Warn on improperly formatted parameter definitions

[timhoffm]

Parameter name and types are split by " : "

val : int

It's a common mistake to leave out the leading space, i.e.

val: int

This results in the whole string being interpreted as "name" and thus incorrect rendering such as:

The relevant code is:

numpydoc/numpydoc/docscrape.py

Lines 229 to 239 in 4dd6fd4

	if " : " in header:
	arg_name, arg_type = header.split(" : ", maxsplit=1)
	else:
	# NOTE: param line with single element should never have a
	# a " :" before the description line, so this should probably
	# warn.
	header = header.removesuffix(" :")
	if single_element_is_type:
	arg_name, arg_type = "", header
	else:
	arg_name, arg_type = header, ""

Would it be desirable to warn in that case? E.g. the most simple check in the else branch could be:

if ": " in header:
    # warn

[larsoner]

I have also hit this bug so agree it's worth finding some way to combat it. Not sure if the proposed simple solution would have some unwanted side effects but it seems worth trying!

[timhoffm]

We should be pretty safe in the else-else branch (l.239), i.e. not single_element_is_type. In this case, we would usually expect "param : type", but haven't found it.

I see little danger in the simple ": " in check. It can only create a false-positive warning if the author has intentionally left out the type and used : in the parameter name. One could try to be more selective and regex match "[param_name]: ", but I don't think that's neccessary.

[larsoner]

@timhoffm PR welcome if you're up for it!