RobertoPrevato
commented
1 year ago Hi @tyzhnenko Thank You for your contribution, I appreciate this very much! I thought many times to adding something similar but always postponed that.
I have two recommendations:
- because gzipping is a CPU bound operation, it is best to use an executor (run_in_executor) when compressing, as the zipping process might block the event loop (depending on the size of the content being zipped). Maybe the GzipMiddleware constructor might accept an optional ThreadPoolExecutor. So by default it would use the default one, and optionally a dedicated thread pool.
- the second recommendation would be to support specifying which content types should be zipped, as some contents should not be zipped. NGINX for example compresses by default only text/html. I would check only for substrings because there are many variants of MIME types in some cases.
Something like:
handled_content_types = {
b"json",
b"xml",
b"yaml",
b"html",
b"text/plain",
b"application/javascript",
b"text/css",
b"text/csv",
}
def handle_content_type(self, content_type: bytes) -> bool:
lower_content_type = content_type.lower()
return any(fragment in lower_content_type for fragment in self.handled_content_types)
# handle request only if the content type should be handledIf you don´t have more time to spare, we can merge sooner and I can add these small changes before releasing to PyPi. 😄
tyzhnenko
codecov-commenter
Add GzipMiddleware class that compresses response body if request Accept-Encoding header contains 'gzip'
Add GzipMiddleware for app_1, app_2, app_3, app_4 integration tests servers
Add unit tests for GzipMiddleware