.c'.

   Known deviations:

   * The end-of-line comment recognition is more in the spirit of the
     Bourne Shell (as specified by RES-1994).  That means that
     "foo#bar" is taken literally, whereas "foo #bar" is interpreted
     as "foo".  The Draft apparently specifies that both should be
     interpreted as "foo".

   * We don't recognize sole CR as the line ending.

   * We don't implement expiry mechanism for /robots.txt specs.  I
     consider it non-necessary for a relatively short-lived
     application such as Wget.  Besides, it is highly questionable
     whether anyone deploys the recommended expiry scheme for
     robots.txt.

   Entry points are functions res_parse, res_parse_from_file,
   res_match_path, res_register_specs, res_get_specs, and
   res_retrieve_file.  */

#ifdef HAVE_CONFIG_H
# include <config.h>
#endif

#include <stdio.h>
#include <stdlib.h>
#ifdef HAVE_STRING_H
# include <string.h>
#else
# include <strings.h>
#endif /* HAVE_STRING_H */
#include <errno.h>
#include <assert.h>

#include "wget.h"
#include "utils.h"
#include "hash.h"
#include "url.h"
#include "retr.h"
#include "res.h"

struct path_info {
  char *path;
  int allowedp;
  int user_agent_exact_p;
};

struct robot_specs {
  int count;
  int size;
  struct path_info *paths;
};

/* Parsing the robot spec. */

/* Check whether AGENT (a string of length LENGTH) equals "wget" or
   "*".  If it is either of them, *matches is set to one.  If it is
   "wget", *exact_match is set to one.  */

static void
match_user_agent (const char *agent, int length,
		  int *matches, int *exact_match)
{
  if (length == 1 && *agent == '*')
    {
      *matches = 1;
      *exact_match = 0;
    }
  else if (BOUNDED_EQUAL_NO_CASE (agent, agent + length, "wget"))
    {
      *matches = 1;
      *exact_match = 1;
    }
  else
    {
      *matches = 0;
      *exact_match = 0;
    }
}

/* Add a path specification between PATH_B and PATH_E as one of the
   paths in SPECS.  */

static void
add_path (struct robot_specs *specs, const char *path_b, const char *path_e,
	  int allowedp, int exactp)
{
  struct path_info pp;
  if (path_b < path_e && *path_b == '/')
    /* Our path representation doesn't use a leading slash, so remove
       one from theirs. */
    ++path_b;
  pp.path     = strdupdelim (path_b, path_e);
  pp.allowedp = allowedp;
  pp.user_agent_exact_p = exactp;
  ++specs->count;
  if (specs->count > specs->size)
    {
      if (specs->size == 0)
	specs->size = 1;
      else
	specs->size <<= 1;
      specs->paths = xrealloc (specs->paths,
			       specs->size * sizeof (struct path_info));
    }
  specs->paths[specs->count - 1] = pp;
}

/* Recreate SPECS->paths with only those paths that have non-zero
   user_agent_exact_p.  */

static void
prune_non_exact (struct robot_specs *specs)
{
  struct path_info *newpaths;
  int i, j, cnt;
  cnt = 0;
  for (i = 0; i < specs->count; i++)
    if (specs->paths[i].user_agent_exact_p)
      ++cnt;
  newpaths = xmalloc (cnt * sizeof (struct path_info));
  for (i = 0, j = 0; i < specs->count; i++)
    if (specs->paths[i].user_agent_exact_p)
      newpaths[j++] = specs->paths[i];
  assert (j == cnt);
  xfree (specs->paths);
  specs->paths = newpaths;
  specs->count = cnt;
  specs->size  = cnt;
}

#define EOL(p) ((p) >= lineend)

#define SKIP_SPACE(p) do {		\
  while (!EOL (p) && ISSPACE (*p))	\
    ++p;				\
} while (0)

#define FIELD_IS(string_literal)	\
  BOUNDED_EQUAL_NO_CASE (field_b, field_e, string_literal)

/* Parse textual RES specs beginning with SOURCE of length LENGTH.
   Return a specs objects ready to be fed to res_match_path.

   The parsing itself is trivial, but creating a correct SPECS object
   is trickier than it seems, because RES is surprisingly byzantine if
   you attempt to implement it correctly.

   A "record" is a block of one or more `User-Agent' lines followed by
   one or more `Allow' or `Disallow' lines.  Record is accepted by
   Wget if one of the `User-Agent' lines was "wget", or if the user
   agent line was "*".

   After all the lines have been read, we examine whether an exact
   ("wget") user-agent field was specified.  If so, we delete all the
   lines read under "User-Agent: *" blocks because we have our own
   Wget-specific blocks.  This enables the admin to say:

       User-Agent: *
       Disallow: /

       User-Agent: google
       User-Agent: wget
       Disallow: /cgi-bin

   This means that to Wget and to Google, /cgi-bin is disallowed,
   whereas for all other crawlers, everything is disallowed.
   res_parse is implemented so that the order of records doesn't
   matter.  In the case above, the "User-Agent: *" could have come
   after the other one.  */

struct robot_specs *
res_parse (const char *source, int length)
{
  int line_count = 1;

  const char *p   = source;
  const char *end = source + length;

  /* non-zero if last applicable user-agent field matches Wget. */
  int user_agent_applies = 0;

  /* non-zero if last applicable user-agent field *exactly* matches
     Wget.  */
  int user_agent_exact = 0;

  /* whether we ever encountered exact user agent. */
  int found_exact = 0;

  /* count of allow/disallow lines in the current "record", i.e. after
     the last `user-agent' instructions.  */
  int record_count = 0;

  struct robot_specs *specs = xmalloc (sizeof (struct robot_specs));
  memset (specs, '\0', sizeof (struct robot_specs));

  while (1)
    {
      const char *lineend, *lineend_real;
      const char *field_b, *field_e;
      const char *value_b, *value_e;

      if (p == end)
	break;
      lineend_real = memchr (p, '\n', end - p);
      if (lineend_real)
	++lineend_real;
      else
	lineend_real = end;
      lineend = lineend_real;

      /* Before doing anything else, check whether the line is empty
	 or comment-only. */
      SKIP_SPACE (p);
      if (EOL (p) || *p == '#')
	goto next;

      /* Make sure the end-of-line comments are respected by setting
	 lineend to a location preceding the first comment.  Real line
	 ending remains in lineend_real.  */
      for (lineend = p; lineend < lineend_real; lineend++)
	if ((lineend == p || ISSPACE (*(lineend - 1)))
	    && *lineend == '#')
	  break;

      /* Ignore trailing whitespace in the same way. */
      while (lineend > p && ISSPACE (*(lineend - 1)))
	--lineend;

      assert (!EOL (p));

      field_b = p;
      while (!EOL (p) && (ISALNUM (*p) || *p == '-'))
	++p;
      field_e = p;

      SKIP_SPACE (p);
      if (field_b == field_e || EOL (p) || *p != ':')
	{
	  DEBUGP (("Ignoring malformed line %d", line_count));
	  goto next;
	}
      ++p;			/* skip ':' */
      SKIP_SPACE (p);

      value_b = p;
      while (!EOL (p))
	++p;
      value_e = p;

      /* Finally, we have a syntactically valid line. */
      if (FIELD_IS ("user-agent"))
	{
	  /* We have to support several cases:

	     --previous records--

	     User-Agent: foo
	     User-Agent: Wget
	     User-Agent: bar
	     ... matching record ...

	     User-Agent: baz
	     User-Agent: qux
	     ... non-matching record ...

	     User-Agent: *
	     ... matching record, but will be pruned later ...

	     We have to respect `User-Agent' at the beginning of each
	     new record simply because we don't know if we're going to
	     encounter "Wget" among the agents or not.  Hence,
	     match_user_agent is called when record_count != 0.

	     But if record_count is 0, we have to keep calling it
	     until it matches, and if that happens, we must not call
	     it any more, until the next record.  Hence the other part
	     of the condition.  */
	  if (record_count != 0 || user_agent_applies == 0)
	    match_user_agent (value_b, value_e - value_b,
			      &user_agent_applies, &user_agent_exact);
	  if (user_agent_exact)
	    found_exact = 1;
	  record_count = 0;
	}
      else if (FIELD_IS ("allow"))
	{
	  if (user_agent_applies)
	    {
	      add_path (specs, value_b, value_e, 1, user_agent_exact);
	    }
	  ++record_count;
	}
      else if (FIELD_IS ("disallow"))
	{
	  if (user_agent_applies)
	    {
	      int allowed = 0;
	      if (value_b == value_e)
		/* Empty "disallow" line means everything is
		   *allowed*!  */
		allowed = 1;
	      add_path (specs, value_b, value_e, allowed, user_agent_exact);
	    }
	  ++record_count;
	}
      else
	{
	  DEBUGP (("Ignoring unknown field at line %d", line_count));
	  goto next;
	}

    next:
      p = lineend_real;
      ++line_count;
    }

  if (found_exact)
    {
      /* We've encountered an exactly matching user-agent.  Throw out
	 all the stuff with user-agent: *.  */
      prune_non_exact (specs);
    }
  else if (specs->size > specs->count)
    {
      /* add_path normally over-allocates specs->paths.  Reallocate it
	 to the correct size in order to conserve some memory.  */
      specs->paths = xrealloc (specs->paths,
			       specs->count * sizeof (struct path_info));
      specs->size = specs->count;
    }

  return specs;
}

/* The same like res_parse, but first map the FILENAME into memory,
   and then parse it.  */

struct robot_specs *
res_parse_from_file (const char *filename)
{
  struct robot_specs *specs;
  struct file_memory *fm = read_file (filename);
  if (!fm)
    {
      logprintf (LOG_NOTQUIET, "Cannot open %s: %s",
		 filename, strerror (errno));
      return NULL;
    }
  specs = res_parse (fm->content, fm->length);
  read_file_free (fm);
  return specs;
}

static void
free_specs (struct robot_specs *specs)
{
  int i;
  for (i = 0; i < specs->count; i++)
    xfree (specs->paths[i].path);
  FREE_MAYBE (specs->paths);
  xfree (specs);
}

/* Matching of a path according to the specs. */

/* If C is '%' and (ptr[1], ptr[2]) form a hexadecimal number, and if
   that number is not a numerical representation of '/', decode C and
   advance the pointer.  */

#define DECODE_MAYBE(c, ptr) do {					\
  if (c == '%' && ISXDIGIT (ptr[1]) && ISXDIGIT (ptr[2]))		\
    {									\
      char decoded							\
	= (XCHAR_TO_XDIGIT (ptr[1]) << 4) + XCHAR_TO_XDIGIT (ptr[2]);	\
      if (decoded != '/')						\
	{								\
	  c = decoded;							\
	  ptr += 2;							\
	}								\
    }									\
} while (0)

/* The inner matching engine: return non-zero if RECORD_PATH matches
   URL_PATH.  The rules for matching are described at
   <http://info.webcrawler.com/mak/projects/robots/norobots-rfc.html>,
   section 3.2.2.  */

static int
matches (const char *record_path, const char *url_path)
{
  const char *rp = record_path;
  const char *up = url_path;

  for (; ; ++rp, ++up)
    {
      char rc = *rp;
      char uc = *up;
      if (!rc)
	return 1;
      if (!uc)
	return 0;
      DECODE_MAYBE(rc, rp);
      DECODE_MAYBE(uc, up);
      if (rc != uc)
	return 0;
    }
}

/* Iterate through all paths in SPECS.  For the first one that
   matches, return its allow/reject status.  If none matches,
   retrieval is by default allowed.  */

int
res_match_path (const struct robot_specs *specs, const char *path)
{
  int i;
  if (!specs)
    return 1;
  for (i = 0; i < specs->count; i++)
    if (matches (specs->paths[i].path, path))
      {
	int allowedp = specs->paths[i].allowedp;
	DEBUGP (("%s path %s because of rule `%s'.\n",
		 allowedp ? "Allowing" : "Rejecting",
		 path, specs->paths[i].path));
	return allowedp;
      }
  return 1;
}

/* Registering the specs. */

static struct hash_table *registered_specs;

/* Stolen from cookies.c. */
#define SET_HOSTPORT(host, port, result) do {		\
  int HP_len = strlen (host);				\
  result = alloca (HP_len + 1 + numdigit (port) + 1);	\
  memcpy (result, host, HP_len);			\
  result[HP_len] = ':';					\
  long_to_string (result + HP_len + 1, port);		\
} while (0)

/* Register RES specs that below to server on HOST:PORT.  They will
   later be retrievable using res_get_specs.  */

void
res_register_specs (const char *host, int port, struct robot_specs *specs)
{
  struct robot_specs *old;
  char *hp, *hp_old;
  SET_HOSTPORT (host, port, hp);

  if (!registered_specs)
    registered_specs = make_nocase_string_hash_table (0);

  /* Required to shut up the compiler. */
  old    = NULL;
  hp_old = NULL;

  if (hash_table_get_pair (registered_specs, hp, hp_old, old))
    {
      if (old)
	free_specs (old);
      hash_table_put (registered_specs, hp_old, specs);
    }
  else
    {
      hash_table_put (registered_specs, xstrdup (hp), specs);
    }
}

/* Get the specs that belong to HOST:PORT. */

struct robot_specs *
res_get_specs (const char *host, int port)
{
  char *hp;
  SET_HOSTPORT (host, port, hp);
  if (!registered_specs)
    return NULL;
  return hash_table_get (registered_specs, hp);
}

/* Loading the robots file.  */

#define RES_SPECS_LOCATION "/robots.txt"

/* Retrieve the robots.txt from the server root of the server that
   serves URL.  The file will be named according to the currently
   active rules, and the file name will be returned in *file.

   Return non-zero if robots were retrieved OK, zero otherwise.  */

int
res_retrieve_file (const char *url, char **file)
{
  uerr_t err;
  char *robots_url = uri_merge (url, RES_SPECS_LOCATION);

  logputs (LOG_VERBOSE, _("Loading robots.txt; please ignore errors.\n"));
  *file = NULL;
  err = retrieve_url (robots_url, file, NULL, NULL, NULL);
  xfree (robots_url);

  if (err != RETROK && *file != NULL)
    {
      /* If the file is not retrieved correctly, but retrieve_url
	 allocated the file name, deallocate is here so that the
	 caller doesn't have to worry about it.  */
      xfree (*file);
      *file = NULL;
    }
  return err == RETROK;
}

static int
cleanup_hash_table_mapper (void *key, void *value, void *arg_ignored)
{
  xfree (key);
  free_specs (value);
  return 0;
}

void
res_cleanup (void)
{
  if (registered_specs)
    {
      hash_table_map (registered_specs, cleanup_hash_table_mapper, NULL);
      hash_table_destroy (registered_specs);
      registered_specs = NULL;
    }
}