blob: 690608998b05fcc2f9aec58a8bfed71c2f168317 [file] [log] [blame]
# (c) 2005 Ian Bicking and contributors; written for Paste
# (http://pythonpaste.org)
# Licensed under the MIT license:
# http://www.opensource.org/licenses/mit-license.php Also licenced under the
# Apache License, 2.0: http://opensource.org/licenses/apache2.0.php Licensed to
# PSF under a Contributor Agreement
"""
Middleware to check for obedience to the WSGI specification.
Some of the things this checks:
* Signature of the application and start_response (including that
keyword arguments are not used).
* Environment checks:
- Environment is a dictionary (and not a subclass).
- That all the required keys are in the environment: REQUEST_METHOD,
SERVER_NAME, SERVER_PORT, wsgi.version, wsgi.input, wsgi.errors,
wsgi.multithread, wsgi.multiprocess, wsgi.run_once
- That HTTP_CONTENT_TYPE and HTTP_CONTENT_LENGTH are not in the
environment (these headers should appear as CONTENT_LENGTH and
CONTENT_TYPE).
- Warns if QUERY_STRING is missing, as the cgi module acts
unpredictably in that case.
- That CGI-style variables (that don't contain a .) have
(non-unicode) string values
- That wsgi.version is a tuple
- That wsgi.url_scheme is 'http' or 'https' (@@: is this too
restrictive?)
- Warns if the REQUEST_METHOD is not known (@@: probably too
restrictive).
- That SCRIPT_NAME and PATH_INFO are empty or start with /
- That at least one of SCRIPT_NAME or PATH_INFO are set.
- That CONTENT_LENGTH is a positive integer.
- That SCRIPT_NAME is not '/' (it should be '', and PATH_INFO should
be '/').
- That wsgi.input has the methods read, readline, readlines, and
__iter__
- That wsgi.errors has the methods flush, write, writelines
* The status is a string, contains a space, starts with an integer,
and that integer is in range (> 100).
* That the headers is a list (not a subclass, not another kind of
sequence).
* That the items of the headers are tuples of strings.
* That there is no 'status' header (that is used in CGI, but not in
WSGI).
* That the headers don't contain newlines or colons, end in _ or -, or
contain characters codes below 037.
* That Content-Type is given if there is content (CGI often has a
default content type, but WSGI does not).
* That no Content-Type is given when there is no content (@@: is this
too restrictive?)
* That the exc_info argument to start_response is a tuple or None.
* That all calls to the writer are with strings, and no other methods
on the writer are accessed.
* That wsgi.input is used properly:
- .read() is called with zero or one argument
- That it returns a string
- That readline, readlines, and __iter__ return strings
- That .close() is not called
- No other methods are provided
* That wsgi.errors is used properly:
- .write() and .writelines() is called with a string, except
with python3
- That .close() is not called, and no other methods are provided.
* The response iterator:
- That it is not a string (it should be a list of a single string; a
string will work, but perform horribly).
- That .next() returns a string
- That the iterator is not iterated over until start_response has
been called (that can signal either a server or application
error).
- That .close() is called (doesn't raise exception, only prints to
sys.stderr, because we only know it isn't called when the object
is garbage collected).
"""
from __future__ import unicode_literals
import collections
import re
import warnings
from six import PY3
from six import binary_type
from six import string_types
header_re = re.compile(r'^[a-zA-Z][a-zA-Z0-9\-_]*$')
bad_header_value_re = re.compile(r'[\000-\037]')
valid_methods = (
'GET', 'HEAD', 'POST', 'OPTIONS', 'PUT', 'DELETE',
'TRACE', 'PATCH',
)
METADATA_TYPE = PY3 and (str, binary_type) or (str,)
# PEP-3333 says that environment variables must be "native strings",
# i.e. str(), which however is something *different* in py2 and py3.
SLASH = str('/')
def to_string(value):
if not isinstance(value, string_types):
return value.decode('latin1')
else:
return value
class WSGIWarning(Warning):
"""
Raised in response to WSGI-spec-related warnings
"""
def middleware(application, global_conf=None):
"""
When applied between a WSGI server and a WSGI application, this
middleware will check for WSGI compliancy on a number of levels.
This middleware does not modify the request or response in any
way, but will throw an AssertionError if anything seems off
(except for a failure to close the application iterator, which
will be printed to stderr -- there's no way to throw an exception
at that point).
"""
def lint_app(*args, **kw):
assert len(args) == 2, "Two arguments required"
assert not kw, "No keyword arguments allowed"
environ, start_response = args
check_environ(environ)
# We use this to check if the application returns without
# calling start_response:
start_response_started = []
def start_response_wrapper(*args, **kw):
assert len(args) == 2 or len(args) == 3, (
"Invalid number of arguments: %s" % args)
assert not kw, "No keyword arguments allowed"
status = args[0]
headers = args[1]
if len(args) == 3:
exc_info = args[2]
else:
exc_info = None
check_status(status)
check_headers(headers)
check_content_type(status, headers)
check_exc_info(exc_info)
start_response_started.append(None)
return WriteWrapper(start_response(*args))
environ['wsgi.input'] = InputWrapper(environ['wsgi.input'])
environ['wsgi.errors'] = ErrorWrapper(environ['wsgi.errors'])
iterator = application(environ, start_response_wrapper)
assert isinstance(iterator, collections.Iterable), (
"The application must return an iterator, if only an empty list")
check_iterator(iterator)
return IteratorWrapper(iterator, start_response_started)
return lint_app
class InputWrapper(object):
def __init__(self, wsgi_input):
self.input = wsgi_input
def read(self, *args):
assert len(args) <= 1
v = self.input.read(*args)
assert type(v) is binary_type
return v
def readline(self, *args):
v = self.input.readline(*args)
assert type(v) is binary_type
return v
def readlines(self, *args):
assert len(args) <= 1
lines = self.input.readlines(*args)
assert isinstance(lines, list)
for line in lines:
assert type(line) is binary_type
return lines
def __iter__(self):
while 1:
line = self.readline()
if not line:
return
yield line
def close(self):
assert 0, "input.close() must not be called"
def seek(self, *a, **kw):
return self.input.seek(*a, **kw)
class ErrorWrapper(object):
def __init__(self, wsgi_errors):
self.errors = wsgi_errors
def write(self, s):
if not PY3:
assert type(s) is binary_type
self.errors.write(s)
def flush(self):
self.errors.flush()
def writelines(self, seq):
for line in seq:
self.write(line)
def close(self):
assert 0, "errors.close() must not be called"
class WriteWrapper(object):
def __init__(self, wsgi_writer):
self.writer = wsgi_writer
def __call__(self, s):
assert type(s) is binary_type
self.writer(s)
class IteratorWrapper(object):
def __init__(self, wsgi_iterator, check_start_response):
self.original_iterator = wsgi_iterator
self.iterator = iter(wsgi_iterator)
self.closed = False
self.check_start_response = check_start_response
def __iter__(self):
return self
def next(self):
assert not self.closed, (
"Iterator read after closed")
v = next(self.iterator)
if self.check_start_response is not None:
assert self.check_start_response, (
"The application returns and we started iterating over its"
" body, but start_response has not yet been called")
self.check_start_response = None
assert isinstance(v, binary_type), (
"Iterator %r returned a non-%r object: %r"
% (self.iterator, binary_type, v))
return v
__next__ = next
def close(self):
self.closed = True
if hasattr(self.original_iterator, 'close'):
self.original_iterator.close()
def __del__(self):
assert self.closed, (
"Iterator garbage collected without being closed")
def check_environ(environ):
assert type(environ) is dict, (
"Environment is not of the right type: %r (environment: %r)"
% (type(environ), environ))
for key in ['REQUEST_METHOD', 'SERVER_NAME', 'SERVER_PORT',
'wsgi.version', 'wsgi.input', 'wsgi.errors',
'wsgi.multithread', 'wsgi.multiprocess',
'wsgi.run_once']:
assert key in environ, (
"Environment missing required key: %r" % key)
for key in ['HTTP_CONTENT_TYPE', 'HTTP_CONTENT_LENGTH']:
assert key not in environ, (
"Environment should not have the key: %s "
"(use %s instead)" % (key, key[5:]))
if 'QUERY_STRING' not in environ:
warnings.warn(
'QUERY_STRING is not in the WSGI environment; the cgi '
'module will use sys.argv when this variable is missing, '
'so application errors are more likely',
WSGIWarning)
for key in environ:
if '.' in key:
# Extension, we don't care about its type
continue
assert type(environ[key]) in METADATA_TYPE, (
"Environmental variable %s is not a string: %r (value: %r)"
% (key, type(environ[key]), environ[key]))
assert type(environ['wsgi.version']) is tuple, (
"wsgi.version should be a tuple (%r)" % environ['wsgi.version'])
assert environ['wsgi.url_scheme'] in ('http', 'https'), (
"wsgi.url_scheme unknown: %r" % environ['wsgi.url_scheme'])
check_input(environ['wsgi.input'])
check_errors(environ['wsgi.errors'])
# @@: these need filling out:
if environ['REQUEST_METHOD'] not in valid_methods:
warnings.warn(
"Unknown REQUEST_METHOD: %r" % environ['REQUEST_METHOD'],
WSGIWarning)
assert (not environ.get('SCRIPT_NAME')
or environ['SCRIPT_NAME'].startswith(SLASH)), (
"SCRIPT_NAME doesn't start with /: %r" % environ['SCRIPT_NAME'])
assert (not environ.get('PATH_INFO')
or environ['PATH_INFO'].startswith(SLASH)), (
"PATH_INFO doesn't start with /: %r" % environ['PATH_INFO'])
if environ.get('CONTENT_LENGTH'):
assert int(environ['CONTENT_LENGTH']) >= 0, (
"Invalid CONTENT_LENGTH: %r" % environ['CONTENT_LENGTH'])
if not environ.get('SCRIPT_NAME'):
assert 'PATH_INFO' in environ, (
"One of SCRIPT_NAME or PATH_INFO are required (PATH_INFO "
"should at least be '/' if SCRIPT_NAME is empty)")
assert environ.get('SCRIPT_NAME') != SLASH, (
"SCRIPT_NAME cannot be '/'; it should instead be '', and "
"PATH_INFO should be '/'")
def check_input(wsgi_input):
for attr in ['read', 'readline', 'readlines', '__iter__']:
assert hasattr(wsgi_input, attr), (
"wsgi.input (%r) doesn't have the attribute %s"
% (wsgi_input, attr))
def check_errors(wsgi_errors):
for attr in ['flush', 'write', 'writelines']:
assert hasattr(wsgi_errors, attr), (
"wsgi.errors (%r) doesn't have the attribute %s"
% (wsgi_errors, attr))
def check_status(status):
assert type(status) in METADATA_TYPE, (
"Status must be a %s (not %r)" % (METADATA_TYPE, status))
status = to_string(status)
assert len(status) > 5, (
"The status string (%r) should be a three-digit "
"integer followed by a single space and a status explanation"
) % status
assert status[:3].isdigit(), (
"The status string (%r) should start with"
"three digits") % status
status_int = int(status[:3])
assert status_int >= 100, (
"The status code must be greater or equal than "
"100 (got %d)") % status_int
assert status[3] == ' ', (
"The status string (%r) should start with three"
"digits and a space (4th characters is not a space here)") % status
def _assert_latin1_py3(string, message):
if PY3 and type(string) is str:
try:
string.encode('latin1')
except UnicodeEncodeError:
raise AssertionError(message)
def check_headers(headers):
assert type(headers) is list, (
"Headers (%r) must be of type list: %r"
% (headers, type(headers)))
for item in headers:
assert type(item) is tuple, (
"Individual headers (%r) must be of type tuple: %r"
% (item, type(item)))
assert len(item) == 2
name, value = item
_assert_latin1_py3(
name,
"Headers values must be latin1 string or bytes."
"%r is not a valid latin1 string" % (value,)
)
str_name = to_string(name)
assert str_name.lower() != 'status', (
"The Status header cannot be used; it conflicts with CGI "
"script, and HTTP status is not given through headers "
"(value: %r)." % value)
assert '\n' not in str_name and ':' not in str_name, (
"Header names may not contain ':' or '\\n': %r" % name)
assert header_re.search(str_name), "Bad header name: %r" % name
assert not str_name.endswith('-') and not str_name.endswith('_'), (
"Names may not end in '-' or '_': %r" % name)
_assert_latin1_py3(
value,
"Headers values must be latin1 string or bytes."
"%r is not a valid latin1 string" % (value,)
)
str_value = to_string(value)
assert not bad_header_value_re.search(str_value), (
"Bad header value: %r (bad char: %r)"
% (str_value, bad_header_value_re.search(str_value).group(0)))
def check_content_type(status, headers):
code = int(status.split(None, 1)[0])
# @@: need one more person to verify this interpretation of RFC 2616
# http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html
NO_MESSAGE_BODY = (201, 204, 304)
NO_MESSAGE_TYPE = (204, 304)
length = None
for name, value in headers:
str_name = to_string(name)
if str_name.lower() == 'content-length' and value.isdigit():
length = int(value)
break
for name, value in headers:
str_name = to_string(name)
if str_name.lower() == 'content-type':
if code not in NO_MESSAGE_TYPE:
return
elif length == 0:
warnings.warn(("Content-Type header found in a %s response, "
"which not return content.") % code,
WSGIWarning)
return
else:
assert 0, (("Content-Type header found in a %s response, "
"which must not return content.") % code)
if code not in NO_MESSAGE_BODY and length is not None and length > 0:
assert 0, "No Content-Type header found in headers (%s)" % headers
def check_exc_info(exc_info):
assert exc_info is None or type(exc_info) is tuple, (
"exc_info (%r) is not a tuple: %r" % (exc_info, type(exc_info)))
# More exc_info checks?
def check_iterator(iterator):
valid_type = PY3 and bytes or str
# Technically a bytes (str for py2.x) is legal, which is why it's a
# really bad idea, because it may cause the response to be returned
# character-by-character
assert not isinstance(iterator, valid_type), (
"You should not return a bytes as your application iterator, "
"instead return a single-item list containing that string.")
__all__ = ['middleware']