QuerryParams adds to URL params with "None" value

[stevstrong]

Hi there,

I used the build_request method of AsyncClient() with the following params:
{'start_time': 1651222800, 'end_time': None}

The resulting URL looks like this:
<my_url>?start_time=1651222800&end_time=

I would expect that the second key having value "None" from the params will not be added at all to the URL.

I tested the prepare() method of the standard requests lib and that returns the following URL (conforming with my expectations not adding the second key as it has "None" value):
<my_url>?start_time=1651222800

Digging in the source files I reached these lines in __init__ method of QueryParams class:

httpx/httpx/_urls.py

Lines 563 to 593 in 3350d7e

	if value is None or isinstance(value, (str, bytes)):
	value = value.decode("ascii") if isinstance(value, bytes) else value
	self._dict = parse_qs(value)
	elif isinstance(value, QueryParams):
	self._dict = {k: list(v) for k, v in value._dict.items()}
	else:
	dict_value: typing.Dict[typing.Any, typing.List[typing.Any]] = {}
	if isinstance(value, (list, tuple)):
	# Convert list inputs like:
	# [("a", "123"), ("a", "456"), ("b", "789")]
	# To a dict representation, like:
	# {"a": ["123", "456"], "b": ["789"]}
	for item in value:
	dict_value.setdefault(item[0], []).append(item[1])
	else:
	# Convert dict inputs like:
	# {"a": "123", "b": ["456", "789"]}
	# To dict inputs where values are always lists, like:
	# {"a": ["123"], "b": ["456", "789"]}
	dict_value = {
	k: list(v) if isinstance(v, (list, tuple)) else [v]
	for k, v in value.items()
	}
	# Ensure that keys and values are neatly coerced to strings.
	# We coerce values `True` and `False` to JSON-like "true" and "false"
	# representations, and coerce `None` values to the empty string.
	self._dict = {
	str(k): [primitive_value_to_str(item) for item in v]
	for k, v in dict_value.items()
	}

Is it possible that a "None" value is not checked and excluded at all?
If yes, is that wanted?

[lovelydinosaur]

I would expect that the second key having value "None" from the params will not be added at all to the URL.

We have different behaviour from requests here, yes.

Personally the "omit a parameter if the value is None" seems like non-intuitive behaviour to me, although I realise that folks may have different expectations here.

There's a brief note here: https://www.python-httpx.org/compatibility/#query-parameters

[lovelydinosaur]

I do think we could improve the documentation phrasing here to make the behaviour more clear...

The requests package omits params where the values are None. For example requests.get("https://www.example.com/", params={"search": None}) will send a request to the URL https://www.example.com/.

In contrast to this, the httpx package includes an empty value for params where the values are None. For example httpx.get("https://www.example.com/", params={"search": None}) will send a request to the URL https://www.example.com/?search=.

Overall we consider this to be more consistent behaviour.

[stevstrong]

I see, thanks. In this case It would definitely make sense to update the docs.

I can (and will) change my code to be inline with the new description, I just thought that httpx looks to keep compatibility with requests lib as far as possible.
If not, it is also fine for me.