bottle.py

# -*- coding: utf-8 -*-
"""
Bottle is a fast and simple micro-framework for small web applications. It
offers request dispatching (Routes) with url parameter support, templates,
a built-in HTTP Server and adapters for many third party WSGI/HTTP-server and
template engines - all in a single file and with no dependencies other than the
Python Standard Library.

Homepage and documentation: http://bottlepy.org/

Copyright (c) 2011, Marcel Hellkamp.
License: MIT (see LICENSE.txt for details)
"""

from __future__ import with_statement

__author__ = 'Marcel Hellkamp'
__version__ = '0.10.dev'
__license__ = 'MIT'

import base64
import cgi
import email.utils
import functools
import hmac
import httplib
import imp
import itertools
import mimetypes
import os
import re
import subprocess
import sys
import tempfile
import thread
import threading
import time
import warnings

from Cookie import SimpleCookie
from tempfile import TemporaryFile
from traceback import format_exc
from urlparse import urljoin, SplitResult as UrlSplitResult
from types import MethodType

# Workaround for a bug in some versions of lib2to3 (fixed on CPython 2.7 and 3.2)
import urllib
urlencode = urllib.urlencode
urlquote = urllib.quote
urlunquote = urllib.unquote

try: from collections import MutableMapping as DictMixin
except ImportError: # pragma: no cover
    from UserDict import DictMixin

try: from urlparse import parse_qs
except ImportError: # pragma: no cover
    from cgi import parse_qs

try: import cPickle as pickle
except ImportError: # pragma: no cover
    import pickle

try: from json import dumps as json_dumps, loads as json_lds
except ImportError: # pragma: no cover
    try: from simplejson import dumps as json_dumps, loads as json_lds
    except ImportError: # pragma: no cover
        try: from django.utils.simplejson import dumps as json_dumps, loads as json_lds
        except ImportError: # pragma: no cover
            def json_dumps(data):
                raise ImportError("JSON support requires Python 2.6 or simplejson.")
            json_lds = json_dumps

py3k = sys.version_info >= (3,0,0)
NCTextIOWrapper = None

if py3k: # pragma: no cover
    json_loads = lambda s: json_lds(touni(s))
    # See Request.POST
    from io import BytesIO
    def touni(x, enc='utf8', err='strict'):
        """ Convert anything to unicode """
        return str(x, enc, err) if isinstance(x, bytes) else str(x)
    if sys.version_info < (3,2,0):
        from io import TextIOWrapper
        class NCTextIOWrapper(TextIOWrapper):
            ''' Garbage collecting an io.TextIOWrapper(buffer) instance closes
                the wrapped buffer. This subclass keeps it open. '''
            def close(self): pass
else:
    json_loads = json_lds
    from StringIO import StringIO as BytesIO
    bytes = str
    def touni(x, enc='utf8', err='strict'):
        """ Convert anything to unicode """
        return x if isinstance(x, unicode) else unicode(str(x), enc, err)

def tob(data, enc='utf8'):
    """ Convert anything to bytes """
    return data.encode(enc) if isinstance(data, unicode) else bytes(data)

# Convert strings and unicode to native strings
if  py3k:
    tonat = touni
else:
    tonat = tob
tonat.__doc__ = """ Convert anything to native strings """


# Backward compatibility
def depr(message, critical=False):
    if critical: raise DeprecationWarning(message)
    warnings.warn(message, DeprecationWarning, stacklevel=3)


# Small helpers
def makelist(data):
    if isinstance(data, (tuple, list, set, dict)): return list(data)
    elif data: return [data]
    else: return []


class DictProperty(object):
    ''' Property that maps to a key in a local dict-like attribute. '''
    def __init__(self, attr, key=None, read_only=False):
        self.attr, self.key, self.read_only = attr, key, read_only

    def __call__(self, func):
        functools.update_wrapper(self, func, updated=[])
        self.getter, self.key = func, self.key or func.__name__
        return self

    def __get__(self, obj, cls):
        if obj is None: return self
        key, storage = self.key, getattr(obj, self.attr)
        if key not in storage: storage[key] = self.getter(obj)
        return storage[key]

    def __set__(self, obj, value):
        if self.read_only: raise AttributeError("Read-Only property.")
        getattr(obj, self.attr)[self.key] = value

    def __delete__(self, obj):
        if self.read_only: raise AttributeError("Read-Only property.")
        del getattr(obj, self.attr)[self.key]

class CachedProperty(object):
    ''' A property that is only computed once per instance and then replaces
        itself with an ordinary attribute. Deleting the attribute resets the
        property. '''

    def __init__(self, func):
        self.func = func

    def __get__(self, obj, cls):
        if obj is None: return self
        value = obj.__dict__[self.func.__name__] = self.func(obj)
        return value

cached_property = CachedProperty


class lazy_attribute(object): # Does not need configuration -> lower-case name
    ''' A property that caches itself to the class object. '''
    def __init__(self, func):
        functools.update_wrapper(self, func, updated=[])
        self.getter = func

    def __get__(self, obj, cls):
        value = self.getter(cls)
        setattr(cls, self.__name__, value)
        return value






###############################################################################
# Exceptions and Events ########################################################
###############################################################################


class BottleException(Exception):
    """ A base class for exceptions used by bottle. """
    pass


#TODO: These should subclass BaseRequest

class HTTPResponse(BottleException):
    """ Used to break execution and immediately finish the response """
    def __init__(self, output='', status=200, header=None):
        super(BottleException, self).__init__("HTTP Response %d" % status)
        self.status = int(status)
        self.output = output
        self.headers = HeaderDict(header) if header else None

    def apply(self, response):
        if self.headers:
            for key, value in self.headers.iterallitems():
                response.headers[key] = value
        response.status = self.status


class HTTPError(HTTPResponse):
    """ Used to generate an error page """
    def __init__(self, code=500, output='Unknown Error', exception=None,
                 traceback=None, header=None):
        super(HTTPError, self).__init__(output, code, header)
        self.exception = exception
        self.traceback = traceback

    def __repr__(self):
        return template(ERROR_PAGE_TEMPLATE, e=self)






###############################################################################
# Routing ######################################################################
###############################################################################


class RouteError(BottleException):
    """ This is a base class for all routing related exceptions """


class RouteReset(BottleException):
    """ If raised by a plugin or request handler, the route is reset and all
        plugins are re-applied. """


class RouteSyntaxError(RouteError):
    """ The route parser found something not supported by this router """


class RouteBuildError(RouteError):
    """ The route could not been built """


class Router(object):
    ''' A Router is an ordered collection of route->target pairs. It is used to
        efficiently match WSGI requests against a number of routes and return
        the first target that satisfies the request. The target may be anything,
        usually a string, ID or callable object. A route consists of a path-rule
        and a HTTP method.

        The path-rule is either a static path (e.g. `/contact`) or a dynamic
        path that contains wildcards (e.g. `/wiki/:page`). By default, wildcards
        consume characters up to the next slash (`/`). To change that, you may
        add a regular expression pattern (e.g. `/wiki/:page#[a-z]+#`).

        For performance reasons, static routes (rules without wildcards) are
        checked first. Dynamic routes are searched in order. Try to avoid
        ambiguous or overlapping rules.

        The HTTP method string matches only on equality, with two exceptions:
          * ´GET´ routes also match ´HEAD´ requests if there is no appropriate
            ´HEAD´ route installed.
          * ´ANY´ routes do match if there is no other suitable route installed.

        An optional ``name`` parameter is used by :meth:`build` to identify
        routes.
    '''

    default = '[^/]+'

    @lazy_attribute
    def syntax(cls):
        return re.compile(r'(?<!\\):([a-zA-Z_][a-zA-Z_0-9]*)?(?:#(.*?)#)?')

    def __init__(self):
        self.routes = {}  # A {rule: {method: target}} mapping
        self.rules  = []  # An ordered list of rules
        self.named  = {}  # A name->(rule, build_info) mapping
        self.static = {}  # Cache for static routes: {path: {method: target}}
        self.dynamic = [] # Cache for dynamic routes. See _compile()

    def add(self, rule, method, target, name=None):
        ''' Add a new route or replace the target for an existing route. '''

        if rule in self.routes:
            self.routes[rule][method.upper()] = target
        else:
            self.routes[rule] = {method.upper(): target}
            self.rules.append(rule)
            if self.static or self.dynamic: # Clear precompiler cache.
                self.static, self.dynamic = {}, {}
        if name:
            self.named[name] = (rule, None)

    def build(self, _name, *anon, **args):
        ''' Return a string that matches a named route. Use keyword arguments
            to fill out named wildcards. Remaining arguments are appended as a
            query string. Raises RouteBuildError or KeyError.'''
        if _name not in self.named:
            raise RouteBuildError("No route with that name.", _name)
        rule, pairs = self.named[_name]
        if not pairs:
            token = self.syntax.split(rule)
            parts = [p.replace('\\:',':') for p in token[::3]]
            names = token[1::3]
            if len(parts) > len(names): names.append(None)
            pairs = zip(parts, names)
            self.named[_name] = (rule, pairs)
        try:
            anon = list(anon)
            url = [s if k is None
                   else s+str(args.pop(k)) if k else s+str(anon.pop())
                   for s, k in pairs]
        except IndexError:
            msg = "Not enough arguments to fill out anonymous wildcards."
            raise RouteBuildError(msg)
        except KeyError, e:
            raise RouteBuildError(*e.args)

        if args: url += ['?', urlencode(args)]
        return ''.join(url)

    def match(self, environ):
        ''' Return a (target, url_agrs) tuple or raise HTTPError(404/405). '''
        targets, urlargs = self._match_path(environ)
        if not targets:
            raise HTTPError(404, "Not found: " + repr(environ['PATH_INFO']))
        method = environ['REQUEST_METHOD'].upper()
        if method in targets:
            return targets[method], urlargs
        if method == 'HEAD' and 'GET' in targets:
            return targets['GET'], urlargs
        if 'ANY' in targets:
            return targets['ANY'], urlargs
        allowed = [verb for verb in targets if verb != 'ANY']
        if 'GET' in allowed and 'HEAD' not in allowed:
            allowed.append('HEAD')
        raise HTTPError(405, "Method not allowed.",
                        header=[('Allow',",".join(allowed))])

    def _match_path(self, environ):
        ''' Optimized PATH_INFO matcher. '''
        path = environ['PATH_INFO'] or '/'
        # Assume we are in a warm state. Search compiled rules first.
        match = self.static.get(path)
        if match: return match, {}
        for combined, rules in self.dynamic:
            match = combined.match(path)
            if not match: continue
            gpat, match = rules[match.lastindex - 1]
            return match, gpat(path).groupdict() if gpat else {}
        # Lazy-check if we are really in a warm state. If yes, stop here.
        if self.static or self.dynamic or not self.routes: return None, {}
        # Cold state: We have not compiled any rules yet. Do so and try again.
        if not environ.get('wsgi.run_once'):
            self._compile()
            return self._match_path(environ)
        # For run_once (CGI) environments, don't compile. Just check one by one.
        epath = path.replace(':','\\:') # Turn path into its own static rule.
        match = self.routes.get(epath) # This returns static rule only.
        if match: return match, {}
        for rule in self.rules:
            #: Skip static routes to reduce re.compile() calls.
            if rule.count(':') < rule.count('\\:'): continue
            match = self._compile_pattern(rule).match(path)
            if match: return self.routes[rule], match.groupdict()
        return None, {}

    def _compile(self):
        ''' Prepare static and dynamic search structures. '''
        self.static = {}
        self.dynamic = []
        def fpat_sub(m):
            return m.group(0) if len(m.group(1)) % 2 else m.group(1) + '(?:'
        for rule in self.rules:
            target = self.routes[rule]
            if not self.syntax.search(rule):
                self.static[rule.replace('\\:',':')] = target
                continue
            gpat = self._compile_pattern(rule)
            fpat = re.sub(r'(\\*)(\(\?P<[^>]*>|\((?!\?))', fpat_sub, gpat.pattern)
            gpat = gpat.match if gpat.groupindex else None
            try:
                combined = '%s|(%s)' % (self.dynamic[-1][0].pattern, fpat)
                self.dynamic[-1] = (re.compile(combined), self.dynamic[-1][1])
                self.dynamic[-1][1].append((gpat, target))
            except (AssertionError, IndexError), e: # AssertionError: Too many groups
                self.dynamic.append((re.compile('(^%s$)'%fpat),
                                    [(gpat, target)]))
            except re.error, e:
                raise RouteSyntaxError("Could not add Route: %s (%s)" % (rule, e))

    def _compile_pattern(self, rule):
        ''' Return a regular expression with named groups for each wildcard. '''
        out = ''
        for i, part in enumerate(self.syntax.split(rule)):
            if i%3 == 0:   out += re.escape(part.replace('\\:',':'))
            elif i%3 == 1: out += '(?P<%s>' % part if part else '(?:'
            else:          out += '%s)' % (part or '[^/]+')
        return re.compile('^%s$'%out)


class Route(object):
    ''' This class wraps a route callback along with route specific metadata and
        configuration and applies Plugins on demand. '''

    def __init__(self, app, rule, method, callback, name=None,
                 plugins=None, skiplist=None, **config):
        #: The application this route is installed to.
        self.app = app
        #: The path-rule string (e.g. ``/wiki/:page``).
        self.rule = rule
        #: The HTTP method as a string (e.g. ``GET``).
        self.method = method
        #: The original callback with no plugins applied. Useful for introspection.
        self.callback = callback
        #: The name of the route (if specified) or ``None``.
        self.name = name or None
        #: A list of route-specific plugins (see :meth:`Bottle.route`).
        self.plugins = plugins or []
        #: A list of plugins to not apply to this route (see :meth:`Bottle.route`).
        self.skiplist = skiplist or []
        #: Additional keyword arguments passed to the :meth:`Bottle.route`
        #: decorator are stored in this dictionary. Used for route-specific
        #: plugin configuration and meta-data.
        self.config = config

    def __call__(self, *a, **ka):
        depr("Some APIs changed to return Route() instances instead of"\
             " callables. Make sure to use the Route.call method and not to"\
             " call Route instances directly.")
        return self.call(*a, **ka)

    @cached_property
    def call(self):
        ''' The route callback with all plugins applied. This property is
            created on demand and then cached to speed ub subsequent requests.'''
        return self._make_callback()

    def reset(self):
        ''' Forget any cached values. The next time :attr:`call` is accessed,
            all plugins are re-applied. '''
        self.__dict__.pop('call', None)

    def prepare(self):
        ''' Do all on-demand work immediately (useful for debugging).'''
        self.call

    @property
    def _context(self):
        depr('Switch to Plugin API v2 and access the Route object directly.')
        return dict(rule=self.rule, method=self.method, callback=self.callback,
                    name=self.name, app=self.app, config=self.config,
                    apply=self.plugins, skip=self.skiplist)

    def all_plugins(self):
        ''' Return a list of all Plugins that affect this route. '''
        return [] if True in self.skiplist else [p for p\
          in reversed(self.app.plugins + self.plugins)\
          if p not in self.skiplist and type(p) not in self.skiplist\
            and getattr(p, 'name', True) not in self.skiplist]

    def _make_callback(self):
        callback = self.callback
        for plugin in self.all_plugins():
            try:
                if hasattr(plugin, 'apply'):
                    api = getattr(plugin, 'api', 1)
                    context = self if api > 1 else self._context
                    callback = plugin.apply(callback, context)
                else:
                    callback = plugin(callback)
            except RouteReset: # Try again with changed configuration.
                return self._make_callback()
            functools.update_wrapper(callback, self.callback)
        return callback






###############################################################################
# Application Object ###########################################################
###############################################################################


class Bottle(object):
    """ WSGI application """

    def __init__(self, catchall=True, autojson=True, config=None):
        """ Create a new bottle instance.
            You usually don't do that. Use `bottle.app.push()` instead.
        """
        self.routes = [] # List of installed :class:`Route` instances.
        self.router = Router() # Maps requests to :class:`Route` instances.
        self.plugins = [] # List of installed plugins.

        self.mounts = {}
        self.error_handler = {}
        #: If true, most exceptions are catched and returned as :exc:`HTTPError`
        self.catchall = catchall
        self.config = config or {}
        # Default plugins
        self.hooks = self.install(HooksPlugin())
        if autojson:
            self.install(JSONPlugin())
        self.install(TemplatePlugin())

    def mount(self, app, prefix, **options):
        ''' Mount an application to a specific URL prefix. The prefix is added
            to SCIPT_PATH and removed from PATH_INFO before the sub-application
            is called.

            :param app: an instance of :class:`Bottle`.
            :param prefix: path prefix used as a mount-point.

            All other parameters are passed to the underlying :meth:`route` call.
        '''
        if not isinstance(app, Bottle):
            raise TypeError('Only Bottle instances are supported for now.')
        prefix = '/'.join(filter(None, prefix.split('/')))
        if not prefix:
            raise TypeError('Empty prefix. Perhaps you want a merge()?')
        for other in self.mounts:
            if other.startswith(prefix):
                raise TypeError('Conflict with existing mount: %s' % other)
        path_depth = prefix.count('/') + 1
        options.setdefault('method', 'ANY')
        options.setdefault('skip', True)
        self.mounts[prefix] = app
        @self.route('/%s/:#.*#' % prefix, **options)
        def mountpoint():
            request.path_shift(path_depth)
            return app._handle(request.environ)

    def install(self, plugin):
        ''' Add a plugin to the list of plugins and prepare it for being
            applied to all routes of this application. A plugin may be a simple
            decorator or an object that implements the :class:`Plugin` API.
        '''
        if hasattr(plugin, 'setup'): plugin.setup(self)
        if not callable(plugin) and not hasattr(plugin, 'apply'):
            raise TypeError("Plugins must be callable or implement .apply()")
        self.plugins.append(plugin)
        self.reset()
        return plugin

    def uninstall(self, plugin):
        ''' Uninstall plugins. Pass an instance to remove a specific plugin, a type
            object to remove all plugins that match that type, a string to remove
            all plugins with a matching ``name`` attribute or ``True`` to remove all
            plugins. Return the list of removed plugins. '''
        removed, remove = [], plugin
        for i, plugin in list(enumerate(self.plugins))[::-1]:
            if remove is True or remove is plugin or remove is type(plugin) \
            or getattr(plugin, 'name', True) == remove:
                removed.append(plugin)
                del self.plugins[i]
                if hasattr(plugin, 'close'): plugin.close()
        if removed: self.reset()
        return removed

    def reset(self, route=None):
        ''' Reset all routes (force plugins to be re-applied) and clear all
            caches. If an ID or route object is given, only that specific route
            is affected. '''
        if route is None: routes = self.routes
        elif isinstance(route, Route): routes = [route]
        else: routes = [self.routes[route]]
        for route in routes: route.reset()
        if DEBUG:
            for route in routes: route.prepare()

    def close(self):
        ''' Close the application and all installed plugins. '''
        for plugin in self.plugins:
            if hasattr(plugin, 'close'): plugin.close()
        self.stopped = True

    def match(self, environ):
        """ Search for a matching route and return a (:class:`Route` , urlargs)
            tuple. The second value is a dictionary with parameters extracted
            from the URL. Raise :exc:`HTTPError` (404/405) on a non-match."""
        route, args = self.router.match(environ)
        return route, args

    def get_url(self, routename, **kargs):
        """ Return a string that matches a named route """
        scriptname = request.environ.get('SCRIPT_NAME', '').strip('/') + '/'
        location = self.router.build(routename, **kargs).lstrip('/')
        return urljoin(urljoin('/', scriptname), location)

    def route(self, path=None, method='GET', callback=None, name=None,
              apply=None, skip=None, **config):
        """ A decorator to bind a function to a request URL. Example::

                @app.route('/hello/:name')
                def hello(name):
                    return 'Hello %s' % name

            The ``:name`` part is a wildcard. See :class:`Router` for syntax
            details.

            :param path: Request path or a list of paths to listen to. If no
              path is specified, it is automatically generated from the
              signature of the function.
            :param method: HTTP method (`GET`, `POST`, `PUT`, ...) or a list of
              methods to listen to. (default: `GET`)
            :param callback: An optional shortcut to avoid the decorator
              syntax. ``route(..., callback=func)`` equals ``route(...)(func)``
            :param name: The name for this route. (default: None)
            :param apply: A decorator or plugin or a list of plugins. These are
              applied to the route callback in addition to installed plugins.
            :param skip: A list of plugins, plugin classes or names. Matching
              plugins are not installed to this route. ``True`` skips all.

            Any additional keyword arguments are stored as route-specific
            configuration and passed to plugins (see :meth:`Plugin.apply`).
        """
        if callable(path): path, callback = None, path
        plugins = makelist(apply)
        skiplist = makelist(skip)
        def decorator(callback):
            for rule in makelist(path) or yieldroutes(callback):
                for verb in makelist(method):
                    verb = verb.upper()
                    route = Route(self, rule, verb, callback, name=name,
                                  plugins=plugins, skiplist=skiplist, **config)
                    self.routes.append(route)
                    self.router.add(rule, verb, route, name=name)
                    if DEBUG: route.prepare()
            return callback
        return decorator(callback) if callback else decorator

    def get(self, path=None, method='GET', **options):
        """ Equals :meth:`route`. """
        return self.route(path, method, **options)

    def post(self, path=None, method='POST', **options):
        """ Equals :meth:`route` with a ``POST`` method parameter. """
        return self.route(path, method, **options)

    def put(self, path=None, method='PUT', **options):
        """ Equals :meth:`route` with a ``PUT`` method parameter. """
        return self.route(path, method, **options)

    def delete(self, path=None, method='DELETE', **options):
        """ Equals :meth:`route` with a ``DELETE`` method parameter. """
        return self.route(path, method, **options)

    def error(self, code=500):
        """ Decorator: Register an output handler for a HTTP error code"""
        def wrapper(handler):
            self.error_handler[int(code)] = handler
            return handler
        return wrapper

    def hook(self, name):
        """ Return a decorator that attaches a callback to a hook. """
        def wrapper(func):
            self.hooks.add(name, func)
            return func
        return wrapper

    def handle(self, path, method='GET'):
        """ (deprecated) Execute the first matching route callback and return
            the result. :exc:`HTTPResponse` exceptions are catched and returned.
            If :attr:`Bottle.catchall` is true, other exceptions are catched as
            well and returned as :exc:`HTTPError` instances (500).
        """
        depr("This method will change semantics in 0.10. Try to avoid it.")
        if isinstance(path, dict):
            return self._handle(path)
        return self._handle({'PATH_INFO': path, 'REQUEST_METHOD': method.upper()})

    def _handle(self, environ):
        try:
            route, args = self.match(environ)
            environ['route.handle'] = environ['bottle.route'] = route
            environ['route.url_args'] = args
            return route.call(**args)
        except HTTPResponse, r:
            return r
        except RouteReset:
            route.reset()
            return self._handle(environ)
        except (KeyboardInterrupt, SystemExit, MemoryError):
            raise
        except Exception, e:
            if not self.catchall: raise
            stacktrace = format_exc(10)
            environ['wsgi.errors'].write(stacktrace)
            return HTTPError(500, "Internal Server Error", e, stacktrace)

    def _cast(self, out, request, response, peek=None):
        """ Try to convert the parameter into something WSGI compatible and set
        correct HTTP headers when possible.
        Support: False, str, unicode, dict, HTTPResponse, HTTPError, file-like,
        iterable of strings and iterable of unicodes
        """

        # Empty output is done here
        if not out:
            response['Content-Length'] = 0
            return []
        # Join lists of byte or unicode strings. Mixed lists are NOT supported
        if isinstance(out, (tuple, list))\
        and isinstance(out[0], (bytes, unicode)):
            out = out[0][0:0].join(out) # b'abc'[0:0] -> b''
        # Encode unicode strings
        if isinstance(out, unicode):
            out = out.encode(response.charset)
        # Byte Strings are just returned
        if isinstance(out, bytes):
            response['Content-Length'] = len(out)
            return [out]
        # HTTPError or HTTPException (recursive, because they may wrap anything)
        # TODO: Handle these explicitly in handle() or make them iterable.
        if isinstance(out, HTTPError):
            out.apply(response)
            out = self.error_handler.get(out.status, repr)(out)
            if isinstance(out, HTTPResponse):
                depr('Error handlers must not return :exc:`HTTPResponse`.') #0.9
            return self._cast(out, request, response)
        if isinstance(out, HTTPResponse):
            out.apply(response)
            return self._cast(out.output, request, response)

        # File-like objects.
        if hasattr(out, 'read'):
            if 'wsgi.file_wrapper' in request.environ:
                return request.environ['wsgi.file_wrapper'](out)
            elif hasattr(out, 'close') or not hasattr(out, '__iter__'):
                return WSGIFileWrapper(out)

        # Handle Iterables. We peek into them to detect their inner type.
        try:
            out = iter(out)
            first = out.next()
            while not first:
                first = out.next()
        except StopIteration:
            return self._cast('', request, response)
        except HTTPResponse, e:
            first = e
        except Exception, e:
            first = HTTPError(500, 'Unhandled exception', e, format_exc(10))
            if isinstance(e, (KeyboardInterrupt, SystemExit, MemoryError))\
            or not self.catchall:
                raise
        # These are the inner types allowed in iterator or generator objects.
        if isinstance(first, HTTPResponse):
            return self._cast(first, request, response)
        if isinstance(first, bytes):
            return itertools.chain([first], out)
        if isinstance(first, unicode):
            return itertools.imap(lambda x: x.encode(response.charset),
                                  itertools.chain([first], out))
        return self._cast(HTTPError(500, 'Unsupported response type: %s'\
                                         % type(first)), request, response)

    def wsgi(self, environ, start_response):
        """ The bottle WSGI-interface. """
        try:
            environ['bottle.app'] = self
            request.bind(environ)
            response.bind()
            out = self._cast(self._handle(environ), request, response)
            # rfc2616 section 4.3
            if response.status_code in (100, 101, 204, 304)\
            or request.method == 'HEAD':
                if hasattr(out, 'close'): out.close()
                out = []
            start_response(response.status_line, list(response.iter_headers()))
            return out
        except (KeyboardInterrupt, SystemExit, MemoryError):
            raise
        except Exception, e:
            if not self.catchall: raise
            err = '<h1>Critical error while processing request: %s</h1>' \
                  % environ.get('PATH_INFO', '/')
            if DEBUG:
                err += '<h2>Error:</h2>\n<pre>%s</pre>\n' % repr(e)
                err += '<h2>Traceback:</h2>\n<pre>%s</pre>\n' % format_exc(10)
            environ['wsgi.errors'].write(err) #TODO: wsgi.error should not get html
            start_response('500 INTERNAL SERVER ERROR', [('Content-Type', 'text/html')])
            return [tob(err)]

    def __call__(self, environ, start_response):
        return self.wsgi(environ, start_response)






###############################################################################
# HTTP and WSGI Tools ##########################################################
###############################################################################


class BaseRequest(DictMixin):
    """ A wrapper for WSGI environment dictionaries that adds a lot of
        convenient access methods and properties. Most of them are read-only."""

    #: Maximum size of memory buffer for :attr:`body` in bytes.
    MEMFILE_MAX = 102400

    def __init__(self, environ):
        """ Wrap a WSGI environ dictionary. """
        #: The wrapped WSGI environ dictionary. This is the only real attribute.
        #: All other attributes actually are read-only properties.
        self.environ = environ
        environ['bottle.request'] = self

    @property
    def path(self):
        ''' The value of ``PATH_INFO`` with exactly one prefixed slash (to fix
            broken clients and avoid the "empty path" edge case). '''
        return '/' + self.environ.get('PATH_INFO','').lstrip('/')

    @property
    def method(self):
        ''' The ``REQUEST_METHOD`` value as an uppercase string. '''
        return self.environ.get('REQUEST_METHOD', 'GET').upper()

    @DictProperty('environ', 'bottle.request.headers', read_only=True)
    def headers(self):
        ''' A :class:`WSGIHeaderDict` that provides case-insensitive access to
            HTTP request headers. '''
        return WSGIHeaderDict(self.environ)

    @DictProperty('environ', 'bottle.request.cookies', read_only=True)
    def cookies(self):
        """ Cookies parsed into a dictionary. Signed cookies are NOT decoded.
            Use :meth:`get_cookie` if you expect signed cookies. """
        raw_dict = SimpleCookie(self.environ.get('HTTP_COOKIE',''))
        cookies = {}
        for cookie in raw_dict.itervalues():
            cookies[cookie.key] = cookie.value
        return cookies

    def get_cookie(self, key, default=None, secret=None):
        """ Return the content of a cookie. To read a `Signed Cookie`, the
            `secret` must match the one used to create the cookie (see
            :meth:`BaseResponse.set_cookie`). If anything goes wrong (missing
            cookie or wrong signature), return a default value. """
        value = self.cookies.get(key)
        if secret and value:
            dec = cookie_decode(value, secret) # (key, value) tuple or None
            return dec[1] if dec and dec[0] == key else default
        return value or default

    @DictProperty('environ', 'bottle.request.query', read_only=True)
    def query(self):
        ''' The :attr:`query_string` parsed into a :class:`MultiDict`. These
            values are sometimes called "URL arguments" or "GET parameters", but
            not to be confused with "URL wildcards" as they are provided by the
            :class:`Router`. '''
        data = parse_qs(self.query_string, keep_blank_values=True)
        get = self.environ['bottle.get'] = MultiDict()
        for key, values in data.iteritems():
            for value in values:
                get[key] = value
        return get

    @DictProperty('environ', 'bottle.request.forms', read_only=True)
    def forms(self):
        """ Form values parsed from an `url-encoded` or `multipart/form-data`
            encoded POST or PUT request body. The result is retuned as a
            :class:`MultiDict`. All keys and values are strings. File uploads
            are stored separately in :attr:`files`. """
        forms = MultiDict()
        for name, item in self.POST.iterallitems():
            if not hasattr(item, 'filename'):
                forms[name] = item
        return forms

    @DictProperty('environ', 'bottle.request.params', read_only=True)
    def params(self):
        """ A :class:`MultiDict` with the combined values of :attr:`query` and
            :attr:`forms`. File uploads are stored in :attr:`files`. """
        params = MultiDict()
        for key, value in self.query.iterallitems():
            params[key] = value
        for key, value in self.forms.iterallitems():
            params[key] = value
        return params

    @DictProperty('environ', 'bottle.request.files', read_only=True)
    def files(self):
        """ File uploads parsed from an `url-encoded` or `multipart/form-data`
            encoded POST or PUT request body. The values are instances of
            :class:`cgi.FieldStorage`. The most important attributes are:

            filename
                The filename, if specified; otherwise None; this is the client
                side filename, *not* the file name on which it is stored (that's
                a temporary file you don't deal with)
            file
                The file(-like) object from which you can read the data.
            value
                The value as a *string*; for file uploads, this transparently
                reads the file every time you request the value. Do not do this
                on big files.
        """
        files = MultiDict()
        for name, item in self.POST.iterallitems():
            if hasattr(item, 'filename'):
                files[name] = item
        return files

    @DictProperty('environ', 'bottle.request.json', read_only=True)
    def json(self):
        ''' If the ``Content-Type`` header is ``application/json``, this
            property holds the parsed content of the request body. Only requests
            smaller than :attr:`MEMFILE_MAX` are processed to avoid memory
            exhaustion. '''
        if self.environ.get('CONTENT_TYPE') == 'application/json' \
        and 0 < self.content_length < self.MEMFILE_MAX:
            return json_loads(self.body.read(self.MEMFILE_MAX))
        return None

    @DictProperty('environ', 'bottle.request.body', read_only=True)
    def _body(self):
        maxread = max(0, self.content_length)
        stream = self.environ['wsgi.input']
        body = BytesIO() if maxread < self.MEMFILE_MAX else TemporaryFile(mode='w+b')
        while maxread > 0:
            part = stream.read(min(maxread, self.MEMFILE_MAX))
            if not part: break
            body.write(part)
            maxread -= len(part)
        self.environ['wsgi.input'] = body
        body.seek(0)
        return body

    @property
    def body(self):
        """ The HTTP request body as a seek-able file-like object. Depending on
            :attr:`MEMFILE_MAX`, this is either a temporary file or a
            :class:`io.BytesIO` instance. Accessing this property for the first
            time reads and replaces the ``wsgi.input`` environ variable.
            Subsequent accesses just do a `seek(0)` on the file object. """
        self._body.seek(0)
        return self._body

    #: An alias for :attr:`query`.
    GET = query

    @DictProperty('environ', 'bottle.request.post', read_only=True)
    def POST(self):
        """ The values of :attr:`forms` and :attr:`files` combined into a single
            :class:`MultiDict`. Values are either strings (form values) or
            instances of :class:`cgi.FieldStorage` (file uploads).
        """
        post = MultiDict()
        safe_env = {'QUERY_STRING':''} # Build a safe environment for cgi
        for key in ('REQUEST_METHOD', 'CONTENT_TYPE', 'CONTENT_LENGTH'):
            if key in self.environ: safe_env[key] = self.environ[key]
        if NCTextIOWrapper:
            fb = NCTextIOWrapper(self.body, encoding='ISO-8859-1', newline='\n')
        else:
            fb = self.body
        data = cgi.FieldStorage(fp=fb, environ=safe_env, keep_blank_values=True)
        for item in data.list or []:
            post[item.name] = item if item.filename else item.value
        return post

    @property
    def COOKIES(self):
        ''' Alias for :attr:`cookies` (deprecated). '''
        depr('BaseRequest.COOKIES was renamed to BaseRequest.cookies (lowercase).')
        return self.cookies

    @property
    def url(self):
        """ The full request URI including hostname and scheme. If your app
            lives behind a reverse proxy or load balancer and you get confusing
            results, make sure that the ``X-Forwarded-Host`` header is set
            correctly. """
        return self.urlparts.geturl()

    @DictProperty('environ', 'bottle.request.urlparts', read_only=True)
    def urlparts(self):
        ''' The :attr:`url` string as an :class:`urlparse.SplitResult` tuple.
            The tuple contains (scheme, host, path, query_string and fragment),
            but the fragment is always empty because it is not visible to the
            server. '''
        env = self.environ
        http = env.get('wsgi.url_scheme', 'http')
        host = env.get('HTTP_X_FORWARDED_HOST') or env.get('HTTP_HOST')
        if not host:
            # HTTP 1.1 requires a Host-header. This is for HTTP/1.0 clients.
            host = env.get('SERVER_NAME', '127.0.0.1')
            port = env.get('SERVER_PORT')
            if port and port != ('80' if http == 'http' else '443'):
                host += ':' + port
        path = urlquote(self.fullpath)
        return UrlSplitResult(http, host, path, env.get('QUERY_STRING'), '')

    @property
    def fullpath(self):
        """ Request path including :attr:`script_name` (if present). """
        return urljoin(self.script_name, self.path.lstrip('/'))

    @property
    def query_string(self):
        """ The raw :attr:`query` part of the URL (everything in between ``?``
            and ``#``) as a string. """
        return self.environ.get('QUERY_STRING', '')

    @property
    def script_name(self):
        ''' The initial portion of the URL's `path` that was removed by a higher
            level (server or routing middleware) before the application was
            called. This property returns an empty string, or a path with
            leading and tailing slashes. '''
        script_name = self.environ.get('SCRIPT_NAME', '').strip('/')
        return '/' + script_name + '/' if script_name else '/'

    def path_shift(self, shift=1):
        ''' Shift path segments from :attr:`path` to :attr:`script_name` and
            vice versa.

           :param shift: The number of path segments to shift. May be negative
                         to change the shift direction. (default: 1)
        '''
        script = self.environ.get('SCRIPT_NAME','/')
        self['SCRIPT_NAME'], self['PATH_INFO'] = path_shift(script, self.path, shift)

    @property
    def content_length(self):
        ''' The request body length as an integer. The client is responsible to
            set this header. Otherwise, the real length of the body is unknown
            and -1 is returned. In this case, :attr:`body` will be empty. '''
        return int(self.environ.get('CONTENT_LENGTH') or -1)

    @property
    def is_xhr(self):
        ''' True if the request was triggered by a XMLHttpRequest. This only
            works with JavaScript libraries that support the `X-Requested-With`
            header (most of the popular libraries do). '''
        requested_with = self.environ.get('HTTP_X_REQUESTED_WITH','')
        return requested_with.lower() == 'xmlhttprequest'

    @property
    def is_ajax(self):
        ''' Alias for :attr:`is_xhr`. "Ajax" is not the right term. '''
        return self.is_xhr

    @property
    def auth(self):
        """ HTTP authentication data as a (user, password) tuple. This
            implementation currently supports basic (not digest) authentication
            only. If the authentication happened at a higher level (e.g. in the
            front web-server or a middleware), the password field is None, but
            the user field is looked up from the ``REMOTE_USER`` environ
            variable. On any errors, None is returned. """
        basic = parse_auth(self.environ.get('HTTP_AUTHORIZATION',''))
        if basic: return basic
        ruser = self.environ.get('REMOTE_USER')
        if ruser: return (ruser, None)
        return None

    @property
    def remote_route(self):
        """ A list of all IPs that were involved in this request, starting with
            the client IP and followed by zero or more proxies. This does only
            work if all proxies support the ```X-Forwarded-For`` header. Note
            that this information can be forged by malicious clients. """
        proxy = self.environ.get('HTTP_X_FORWARDED_FOR')
        if proxy: return [ip.strip() for ip in proxy.split(',')]
        remote = self.environ.get('REMOTE_ADDR')
        return [remote] if remote else []

    @property
    def remote_addr(self):
        """ The client IP as a string. Note that this information can be forged
            by malicious clients. """
        route = self.remote_route
        return route[0] if route else None

    def copy(self):
        """ Return a new :class:`Request` with a shallow :attr:`environ` copy. """
        return Request(self.environ.copy())

    def __getitem__(self, key): return self.environ[key]
    def __delitem__(self, key): self[key] = ""; del(self.environ[key])
    def __iter__(self): return iter(self.environ)
    def __len__(self): return len(self.environ)
    def keys(self): return self.environ.keys()
    def __setitem__(self, key, value):
        """ Change an environ value and clear all caches that depend on it. """

        if self.environ.get('bottle.request.readonly'):
            raise KeyError('The environ dictionary is read-only.')

        self.environ[key] = value
        todelete = ()

        if key == 'wsgi.input':
            todelete = ('body', 'forms', 'files', 'params', 'post', 'json')
        elif key == 'QUERY_STRING':
            todelete = ('query', 'params')
        elif key.startswith('HTTP_'):
            todelete = ('headers', 'cookies')

        for key in todelete:
            self.environ.pop('bottle.request.'+key, None)

    def __repr__(self):
        return '<%s: %s %s>' % (self.__class__.__name__, self.method, self.url)

def _hkey(s):
    return s.title().replace('_','-')


class HeaderProperty(object):
    def __init__(self, name, reader=None, writer=str, default=''):
        self.name, self.reader, self.writer, self.default = name, reader, writer, default
        self.__doc__ = 'Current value of the %r header.' % name.title()

    def __get__(self, obj, cls):
        if obj is None: return self
        value = obj.headers.get(self.name)
        return self.reader(value) if (value and self.reader) else (value or self.default)

    def __set__(self, obj, value):
        if self.writer: value = self.writer(value)
        obj.headers[self.name] = value

    def __delete__(self, obj):
        if self.name in obj.headers:
            del obj.headers[self.name]


class BaseResponse(object):
    """ Storage class for a response body as well as headers and cookies.

        This class does support dict-like case-insensitive item-access to
        headers, but is NOT a dict. Most notably, iterating over a response
        yields parts of the body and not the headers.
    """

    default_status = 200
    default_content_type = 'text/html; charset=UTF-8'

    # Header blacklist for specific response codes
    # (rfc2616 section 10.2.3 and 10.3.5)
    bad_headers = {
        204: set(('Content-Type',)),
        304: set(('Allow', 'Content-Encoding', 'Content-Language',
                  'Content-Length', 'Content-Range', 'Content-Type',
                  'Content-Md5', 'Last-Modified'))}

    def __init__(self, body='', status=None, **headers):
        #: The HTTP status code as an integer (e.g. 404).
        #: Do not change the value manually, use :attr:`status` instead.
        self.status_code = None
        #: The HTTP status line as a string (e.g. "404 Not Found").
        #: Do not change the value manually, use :attr:`status` instead.
        self.status_line = None
        #: The response body as one of the supported data types.
        self.body = body
        self._cookies = None
        self._headers = {'Content-Type': [self.default_content_type]}
        self.status = status or self.default_status
        if headers:
            for name, value in headers.items():
                self[name] = value

    def copy(self):
        ''' Returns a copy of self. '''
        copy = Response()
        copy.status = self.status
        copy._headers = dict((k, v[:]) for (k, v) in self._headers.items())
        return copy

    def __iter__(self):
        return iter(self.body)

    def close(self):
        if hasattr(self.body, 'close'):
            self.body.close()

    def _set_status(self, status):
        if isinstance(status, int):
            code, status = status, _HTTP_STATUS_LINES.get(status)
        elif ' ' in status:
            status = status.strip()
            code   = int(status.split()[0])
        else:
            raise ValueError('String status line without a reason phrase.')
        if not 100 <= code <= 999: raise ValueError('Status code out of range.')
        self.status_code = code
        self.status_line = status or ('%d Unknown' % code)

    status = property(lambda self: self.status_code, _set_status, None,
        ''' A writeable property to change the HTTP response status. It accepts
            either a numeric code (100-999) or a string with a custom reason
            phrase (e.g. "404 Brain not found"). Both :data:`status_line` and
            :data:`status_code` are updates accordingly. The return value is
            always a numeric code. ''')
    del _set_status

    @property
    def headers(self):
        ''' An instance of :class:`HeaderDict`, a case-insensitive dict-like
            view on the response headers. '''
        self.__dict__['headers'] = hdict = HeaderDict()
        hdict.dict = self._headers
        return hdict

    def __contains__(self, name): return _hkey(name) in self._headers
    def __delitem__(self, name):  del self._headers[_hkey(name)]
    def __getitem__(self, name):  return self._headers[_hkey(name)][-1]
    def __setitem__(self, name, value): self._headers[_hkey(name)] = [str(value)]

    def get_header(self, name, default=None):
        ''' Return the value of a previously defined header. If there is no
            header with that name, return a default value. '''
        return self._headers.get(_hkey(name), [default])[-1]

    def set_header(self, name, value, append=False):
        ''' Create a new response header, replacing any previously defined
            headers with the same name. This equals ``response[name] = value``.

            :param append: Do not delete previously defined headers. This can
                           result in two (or more) headers having the same name.
        '''
        if append:
            self._headers.setdefault(_hkey(name), []).append(str(value))
        else:
            self._headers[_hkey(name)] = [str(value)]

    def iter_headers(self):
        ''' Yield (header, value) tuples, skipping headers that are not
            allowed with the current response status code. '''
        headers = self._headers.iteritems()
        bad_headers = self.bad_headers.get(self.status_code)
        if bad_headers:
            headers = (h for h in headers if h[0] not in bad_headers)
        for name, values in headers:
            for value in values:
                yield name, value
        if self._cookies:
            for c in self._cookies.values():
                yield 'Set-Cookie', c.OutputString()

    def wsgiheader(self):
        depr('The wsgiheader method is deprecated. See headerlist.') #0.10
        return self.headerlist

    @property
    def headerlist(self):
        ''' WSGI conform list of (header, value) tuples. '''
        return list(self.iter_headers())

    content_type = HeaderProperty('Content-Type')
    content_length = HeaderProperty('Content-Length', reader=int)

    @property
    def charset(self):
        """ Return the charset specified in the content-type header (default: utf8). """
        if 'charset=' in self.content_type:
            return self.content_type.split('charset=')[-1].split(';')[0].strip()
        return 'UTF-8'

    @property
    def COOKIES(self):
        """ A dict-like SimpleCookie instance. This should not be used directly.
            See :meth:`set_cookie`. """
        depr('The COOKIES dict is deprecated. Use `set_cookie()` instead.') # 0.10
        if not self._cookies:
            self._cookies = SimpleCookie()
        return self._cookies

    def set_cookie(self, key, value, secret=None, **options):
        ''' Create a new cookie or replace an old one. If the `secret` parameter is
            set, create a `Signed Cookie` (described below).

            :param key: the name of the cookie.
            :param value: the value of the cookie.
            :param secret: a signature key required for signed cookies.

            Additionally, this method accepts all RFC 2109 attributes that are
            supported by :class:`cookie.Morsel`, including:

            :param max_age: maximum age in seconds. (default: None)
            :param expires: a datetime object or UNIX timestamp. (default: None)
            :param domain: the domain that is allowed to read the cookie.
              (default: current domain)
            :param path: limits the cookie to a given path (default: ``/``)
            :param secure: limit the cookie to HTTPS connections (default: off).
            :param httponly: prevents client-side javascript to read this cookie
              (default: off, requires Python 2.6 or newer).

            If neither `expires` nor `max_age` is set (default), the cookie will
            expire at the end of the browser session (as soon as the browser
            window is closed).

            Signed cookies may store any pickle-able object and are
            cryptographically signed to prevent manipulation. Keep in mind that
            cookies are limited to 4kb in most browsers.

            Warning: Signed cookies are not encrypted (the client can still see
            the content) and not copy-protected (the client can restore an old
            cookie). The main intention is to make pickling and unpickling
            save, not to store secret information at client side.
        '''
        if not self._cookies:
            self._cookies = SimpleCookie()

        if secret:
            value = touni(cookie_encode((key, value), secret))
        elif not isinstance(value, basestring):
            raise TypeError('Secret key missing for non-string Cookie.')

        self._cookies[key] = value
        for k, v in options.iteritems():
            self._cookies[key][k.replace('_', '-')] = v

    def delete_cookie(self, key, **kwargs):
        ''' Delete a cookie. Be sure to use the same `domain` and `path`
            settings as used to create the cookie. '''
        kwargs['max_age'] = -1
        kwargs['expires'] = 0
        self.set_cookie(key, '', **kwargs)


class LocalRequest(BaseRequest, threading.local):
    ''' A thread-local subclass of :class:`BaseRequest`. '''
    def __init__(self): pass
    bind = BaseRequest.__init__


class LocalResponse(BaseResponse, threading.local):
    ''' A thread-local subclass of :class:`BaseResponse`. '''
    bind = BaseResponse.__init__

Response = LocalResponse # BC 0.9
Request  = LocalRequest  # BC 0.9






###############################################################################
# Plugins ######################################################################
###############################################################################


class JSONPlugin(object):
    name = 'json'
    api  = 2

    def __init__(self, json_dumps=json_dumps):
        self.json_dumps = json_dumps

    def apply(self, callback, context):
        dumps = self.json_dumps
        if not dumps: return callback
        def wrapper(*a, **ka):
            rv = callback(*a, **ka)
            if isinstance(rv, dict):
                #Attempt to serialize, raises exception on failure
                json_response = dumps(rv)
                #Set content type only if serialization succesful
                response.content_type = 'application/json'
                return json_response
            return rv
        return wrapper


class HooksPlugin(object):
    name = 'hooks'
    api  = 2

    def __init__(self):
        self.hooks = {'before_request': [], 'after_request': []}
        self.app = None

    def _empty(self):
        return not (self.hooks['before_request'] or self.hooks['after_request'])

    def setup(self, app):
        self.app = app

    def add(self, name, func):
        ''' Attach a callback to a hook. '''
        if name not in self.hooks:
            raise ValueError("Unknown hook name %s" % name)
        was_empty = self._empty()
        self.hooks[name].append(func)
        if self.app and was_empty and not self._empty(): self.app.reset()

    def remove(self, name, func):
        ''' Remove a callback from a hook. '''
        if name not in self.hooks:
            raise ValueError("Unknown hook name %s" % name)
        was_empty = self._empty()
        self.hooks[name].remove(func)
        if self.app and not was_empty and self._empty(): self.app.reset()

    def apply(self, callback, context):
        if self._empty(): return callback
        before_request = self.hooks['before_request']
        after_request  = self.hooks['after_request']
        def wrapper(*a, **ka):
            for hook in before_request: hook()
            rv = callback(*a, **ka)
            for hook in after_request[::-1]: hook()
            return rv
        return wrapper


class TemplatePlugin(object):
    ''' This plugin applies the :func:`view` decorator to all routes with a
        `template` config parameter. If the parameter is a tuple, the second
        element must be a dict with additional options (e.g. `template_engine`)
        or default variables for the template. '''
    name = 'template'
    api  = 2

    def apply(self, callback, route):
        conf = route.config.get('template')
        if isinstance(conf, (tuple, list)) and len(conf) == 2:
            return view(conf[0], **conf[1])(callback)
        elif isinstance(conf, str) and 'template_opts' in route.config:
            depr('The `template_opts` parameter is deprecated.') #0.9
            return view(conf, **route.config['template_opts'])(callback)
        elif isinstance(conf, str):
            return view(conf)(callback)
        else:
            return callback


#: Not a plugin, but part of the plugin API. TODO: Find a better place.
class _ImportRedirect(object):
    def __init__(self, name, impmask):
        ''' Create a virtual package that redirects imports (see PEP 302). '''
        self.name = name
        self.impmask = impmask
        self.module = sys.modules.setdefault(name, imp.new_module(name))
        self.module.__dict__.update({'__file__': __file__, '__path__': [],
                                    '__all__': [], '__loader__': self})
        sys.meta_path.append(self)

    def find_module(self, fullname, path=None):
        if '.' not in fullname: return
        packname, modname = fullname.rsplit('.', 1)
        if packname != self.name: return
        return self

    def load_module(self, fullname):
        if fullname in sys.modules: return sys.modules[fullname]
        packname, modname = fullname.rsplit('.', 1)
        realname = self.impmask % modname
        __import__(realname)
        module = sys.modules[fullname] = sys.modules[realname]
        setattr(self.module, modname, module)
        module.__loader__ = self
        return module






###############################################################################
# Common Utilities #############################################################
###############################################################################


class MultiDict(DictMixin):
    """ This dict stores multiple values per key, but behaves exactly like a
        normal dict in that it returns only the newest value for any given key.
        There are special methods available to access the full list of values.
    """

    def __init__(self, *a, **k):
        self.dict = dict((k, [v]) for k, v in dict(*a, **k).iteritems())
    def __len__(self): return len(self.dict)
    def __iter__(self): return iter(self.dict)
    def __contains__(self, key): return key in self.dict
    def __delitem__(self, key): del self.dict[key]
    def __getitem__(self, key): return self.dict[key][-1]
    def __setitem__(self, key, value): self.append(key, value)
    def iterkeys(self): return self.dict.iterkeys()
    def itervalues(self): return (v[-1] for v in self.dict.itervalues())
    def iteritems(self): return ((k, v[-1]) for (k, v) in self.dict.iteritems())
    def iterallitems(self):
        for key, values in self.dict.iteritems():
            for value in values:
                yield key, value

    # 2to3 is not able to fix these automatically.
    keys     = iterkeys     if py3k else lambda self: list(self.iterkeys())
    values   = itervalues   if py3k else lambda self: list(self.itervalues())
    items    = iteritems    if py3k else lambda self: list(self.iteritems())
    allitems = iterallitems if py3k else lambda self: list(self.iterallitems())

    def get(self, key, default=None, index=-1):
        ''' Return the current value for a key. The third `index` parameter
            defaults to -1 (last value). '''
        if key in self.dict or default is KeyError:
            return self.dict[key][index]
        return default

    #: Alias for :meth:`get` to mimic other multi-dict APIs (Django)
    getone = get

    def append(self, key, value):
        ''' Add a new value to the list of values for this key. '''
        self.dict.setdefault(key, []).append(value)

    def replace(self, key, value):
        ''' Replace the list of values with a single value. '''
        self.dict[key] = [value]

    def getall(self, key):
        ''' Return a (possibly empty) list of values for a key. '''
        return self.dict.get(key) or []


class HeaderDict(MultiDict):
    """ A case-insensitive version of :class:`MultiDict` that defaults to
        replace the old value instead of appending it. """

    def __init__(self, *a, **ka):
        self.dict = {}
        if a or ka: self.update(*a, **ka)

    def __contains__(self, key): return _hkey(key) in self.dict
    def __delitem__(self, key): del self.dict[_hkey(key)]
    def __getitem__(self, key): return self.dict[_hkey(key)][-1]
    def __setitem__(self, key, value): self.dict[_hkey(key)] = [str(value)]
    def append(self, key, value):
        self.dict.setdefault(_hkey(key), []).append(str(value))
    def replace(self, key, value): self.dict[_hkey(key)] = [str(value)]
    def getall(self, key): return self.dict.get(_hkey(key)) or []
    def get(self, key, default=None, index=-1):
        return MultiDict.get(self, _hkey(key), default, index)
    def filter(self, names):
        for name in map(_hkey, names):
            if name in self.dict:
                del self.dict[name]


class WSGIHeaderDict(DictMixin):
    ''' This dict-like class wraps a WSGI environ dict and provides convenient
        access to HTTP_* fields. Keys and values are native strings
        (2.x bytes or 3.x unicode) and keys are case-insensitive. If the WSGI
        environment contains non-native string values, these are de- or encoded
        using a lossless 'latin1' character set.

        The API will remain stable even on changes to the relevant PEPs.
        Currently PEP 333, 444 and 3333 are supported. (PEP 444 is the only one
        that uses non-native strings.)
    '''
    #: List of keys that do not have a 'HTTP_' prefix.
    cgikeys = ('CONTENT_TYPE', 'CONTENT_LENGTH')

    def __init__(self, environ):
        self.environ = environ

    def _ekey(self, key):
        ''' Translate header field name to CGI/WSGI environ key. '''
        key = key.replace('-','_').upper()
        if key in self.cgikeys:
            return key
        return 'HTTP_' + key

    def raw(self, key, default=None):
        ''' Return the header value as is (may be bytes or unicode). '''
        return self.environ.get(self._ekey(key), default)

    def __getitem__(self, key):
        return tonat(self.environ[self._ekey(key)], 'latin1')

    def __setitem__(self, key, value):
        raise TypeError("%s is read-only." % self.__class__)

    def __delitem__(self, key):
        raise TypeError("%s is read-only." % self.__class__)

    def __iter__(self):
        for key in self.environ:
            if key[:5] == 'HTTP_':
                yield key[5:].replace('_', '-').title()
            elif key in self.cgikeys:
                yield key.replace('_', '-').title()

    def keys(self): return [x for x in self]
    def __len__(self): return len(self.keys())
    def __contains__(self, key): return self._ekey(key) in self.environ


class AppStack(list):
    """ A stack-like list. Calling it returns the head of the stack. """

    def __call__(self):
        """ Return the current default application. """
        return self[-1]

    def push(self, value=None):
        """ Add a new :class:`Bottle` instance to the stack """
        if not isinstance(value, Bottle):
            value = Bottle()
        self.append(value)
        return value


class WSGIFileWrapper(object):

   def __init__(self, fp, buffer_size=1024*64):
       self.fp, self.buffer_size = fp, buffer_size
       for attr in ('fileno', 'close', 'read', 'readlines'):
           if hasattr(fp, attr): setattr(self, attr, getattr(fp, attr))

   def __iter__(self):
       read, buff = self.fp.read, self.buffer_size
       while True:
           part = read(buff)
           if not part: break
           yield part






###############################################################################
# Application Helper ###########################################################
###############################################################################


def abort(code=500, text='Unknown Error: Application stopped.'):
    """ Aborts execution and causes a HTTP error. """
    raise HTTPError(code, text)


def redirect(url, code=None):
    """ Aborts execution and causes a 303 or 302 redirect, depending on
        the HTTP protocol version. """
    if code is None:
        code = 303 if request.get('SERVER_PROTOCOL') == "HTTP/1.1" else 302
    location = urljoin(request.url, url)
    raise HTTPResponse("", status=code, header=dict(Location=location))


def static_file(filename, root, mimetype='auto', download=False):
    """ Open a file in a safe way and return :exc:`HTTPResponse` with status
        code 200, 305, 401 or 404. Set Content-Type, Content-Encoding,
        Content-Length and Last-Modified header. Obey If-Modified-Since header
        and HEAD requests.
    """
    root = os.path.abspath(root) + os.sep
    filename = os.path.abspath(os.path.join(root, filename.strip('/\\')))
    header = dict()

    if not filename.startswith(root):
        return HTTPError(403, "Access denied.")
    if not os.path.exists(filename) or not os.path.isfile(filename):
        return HTTPError(404, "File does not exist.")
    if not os.access(filename, os.R_OK):
        return HTTPError(403, "You do not have permission to access this file.")

    if mimetype == 'auto':
        mimetype, encoding = mimetypes.guess_type(filename)
        if mimetype: header['Content-Type'] = mimetype
        if encoding: header['Content-Encoding'] = encoding
    elif mimetype:
        header['Content-Type'] = mimetype

    if download:
        download = os.path.basename(filename if download == True else download)
        header['Content-Disposition'] = 'attachment; filename="%s"' % download

    stats = os.stat(filename)
    header['Content-Length'] = stats.st_size
    lm = time.strftime("%a, %d %b %Y %H:%M:%S GMT", time.gmtime(stats.st_mtime))
    header['Last-Modified'] = lm

    ims = request.environ.get('HTTP_IF_MODIFIED_SINCE')
    if ims:
        ims = parse_date(ims.split(";")[0].strip())
    if ims is not None and ims >= int(stats.st_mtime):
        header['Date'] = time.strftime("%a, %d %b %Y %H:%M:%S GMT", time.gmtime())
        return HTTPResponse(status=304, header=header)

    body = '' if request.method == 'HEAD' else open(filename, 'rb')
    return HTTPResponse(body, header=header)






###############################################################################
# HTTP Utilities and MISC (TODO) ###############################################
###############################################################################


def debug(mode=True):
    """ Change the debug level.
    There is only one debug level supported at the moment."""
    global DEBUG
    DEBUG = bool(mode)


def parse_date(ims):
    """ Parse rfc1123, rfc850 and asctime timestamps and return UTC epoch. """
    try:
        ts = email.utils.parsedate_tz(ims)
        return time.mktime(ts[:8] + (0,)) - (ts[9] or 0) - time.timezone
    except (TypeError, ValueError, IndexError, OverflowError):
        return None


def parse_auth(header):
    """ Parse rfc2617 HTTP authentication header string (basic) and return (user,pass) tuple or None"""
    try:
        method, data = header.split(None, 1)
        if method.lower() == 'basic':
            #TODO: Add 2to3 save base64[encode/decode] functions.
            user, pwd = touni(base64.b64decode(tob(data))).split(':',1)
            return user, pwd
    except (KeyError, ValueError):
        return None


def _lscmp(a, b):
    ''' Compares two strings in a cryptographically save way:
        Runtime is not affected by length of common prefix. '''
    return not sum(0 if x==y else 1 for x, y in zip(a, b)) and len(a) == len(b)


def cookie_encode(data, key):
    ''' Encode and sign a pickle-able object. Return a (byte) string '''
    msg = base64.b64encode(pickle.dumps(data, -1))
    sig = base64.b64encode(hmac.new(key, msg).digest())
    return tob('!') + sig + tob('?') + msg


def cookie_decode(data, key):
    ''' Verify and decode an encoded string. Return an object or None.'''
    data = tob(data)
    if cookie_is_encoded(data):
        sig, msg = data.split(tob('?'), 1)
        if _lscmp(sig[1:], base64.b64encode(hmac.new(key, msg).digest())):
            return pickle.loads(base64.b64decode(msg))
    return None


def cookie_is_encoded(data):
    ''' Return True if the argument looks like a encoded cookie.'''
    return bool(data.startswith(tob('!')) and tob('?') in data)


def html_escape(string):
    ''' Escape HTML special characters ``&<>`` and quotes ``'"``. '''
    return string.replace('&','&amp;').replace('<','&lt;').replace('>','&gt;')\
                 .replace('"','&quot;').replace("'",'&#039;')


def html_quote(string):
    ''' Escape and quote a string to be used as an HTTP attribute.'''
    return '"%s"' % html_escape(string).replace('\n','%#10;')\
                    .replace('\r','&#13;').replace('\t','&#9;')


def yieldroutes(func):
    """ Return a generator for routes that match the signature (name, args)
    of the func parameter. This may yield more than one route if the function
    takes optional keyword arguments. The output is best described by example::

        a()         -> '/a'
        b(x, y)     -> '/b/:x/:y'
        c(x, y=5)   -> '/c/:x' and '/c/:x/:y'
        d(x=5, y=6) -> '/d' and '/d/:x' and '/d/:x/:y'
    """
    import inspect # Expensive module. Only import if necessary.
    path = '/' + func.__name__.replace('__','/').lstrip('/')
    spec = inspect.getargspec(func)
    argc = len(spec[0]) - len(spec[3] or [])
    path += ('/:%s' * argc) % tuple(spec[0][:argc])
    yield path
    for arg in spec[0][argc:]:
        path += '/:%s' % arg
        yield path


def path_shift(script_name, path_info, shift=1):
    ''' Shift path fragments from PATH_INFO to SCRIPT_NAME and vice versa.

        :return: The modified paths.
        :param script_name: The SCRIPT_NAME path.
        :param script_name: The PATH_INFO path.
        :param shift: The number of path fragments to shift. May be negative to
          change the shift direction. (default: 1)
    '''
    if shift == 0: return script_name, path_info
    pathlist = path_info.strip('/').split('/')
    scriptlist = script_name.strip('/').split('/')
    if pathlist and pathlist[0] == '': pathlist = []
    if scriptlist and scriptlist[0] == '': scriptlist = []
    if shift > 0 and shift <= len(pathlist):
        moved = pathlist[:shift]
        scriptlist = scriptlist + moved
        pathlist = pathlist[shift:]
    elif shift < 0 and shift >= -len(scriptlist):
        moved = scriptlist[shift:]
        pathlist = moved + pathlist
        scriptlist = scriptlist[:shift]
    else:
        empty = 'SCRIPT_NAME' if shift < 0 else 'PATH_INFO'
        raise AssertionError("Cannot shift. Nothing left from %s" % empty)
    new_script_name = '/' + '/'.join(scriptlist)
    new_path_info = '/' + '/'.join(pathlist)
    if path_info.endswith('/') and pathlist: new_path_info += '/'
    return new_script_name, new_path_info


def validate(**vkargs):
    """
    Validates and manipulates keyword arguments by user defined callables.
    Handles ValueError and missing arguments by raising HTTPError(403).
    """
    def decorator(func):
        def wrapper(**kargs):
            for key, value in vkargs.iteritems():
                if key not in kargs:
                    abort(403, 'Missing parameter: %s' % key)
                try:
                    kargs[key] = value(kargs[key])
                except ValueError:
                    abort(403, 'Wrong parameter format for: %s' % key)
            return func(**kargs)
        return wrapper
    return decorator


def auth_basic(check, realm="private", text="Access denied"):
    ''' Callback decorator to require HTTP auth (basic).
        TODO: Add route(check_auth=...) parameter. '''
    def decorator(func):
      def wrapper(*a, **ka):
        user, password = request.auth or (None, None)
        if user is None or not check(user, password):
          response.headers['WWW-Authenticate'] = 'Basic realm="%s"' % realm
          return HTTPError(401, text)
        return func(*a, **ka)
      return wrapper
    return decorator


def make_default_app_wrapper(name):
    ''' Return a callable that relays calls to the current default app. '''
    @functools.wraps(getattr(Bottle, name))
    def wrapper(*a, **ka):
        return getattr(app(), name)(*a, **ka)
    return wrapper


for name in '''route get post put delete error mount
               hook install uninstall'''.split():
    globals()[name] = make_default_app_wrapper(name)
url = make_default_app_wrapper('get_url')
del name






###############################################################################
# Server Adapter ###############################################################
###############################################################################


class ServerAdapter(object):
    quiet = False
    def __init__(self, host='127.0.0.1', port=8080, **config):
        self.options = config
        self.host = host
        self.port = int(port)

    def run(self, handler): # pragma: no cover
        pass

    def __repr__(self):
        args = ', '.join(['%s=%s'%(k,repr(v)) for k, v in self.options.items()])
        return "%s(%s)" % (self.__class__.__name__, args)


class CGIServer(ServerAdapter):
    quiet = True
    def run(self, handler): # pragma: no cover
        from wsgiref.handlers import CGIHandler
        def fixed_environ(environ, start_response):
            environ.setdefault('PATH_INFO', '')
            return handler(environ, start_response)
        CGIHandler().run(fixed_environ)


class FlupFCGIServer(ServerAdapter):
    def run(self, handler): # pragma: no cover
        import flup.server.fcgi
        self.options.setdefault('bindAddress', (self.host, self.port))
        flup.server.fcgi.WSGIServer(handler, **self.options).run()


class WSGIRefServer(ServerAdapter):
    def run(self, handler): # pragma: no cover
        from wsgiref.simple_server import make_server, WSGIRequestHandler
        if self.quiet:
            class QuietHandler(WSGIRequestHandler):
                def log_request(*args, **kw): pass
            self.options['handler_class'] = QuietHandler
        srv = make_server(self.host, self.port, handler, **self.options)
        srv.serve_forever()


class CherryPyServer(ServerAdapter):
    def run(self, handler): # pragma: no cover
        from cherrypy import wsgiserver
        server = wsgiserver.CherryPyWSGIServer((self.host, self.port), handler)
        try:
            server.start()
        finally:
            server.stop()


class PasteServer(ServerAdapter):
    def run(self, handler): # pragma: no cover
        from paste import httpserver
        if not self.quiet:
            from paste.translogger import TransLogger
            handler = TransLogger(handler)
        httpserver.serve(handler, host=self.host, port=str(self.port),
                         **self.options)


class MeinheldServer(ServerAdapter):
    def run(self, handler):
        from meinheld import server
        server.listen((self.host, self.port))
        server.run(handler)


class FapwsServer(ServerAdapter):
    """ Extremely fast webserver using libev. See http://www.fapws.org/ """
    def run(self, handler): # pragma: no cover
        import fapws._evwsgi as evwsgi
        from fapws import base, config
        port = self.port
        if float(config.SERVER_IDENT[-2:]) > 0.4:
            # fapws3 silently changed its API in 0.5
            port = str(port)
        evwsgi.start(self.host, port)
        # fapws3 never releases the GIL. Complain upstream. I tried. No luck.
        if 'BOTTLE_CHILD' in os.environ and not self.quiet:
            print "WARNING: Auto-reloading does not work with Fapws3."
            print "         (Fapws3 breaks python thread support)"
        evwsgi.set_base_module(base)
        def app(environ, start_response):
            environ['wsgi.multiprocess'] = False
            return handler(environ, start_response)
        evwsgi.wsgi_cb(('', app))
        evwsgi.run()


class TornadoServer(ServerAdapter):
    """ The super hyped asynchronous server by facebook. Untested. """
    def run(self, handler): # pragma: no cover
        import tornado.wsgi
        import tornado.httpserver
        import tornado.ioloop
        container = tornado.wsgi.WSGIContainer(handler)
        server = tornado.httpserver.HTTPServer(container)
        server.listen(port=self.port)
        tornado.ioloop.IOLoop.instance().start()


class AppEngineServer(ServerAdapter):
    """ Adapter for Google App Engine. """
    quiet = True
    def run(self, handler):
        from google.appengine.ext.webapp import util
        # A main() function in the handler script enables 'App Caching'.
        # Lets makes sure it is there. This _really_ improves performance.
        module = sys.modules.get('__main__')
        if module and not hasattr(module, 'main'):
            module.main = lambda: util.run_wsgi_app(handler)
        util.run_wsgi_app(handler)


class TwistedServer(ServerAdapter):
    """ Untested. """
    def run(self, handler):
        from twisted.web import server, wsgi
        from twisted.python.threadpool import ThreadPool
        from twisted.internet import reactor
        thread_pool = ThreadPool()
        thread_pool.start()
        reactor.addSystemEventTrigger('after', 'shutdown', thread_pool.stop)
        factory = server.Site(wsgi.WSGIResource(reactor, thread_pool, handler))
        reactor.listenTCP(self.port, factory, interface=self.host)
        reactor.run()


class DieselServer(ServerAdapter):
    """ Untested. """
    def run(self, handler):
        from diesel.protocols.wsgi import WSGIApplication
        app = WSGIApplication(handler, port=self.port)
        app.run()


class GeventServer(ServerAdapter):
    """ Untested. Options:

        * `monkey` (default: True) fixes the stdlib to use greenthreads.
        * `fast` (default: False) uses libevent's http server, but has some
          issues: No streaming, no pipelining, no SSL.
    """
    def run(self, handler):
        from gevent import wsgi as wsgi_fast, pywsgi, monkey, local
        if self.options.get('monkey', True):
            if not threading.local is local.local: monkey.patch_all()
        wsgi = wsgi_fast if self.options.get('fast') else pywsgi
        wsgi.WSGIServer((self.host, self.port), handler).serve_forever()


class GunicornServer(ServerAdapter):
    """ Untested. """
    def run(self, handler):
        from gunicorn.arbiter import Arbiter
        from gunicorn.config import Config
        handler.cfg = Config({'bind': "%s:%d" % (self.host, self.port), 'workers': 4})
        arbiter = Arbiter(handler)
        arbiter.run()


class EventletServer(ServerAdapter):
    """ Untested """
    def run(self, handler):
        from eventlet import wsgi, listen
        wsgi.server(listen((self.host, self.port)), handler)


class RocketServer(ServerAdapter):
    """ Untested. """
    def run(self, handler):
        from rocket import Rocket
        server = Rocket((self.host, self.port), 'wsgi', { 'wsgi_app' : handler })
        server.start()


class BjoernServer(ServerAdapter):
    """ Fast server written in C: https://github.com/jonashaag/bjoern """
    def run(self, handler):
        from bjoern import run
        run(handler, self.host, self.port)


class AutoServer(ServerAdapter):
    """ Untested. """
    adapters = [PasteServer, CherryPyServer, TwistedServer, WSGIRefServer]
    def run(self, handler):
        for sa in self.adapters:
            try:
                return sa(self.host, self.port, **self.options).run(handler)
            except ImportError:
                pass

server_names = {
    'cgi': CGIServer,
    'flup': FlupFCGIServer,
    'wsgiref': WSGIRefServer,
    'cherrypy': CherryPyServer,
    'paste': PasteServer,
    'fapws3': FapwsServer,
    'tornado': TornadoServer,
    'gae': AppEngineServer,
    'twisted': TwistedServer,
    'diesel': DieselServer,
    'meinheld': MeinheldServer,
    'gunicorn': GunicornServer,
    'eventlet': EventletServer,
    'gevent': GeventServer,
    'rocket': RocketServer,
    'bjoern' : BjoernServer,
    'auto': AutoServer,
}






###############################################################################
# Application Control ##########################################################
###############################################################################


def load(target, **namespace):
    """ Import a module or fetch an object from a module.

        * ``package.module`` returns `module` as a module object.
        * ``pack.mod:name`` returns the module variable `name` from `pack.mod`.
        * ``pack.mod:func()`` calls `pack.mod.func()` and returns the result.

        The last form accepts not only function calls, but any type of
        expression. Keyword arguments passed to this function are available as
        local variables. Example: ``import_string('re:compile(x)', x='[a-z]')``
    """
    module, target = target.split(":", 1) if ':' in target else (target, None)
    if module not in sys.modules: __import__(module)
    if not target: return sys.modules[module]
    if target.isalnum(): return getattr(sys.modules[module], target)
    package_name = module.split('.')[0]
    namespace[package_name] = sys.modules[package_name]
    return eval('%s.%s' % (module, target), namespace)


def load_app(target):
    """ Load a bottle application from a module and make sure that the import
        does not affect the current default application, but returns a separate
        application object. See :func:`load` for details. """
    tmp = app.push() # Create a new "default application"
    rv = _load(target) # Import the target module
    app.remove(tmp) # Remove the temporary added default application
    return rv if isinstance(rv, Bottle) else tmp


def run(app=None, server='wsgiref', host='127.0.0.1', port=8080,
        interval=1, reloader=False, quiet=False, **kargs):
    """ Start a server instance. This method blocks until the server terminates.

        :param app: WSGI application or target string supported by
               :func:`load_app`. (default: :func:`default_app`)
        :param server: Server adapter to use. See :data:`server_names` keys
               for valid names or pass a :class:`ServerAdapter` subclass.
               (default: `wsgiref`)
        :param host: Server address to bind to. Pass ``0.0.0.0`` to listens on
               all interfaces including the external one. (default: 127.0.0.1)
        :param port: Server port to bind to. Values below 1024 require root
               privileges. (default: 8080)
        :param reloader: Start auto-reloading server? (default: False)
        :param interval: Auto-reloader interval in seconds (default: 1)
        :param quiet: Suppress output to stdout and stderr? (default: False)
        :param options: Options passed to the server adapter.
     """
    app = app or default_app()
    if isinstance(app, basestring):
        app = load_app(app)
    if isinstance(server, basestring):
        server = server_names.get(server)
    if isinstance(server, type):
        server = server(host=host, port=port, **kargs)
    if not isinstance(server, ServerAdapter):
        raise RuntimeError("Server must be a subclass of ServerAdapter")
    server.quiet = server.quiet or quiet
    if not server.quiet and not os.environ.get('BOTTLE_CHILD'):
        print "Bottle server starting up (using %s)..." % repr(server)
        print "Listening on http://%s:%d/" % (server.host, server.port)
        print "Use Ctrl-C to quit."
        print
    try:
        if reloader:
            interval = min(interval, 1)
            if os.environ.get('BOTTLE_CHILD'):
                _reloader_child(server, app, interval)
            else:
                _reloader_observer(server, app, interval)
        else:
            server.run(app)
    except KeyboardInterrupt:
        pass
    if not server.quiet and not os.environ.get('BOTTLE_CHILD'):
        print "Shutting down..."


class FileCheckerThread(threading.Thread):
    ''' Thread that periodically checks for changed module files. '''

    def __init__(self, lockfile, interval):
        threading.Thread.__init__(self)
        self.lockfile, self.interval = lockfile, interval
        #1: lockfile to old; 2: lockfile missing
        #3: module file changed; 5: external exit
        self.status = 0

    def run(self):
        exists = os.path.exists
        mtime = lambda path: os.stat(path).st_mtime
        files = dict()
        for module in sys.modules.values():
            path = getattr(module, '__file__', '')
            if path[-4:] in ('.pyo', '.pyc'): path = path[:-1]
            if path and exists(path): files[path] = mtime(path)
        while not self.status:
            for path, lmtime in files.iteritems():
                if not exists(path) or mtime(path) > lmtime:
                    self.status = 3
            if not exists(self.lockfile):
                self.status = 2
            elif mtime(self.lockfile) < time.time() - self.interval - 5:
                self.status = 1
            if not self.status:
                time.sleep(self.interval)
        if self.status != 5:
            thread.interrupt_main()


def _reloader_child(server, app, interval):
    ''' Start the server and check for modified files in a background thread.
        As soon as an update is detected, KeyboardInterrupt is thrown in
        the main thread to exit the server loop. The process exists with status
        code 3 to request a reload by the observer process. If the lockfile
        is not modified in 2*interval second or missing, we assume that the
        observer process died and exit with status code 1 or 2.
    '''
    lockfile = os.environ.get('BOTTLE_LOCKFILE')
    bgcheck = FileCheckerThread(lockfile, interval)
    try:
        bgcheck.start()
        server.run(app)
    except KeyboardInterrupt:
        pass
    bgcheck.status, status = 5, bgcheck.status
    bgcheck.join() # bgcheck.status == 5 --> silent exit
    if status: sys.exit(status)


def _reloader_observer(server, app, interval):
    ''' Start a child process with identical commandline arguments and restart
        it as long as it exists with status code 3. Also create a lockfile and
        touch it (update mtime) every interval seconds.
    '''
    fd, lockfile = tempfile.mkstemp(prefix='bottle-reloader.', suffix='.lock')
    os.close(fd) # We only need this file to exist. We never write to it
    try:
        while os.path.exists(lockfile):
            args = [sys.executable] + sys.argv
            environ = os.environ.copy()
            environ['BOTTLE_CHILD'] = 'true'
            environ['BOTTLE_LOCKFILE'] = lockfile
            p = subprocess.Popen(args, env=environ)
            while p.poll() is None: # Busy wait...
                os.utime(lockfile, None) # I am alive!
                time.sleep(interval)
            if p.poll() != 3:
                if os.path.exists(lockfile): os.unlink(lockfile)
                sys.exit(p.poll())
            elif not server.quiet:
                print "Reloading server..."
    except KeyboardInterrupt:
        pass
    if os.path.exists(lockfile): os.unlink(lockfile)






###############################################################################
# Template Adapters ############################################################
###############################################################################


class TemplateError(HTTPError):
    def __init__(self, message):
        HTTPError.__init__(self, 500, message)


class BaseTemplate(object):
    """ Base class and minimal API for template adapters """
    extensions = ['tpl','html','thtml','stpl']
    settings = {} #used in prepare()
    defaults = {} #used in render()

    def __init__(self, source=None, name=None, lookup=[], encoding='utf8', **settings):
        """ Create a new template.
        If the source parameter (str or buffer) is missing, the name argument
        is used to guess a template filename. Subclasses can assume that
        self.source and/or self.filename are set. Both are strings.
        The lookup, encoding and settings parameters are stored as instance
        variables.
        The lookup parameter stores a list containing directory paths.
        The encoding parameter should be used to decode byte strings or files.
        The settings parameter contains a dict for engine-specific settings.
        """
        self.name = name
        self.source = source.read() if hasattr(source, 'read') else source
        self.filename = source.filename if hasattr(source, 'filename') else None
        self.lookup = map(os.path.abspath, lookup)
        self.encoding = encoding
        self.settings = self.settings.copy() # Copy from class variable
        self.settings.update(settings) # Apply
        if not self.source and self.name:
            self.filename = self.search(self.name, self.lookup)
            if not self.filename:
                raise TemplateError('Template %s not found.' % repr(name))
        if not self.source and not self.filename:
            raise TemplateError('No template specified.')
        self.prepare(**self.settings)

    @classmethod
    def search(cls, name, lookup=[]):
        """ Search name in all directories specified in lookup.
        First without, then with common extensions. Return first hit. """
        if os.path.isfile(name): return name
        for spath in lookup:
            fname = os.path.join(spath, name)
            if os.path.isfile(fname):
                return fname
            for ext in cls.extensions:
                if os.path.isfile('%s.%s' % (fname, ext)):
                    return '%s.%s' % (fname, ext)

    @classmethod
    def global_config(cls, key, *args):
        ''' This reads or sets the global settings stored in class.settings. '''
        if args:
            cls.settings[key] = args[0]
        else:
            return cls.settings[key]

    def prepare(self, **options):
        """ Run preparations (parsing, caching, ...).
        It should be possible to call this again to refresh a template or to
        update settings.
        """
        raise NotImplementedError

    def render(self, *args, **kwargs):
        """ Render the template with the specified local variables and return
        a single byte or unicode string. If it is a byte string, the encoding
        must match self.encoding. This method must be thread-safe!
        Local variables may be provided in dictionaries (*args)
        or directly, as keywords (**kwargs).
        """
        raise NotImplementedError


class MakoTemplate(BaseTemplate):
    def prepare(self, **options):
        from mako.template import Template
        from mako.lookup import TemplateLookup
        options.update({'input_encoding':self.encoding})
        options.setdefault('format_exceptions', bool(DEBUG))
        lookup = TemplateLookup(directories=self.lookup, **options)
        if self.source:
            self.tpl = Template(self.source, lookup=lookup, **options)
        else:
            self.tpl = Template(uri=self.name, filename=self.filename, lookup=lookup, **options)

    def render(self, *args, **kwargs):
        for dictarg in args: kwargs.update(dictarg)
        _defaults = self.defaults.copy()
        _defaults.update(kwargs)
        return self.tpl.render(**_defaults)


class CheetahTemplate(BaseTemplate):
    def prepare(self, **options):
        from Cheetah.Template import Template
        self.context = threading.local()
        self.context.vars = {}
        options['searchList'] = [self.context.vars]
        if self.source:
            self.tpl = Template(source=self.source, **options)
        else:
            self.tpl = Template(file=self.filename, **options)

    def render(self, *args, **kwargs):
        for dictarg in args: kwargs.update(dictarg)
        self.context.vars.update(self.defaults)
        self.context.vars.update(kwargs)
        out = str(self.tpl)
        self.context.vars.clear()
        return out


class Jinja2Template(BaseTemplate):
    def prepare(self, filters=None, tests=None, **kwargs):
        from jinja2 import Environment, FunctionLoader
        if 'prefix' in kwargs: # TODO: to be removed after a while
            raise RuntimeError('The keyword argument `prefix` has been removed. '
                'Use the full jinja2 environment name line_statement_prefix instead.')
        self.env = Environment(loader=FunctionLoader(self.loader), **kwargs)
        if filters: self.env.filters.update(filters)
        if tests: self.env.tests.update(tests)
        if self.source:
            self.tpl = self.env.from_string(self.source)
        else:
            self.tpl = self.env.get_template(self.filename)

    def render(self, *args, **kwargs):
        for dictarg in args: kwargs.update(dictarg)
        _defaults = self.defaults.copy()
        _defaults.update(kwargs)
        return self.tpl.render(**_defaults)

    def loader(self, name):
        fname = self.search(name, self.lookup)
        if fname:
            with open(fname, "rb") as f:
                return f.read().decode(self.encoding)


class SimpleTALTemplate(BaseTemplate):
    ''' Untested! '''
    def prepare(self, **options):
        from simpletal import simpleTAL
        # TODO: add option to load METAL files during render
        if self.source:
            self.tpl = simpleTAL.compileHTMLTemplate(self.source)
        else:
            with open(self.filename, 'rb') as fp:
                self.tpl = simpleTAL.compileHTMLTemplate(tonat(fp.read()))

    def render(self, *args, **kwargs):
        from simpletal import simpleTALES
        from StringIO import StringIO
        for dictarg in args: kwargs.update(dictarg)
        # TODO: maybe reuse a context instead of always creating one
        context = simpleTALES.Context()
        for k,v in self.defaults.items():
            context.addGlobal(k, v)
        for k,v in kwargs.items():
            context.addGlobal(k, v)
        output = StringIO()
        self.tpl.expand(context, output)
        return output.getvalue()


class SimpleTemplate(BaseTemplate):
    blocks = ('if', 'elif', 'else', 'try', 'except', 'finally', 'for', 'while',
              'with', 'def', 'class')
    dedent_blocks = ('elif', 'else', 'except', 'finally')

    @lazy_attribute
    def re_pytokens(cls):
        ''' This matches comments and all kinds of quoted strings but does
            NOT match comments (#...) within quoted strings. (trust me) '''
        return re.compile(r'''
            (''(?!')|""(?!")|'{6}|"{6}    # Empty strings (all 4 types)
             |'(?:[^\\']|\\.)+?'          # Single quotes (')
             |"(?:[^\\"]|\\.)+?"          # Double quotes (")
             |'{3}(?:[^\\]|\\.|\n)+?'{3}  # Triple-quoted strings (')
             |"{3}(?:[^\\]|\\.|\n)+?"{3}  # Triple-quoted strings (")
             |\#.*                        # Comments
            )''', re.VERBOSE)

    def prepare(self, escape_func=html_escape, noescape=False, **kwargs):
        self.cache = {}
        enc = self.encoding
        self._str = lambda x: touni(x, enc)
        self._escape = lambda x: escape_func(touni(x, enc))
        if noescape:
            self._str, self._escape = self._escape, self._str

    @classmethod
    def split_comment(cls, code):
        """ Removes comments (#...) from python code. """
        if '#' not in code: return code
        #: Remove comments only (leave quoted strings as they are)
        subf = lambda m: '' if m.group(0)[0]=='#' else m.group(0)
        return re.sub(cls.re_pytokens, subf, code)

    @cached_property
    def co(self):
        return compile(self.code, self.filename or '<string>', 'exec')

    @cached_property
    def code(self):
        stack = [] # Current Code indentation
        lineno = 0 # Current line of code
        ptrbuffer = [] # Buffer for printable strings and token tuple instances
        codebuffer = [] # Buffer for generated python code
        multiline = dedent = oneline = False
        template = self.source if self.source else open(self.filename).read()

        def yield_tokens(line):
            for i, part in enumerate(re.split(r'\{\{(.*?)\}\}', line)):
                if i % 2:
                    if part.startswith('!'): yield 'RAW', part[1:]
                    else: yield 'CMD', part
                else: yield 'TXT', part

        def flush(): # Flush the ptrbuffer
            if not ptrbuffer: return
            cline = ''
            for line in ptrbuffer:
                for token, value in line:
                    if token == 'TXT': cline += repr(value)
                    elif token == 'RAW': cline += '_str(%s)' % value
                    elif token == 'CMD': cline += '_escape(%s)' % value
                    cline +=  ', '
                cline = cline[:-2] + '\\\n'
            cline = cline[:-2]
            if cline[:-1].endswith('\\\\\\\\\\n'):
                cline = cline[:-7] + cline[-1] # 'nobr\\\\\n' --> 'nobr'
            cline = '_printlist([' + cline + '])'
            del ptrbuffer[:] # Do this before calling code() again
            code(cline)

        def code(stmt):
            for line in stmt.splitlines():
                codebuffer.append('  ' * len(stack) + line.strip())

        for line in template.splitlines(True):
            lineno += 1
            line = line if isinstance(line, unicode)\
                        else unicode(line, encoding=self.encoding)
            if lineno <= 2:
                m = re.search(r"%.*coding[:=]\s*([-\w\.]+)", line)
                if m: self.encoding = m.group(1)
                if m: line = line.replace('coding','coding (removed)')
            if line.strip()[:2].count('%') == 1:
                line = line.split('%',1)[1].lstrip() # Full line following the %
                cline = self.split_comment(line).strip()
                cmd = re.split(r'[^a-zA-Z0-9_]', cline)[0]
                flush() # You are actually reading this? Good luck, it's a mess :)
                if cmd in self.blocks or multiline:
                    cmd = multiline or cmd
                    dedent = cmd in self.dedent_blocks # "else:"
                    if dedent and not oneline and not multiline:
                        cmd = stack.pop()
                    code(line)
                    oneline = not cline.endswith(':') # "if 1: pass"
                    multiline = cmd if cline.endswith('\\') else False
                    if not oneline and not multiline:
                        stack.append(cmd)
                elif cmd == 'end' and stack:
                    code('#end(%s) %s' % (stack.pop(), line.strip()[3:]))
                elif cmd == 'include':
                    p = cline.split(None, 2)[1:]
                    if len(p) == 2:
                        code("_=_include(%s, _stdout, %s)" % (repr(p[0]), p[1]))
                    elif p:
                        code("_=_include(%s, _stdout)" % repr(p[0]))
                    else: # Empty %include -> reverse of %rebase
                        code("_printlist(_base)")
                elif cmd == 'rebase':
                    p = cline.split(None, 2)[1:]
                    if len(p) == 2:
                        code("globals()['_rebase']=(%s, dict(%s))" % (repr(p[0]), p[1]))
                    elif p:
                        code("globals()['_rebase']=(%s, {})" % repr(p[0]))
                else:
                    code(line)
            else: # Line starting with text (not '%') or '%%' (escaped)
                if line.strip().startswith('%%'):
                    line = line.replace('%%', '%', 1)
                ptrbuffer.append(yield_tokens(line))
        flush()
        return '\n'.join(codebuffer) + '\n'

    def subtemplate(self, _name, _stdout, *args, **kwargs):
        for dictarg in args: kwargs.update(dictarg)
        if _name not in self.cache:
            self.cache[_name] = self.__class__(name=_name, lookup=self.lookup)
        return self.cache[_name].execute(_stdout, kwargs)

    def execute(self, _stdout, *args, **kwargs):
        for dictarg in args: kwargs.update(dictarg)
        env = self.defaults.copy()
        env.update({'_stdout': _stdout, '_printlist': _stdout.extend,
               '_include': self.subtemplate, '_str': self._str,
               '_escape': self._escape})
        env.update(kwargs)
        eval(self.co, env)
        if '_rebase' in env:
            subtpl, rargs = env['_rebase']
            subtpl = self.__class__(name=subtpl, lookup=self.lookup)
            rargs['_base'] = _stdout[:] #copy stdout
            del _stdout[:] # clear stdout
            return subtpl.execute(_stdout, rargs)
        return env

    def render(self, *args, **kwargs):
        """ Render the template using keyword arguments as local variables. """
        for dictarg in args: kwargs.update(dictarg)
        stdout = []
        self.execute(stdout, kwargs)
        return ''.join(stdout)


def template(*args, **kwargs):
    '''
    Get a rendered template as a string iterator.
    You can use a name, a filename or a template string as first parameter.
    Template rendering arguments can be passed as dictionaries
    or directly (as keyword arguments).
    '''
    tpl = args[0] if args else None
    template_adapter = kwargs.pop('template_adapter', SimpleTemplate)
    if tpl not in TEMPLATES or DEBUG:
        settings = kwargs.pop('template_settings', {})
        lookup = kwargs.pop('template_lookup', TEMPLATE_PATH)
        if isinstance(tpl, template_adapter):
            TEMPLATES[tpl] = tpl
            if settings: TEMPLATES[tpl].prepare(**settings)
        elif "\n" in tpl or "{" in tpl or "%" in tpl or '$' in tpl:
            TEMPLATES[tpl] = template_adapter(source=tpl, lookup=lookup, **settings)
        else:
            TEMPLATES[tpl] = template_adapter(name=tpl, lookup=lookup, **settings)
    if not TEMPLATES[tpl]:
        abort(500, 'Template (%s) not found' % tpl)
    for dictarg in args[1:]: kwargs.update(dictarg)
    return TEMPLATES[tpl].render(kwargs)

mako_template = functools.partial(template, template_adapter=MakoTemplate)
cheetah_template = functools.partial(template, template_adapter=CheetahTemplate)
jinja2_template = functools.partial(template, template_adapter=Jinja2Template)
simpletal_template = functools.partial(template, template_adapter=SimpleTALTemplate)


def view(tpl_name, **defaults):
    ''' Decorator: renders a template for a handler.
        The handler can control its behavior like that:

          - return a dict of template vars to fill out the template
          - return something other than a dict and the view decorator will not
            process the template, but return the handler result as is.
            This includes returning a HTTPResponse(dict) to get,
            for instance, JSON with autojson or other castfilters.
    '''
    def decorator(func):
        @functools.wraps(func)
        def wrapper(*args, **kwargs):
            result = func(*args, **kwargs)
            if isinstance(result, (dict, DictMixin)):
                tplvars = defaults.copy()
                tplvars.update(result)
                return template(tpl_name, **tplvars)
            return result
        return wrapper
    return decorator

mako_view = functools.partial(view, template_adapter=MakoTemplate)
cheetah_view = functools.partial(view, template_adapter=CheetahTemplate)
jinja2_view = functools.partial(view, template_adapter=Jinja2Template)
simpletal_view = functools.partial(view, template_adapter=SimpleTALTemplate)






###############################################################################
# Constants and Globals ########################################################
###############################################################################


TEMPLATE_PATH = ['./', './views/']
TEMPLATES = {}
DEBUG = False

#: A dict to map HTTP status codes (e.g. 404) to phrases (e.g. 'Not Found')
HTTP_CODES = httplib.responses
HTTP_CODES[418] = "I'm a teapot" # RFC 2324
_HTTP_STATUS_LINES = dict((k, '%d %s'%(k,v)) for (k,v) in HTTP_CODES.iteritems())

#: The default template used for error pages. Override with @error()
ERROR_PAGE_TEMPLATE = """
%try:
    %from bottle import DEBUG, HTTP_CODES, request, touni
    %status_name = HTTP_CODES.get(e.status, 'Unknown').title()
    <!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
    <html>
        <head>
            <title>Error {{e.status}}: {{status_name}}</title>
            <style type="text/css">
              html {background-color: #eee; font-family: sans;}
              body {background-color: #fff; border: 1px solid #ddd;
                    padding: 15px; margin: 15px;}
              pre {background-color: #eee; border: 1px solid #ddd; padding: 5px;}
            </style>
        </head>
        <body>
            <h1>Error {{e.status}}: {{status_name}}</h1>
            <p>Sorry, the requested URL <tt>{{repr(request.url)}}</tt>
               caused an error:</p>
            <pre>{{e.output}}</pre>
            %if DEBUG and e.exception:
              <h2>Exception:</h2>
              <pre>{{repr(e.exception)}}</pre>
            %end
            %if DEBUG and e.traceback:
              <h2>Traceback:</h2>
              <pre>{{e.traceback}}</pre>
            %end
        </body>
    </html>
%except ImportError:
    <b>ImportError:</b> Could not generate the error page. Please add bottle to
    the import path.
%end
"""

#: A thread-save instance of :class:`Request` representing the `current` request.
request = Request()

#: A thread-save instance of :class:`Response` used to build the HTTP response.
response = Response()

#: A thread-save namepsace. Not used by Bottle.
local = threading.local()

# Initialize app stack (create first empty Bottle app)
# BC: 0.6.4 and needed for run()
app = default_app = AppStack()
app.push()

#: A virtual package that redirects import statements.
#: Example: ``import bottle.ext.sqlite`` actually imports `bottle_sqlite`.
ext = _ImportRedirect(__name__+'.ext', 'bottle_%s').module

# THE END