1from django.utils.cache import patch_vary_headers
2from django.utils.deprecation import MiddlewareMixin
3from django.utils.regex_helper import _lazy_re_compile
4from django.utils.text import compress_sequence, compress_string
5
6re_accepts_gzip = _lazy_re_compile(r"\bgzip\b")
7
8
9class GZipMiddleware(MiddlewareMixin):
10 """
11 Compress content if the browser allows gzip compression.
12 Set the Vary header accordingly, so that caches will base their storage
13 on the Accept-Encoding header.
14 """
15
16 max_random_bytes = 100
17
18 def process_response(self, request, response):
19 # It's not worth attempting to compress really short responses.
20 if not response.streaming and len(response.content) < 200:
21 return response
22
23 # Avoid gzipping if we've already got a content-encoding.
24 if response.has_header("Content-Encoding"):
25 return response
26
27 patch_vary_headers(response, ("Accept-Encoding",))
28
29 ae = request.META.get("HTTP_ACCEPT_ENCODING", "")
30 if not re_accepts_gzip.search(ae):
31 return response
32
33 if response.streaming:
34 if response.is_async:
35 # pull to lexical scope to capture fixed reference in case
36 # streaming_content is set again later.
37 orignal_iterator = response.streaming_content
38
39 async def gzip_wrapper():
40 async for chunk in orignal_iterator:
41 yield compress_string(
42 chunk,
43 max_random_bytes=self.max_random_bytes,
44 )
45
46 response.streaming_content = gzip_wrapper()
47 else:
48 response.streaming_content = compress_sequence(
49 response.streaming_content,
50 max_random_bytes=self.max_random_bytes,
51 )
52 # Delete the `Content-Length` header for streaming content, because
53 # we won't know the compressed size until we stream it.
54 del response.headers["Content-Length"]
55 else:
56 # Return the compressed content only if it's actually shorter.
57 compressed_content = compress_string(
58 response.content,
59 max_random_bytes=self.max_random_bytes,
60 )
61 if len(compressed_content) >= len(response.content):
62 return response
63 response.content = compressed_content
64 response.headers["Content-Length"] = str(len(response.content))
65
66 # If there is a strong ETag, make it weak to fulfill the requirements
67 # of RFC 9110 Section 8.8.1 while also allowing conditional request
68 # matches on ETags.
69 etag = response.get("ETag")
70 if etag and etag.startswith('"'):
71 response.headers["ETag"] = "W/" + etag
72 response.headers["Content-Encoding"] = "gzip"
73
74 return response