Fix #126 #148
Fix #126 #148
Conversation
|
Sorry, code for process chunks looks too complex to me. I would like to cleanup existing chunk processing first before adding new features. Guys, thoughts? |
|
Everything before 7d036de is cleanup. Currently aiohttp changes I do agree that there should be more cleaning-up, and I do agree the fix 7d036de and 6731131 is not beautiful. Nonetheless, something came up in my life yesterday, and I can no longer work on this issue. Sorry. |
| def update_transfer_encoding(self): | ||
| """Analyze transfer-encoding header.""" | ||
| te = self.headers.get('TRANSFER-ENCODING', '').lower() | ||
|
|
||
| if self.compress: | ||
| if self.chunked is False: | ||
| raise ValueError("Compress is enabled while chunked is disabled.") |
There was a problem hiding this comment.
Actually, you're trying to forbid pretty valid case. Chunks and compression could be used together to represent two semantic principals: 1) stream chunks, join and decode on server 2) stream chunks and decode on server on fly.
There was a problem hiding this comment.
It is a valid case. But aiohttp turned on chunked when compress is True -- even if chunked is explicitly set to False. This is not a correct behavious, but to keep the old quirks.
|
I agree with @asvetlov that this PR turns request generation on more higher complexity level. |
|
@kxepal, This PR should've been closed. My changes kept all the quirks that aiohttp had. I wanted to keep the changes to a minimal, meaning no structure modifications. And I added some other ugly shit to make the tests pass. In retrospect, that was not a good idea, maybe somebody will come up with a proper clean way. |
This pull request fixes mainly #126, and some other things along the way.
Tests added.