Version 0.15.0

[lovelydinosaur]

Pushing this up so that we can figure out what else we want to block a 0.15 release on.

At the moment my thoughts would be...

• Change the Transport API to .request, .arequest, since it'll otherwise end up blocking other work.
• Possibly also switching the transport API to request(method, url, headers, stream, ext) -> status_code, headers, stream, ext. See Finalising the Transport API for 1.0. #1274 (comment) - Might need discussion, but it's an easy switch to make, and better to make it sooner rather than later.
• Possibly get in the CLI. Because it's lovely, and might be a nice driver of adoption. Add an HTTPX cli. 😇 #1212
• Go through an escalate any deprecations that we've got in the codebase into outright removals.

I think other stuff can wait for nice smaller increments in the 0.15.x series. The only remaining API changes expected at that point are the possible transport API -> transport API as a context manager switch, and some very minor review over all the model classes. (Note that URL is now complete in this regard.)

What do we think, and am I missing anything critical here?

0.15.0

Added

• Added support for curio. (Pull Implement curio backend httpcore#168)
• Added support for event hooks. (Pull Event hooks #1246)
• Added support for authentication flows which require either sync or async I/O. (Pull Add support for sync-specific or async-specific auth flows #1217)
• Added support for monitoring download progress with response.num_bytes_downloaded. (Pull Add progress to streaming download #1268)
• Added Request(content=...) for byte content, instead of overloading Request(data=...) (Pull Seperate content=... and data=... parameters #1266)
• Added support for all URL components as parameter names when using url.copy_with(...). (Pull Finesse URL properties #1285)
• Neater split between automatically populated headers on Request instances, vs default client.headers. (Pull Header refinements #1248)
• Unclosed AsyncClient instances will now raise warnings if garbage collected. (Pull Issue warning on unclosed AsyncClient. #1197)
• Support Response(content=..., text=..., html=..., json=...) for creating usable response instances in code. (Pull Support Response(content=<bytes iterator>) #1265, Support Response(text=...), Response(html=...), Response(json=...) #1297)
• Support instantiating requests from the low-level transport API. (Pull Requests from transport API #1293)
• Raise errors on invalid URL types. (Pull Type error on invalid url type #1259)

Changed

• Cleaned up expected behaviour for URL escaping. url.path is now URL escaped. (Pull Finesse URL properties #1285)
• Cleaned up expected behaviour for bytes vs str in URL components. url.userinfo and url.query are not URL escaped, and so return bytes. (Pull Finesse URL properties #1285)
• Drop url.authority property in favour of url.netloc, since "authority" was semantically incorrect. (Pull Finesse URL properties #1285)
• Drop url.full_path property in favour of url.raw_path, for better consistency with other parts of the API. (Pull Finesse URL properties #1285)
• No longer use the chardet library for auto-detecting charsets, instead defaulting to a simpler approach when no charset is specified. (Drop chardet #1269)

Fixed

• Swapped ordering of redirects and authentication flow. (Pull Swap auth/redirects ordering #1267)
• .netrc lookups should use host, not host+port. (Pull NetRC lookups should use host, not host+port #1298)

Removed

• The URLLib3Transport class no longer exists. We've published it instead as an example of a custom transport class. (Pull Drop urllib3 in favor of public gist #1182)
• Drop request.timer attribute, which was being used internally to set response.elapsed. (Pull Drop request.timer attribute. #1249)
• Drop response.decoder attribute, which was being used internally. (Pull Minor decoder refactoring. #1276)
• Request.prepare() is now a private method. (Pull Refactoring of models api #1284)

[florimondmanca]

Sounds good to me, great one!

[lovelydinosaur]

Right, let's keep the CLI separate. No good reason for us to bog ourselves down in a whole extra layer of API considerations (eg. which flags do we support on the command line) when we're so very close to a 1.0 release now.

I'm going to push on again, and get this released.

Where does this leave us after this?...

Our 1.0 milestone is here.

• There's one significantly complex bug that needs resolving in httpcore around connection pooling. PoolTimeout when num tasks in asyncio.gather() exceeds client max_connections #1171 - It'll be a decent chunk of work, and it's right in the guts of things, so I want to deal with this in a dedicated httpcore minor point release, that should only include that bugfix and nothing else. Eg. 0.11.1, or wherever we're at.
• Review any other bug-labelled tickets.
• Mount API, httpx.HTTPTransport, httpx.MockTransport - Mount API #977, Add httpx.HTTPTransport() and httpx.AsyncHTTPTransport() #1302, Add httpx.MockTransport() #1303 are all in the same area. I figure we'll have a minor httpx release that's dedicated just to adding those, and sprucing up our docs on mounting transports etc.
• chunk_size Allow setting chunk_size for Response.iter_bytes() etc... #394 - We've got an outstanding PR for this, but I'd like to pull it in, in it's own minor point release. Since all response.text/response.content runs through that code I'd like it to be really nicely isolated from any other changes, just in case we needed to very quickly identify any unexpected problems.
• Tightening public API on models. Tightening public API on models #1275 I think this might just be Cookies that's left to review here now.
• Preserving HTTP Header casing. Blocked on Preserve header casing. Take two. 🎬 python-hyper/h11#104
• Retries Retries #1141
• Transport API as a context manager. Our last API change still to be made. Add support for streaming responses to ASGITransport #998 and Finalising the Transport API for 1.0. #1274

We'll then also want to go through the API docs, and make sure that all the public API is being properly exposed in the docs.