Url type port limit not checked

[flapili]

pydantic version: 1.5.1
pydantic compiled: False
install path: /opt/fastapi/backend/venv/lib/python3.8/site-packages/pydantic
python version: 3.8.3 (default, Jun 21 2020, 00:03:19)  [GCC 8.3.0]
platform: Linux-4.19.0-9-amd64-x86_64-with-glibc2.28
optional deps. installed: ['email-validator']

Hi,

The port limit is not checked by URLs types:

>>> from pydantic import BaseModel, stricturl
>>> class Test(BaseModel):
	url: stricturl(allowed_schemes=["tcp"])

	
>>> Test(url="tcp://127.0.0.1:100000000000000000000000000000")
Test(url=UrlValue('tcp://127.0.0.1:100000000000000000000000000000', scheme='tcp', host='127.0.0.1', host_type='ipv4', port='100000000000000000000000000000'))

Excepted result: raise an error,

my current workaround

>>> from pydantic import BaseModel, stricturl, validator
>>> from pydantic.networks import url_regex
>>> class Test(BaseModel):
	url: stricturl(allowed_schemes=["tcp"])
	@validator("url")
	def validate_port(cls, v):
		port = url_regex().match(v).group("port")
		if port is None:
		    return v

		if int(port) > 2**16:
			raise ValueError("port overflow")
		return v

	
>>> Test(url="tcp://127.0.0.1:900000")
Traceback (most recent call last):
  File "<pyshell#84>", line 1, in <module>
    Test(url="tcp://127.0.0.1:900000")
  File "pydantic\main.py", line 338, in pydantic.main.BaseModel.__init__
pydantic.error_wrappers.ValidationError: 1 validation error for Test
url
  port overflow (type=value_error)

[samuelcolvin]

Is there a nice RFC explicitly stating a limit for port? If so I guess we could add a check, but this is not a bug, just a limit of the current validation.

You can implement a much cleaner validator:

class Test(BaseModel):
    url: stricturl(allowed_schemes=['tcp'])

    @validator('url')
    def validate_port(cls, v):
        if int(v.port) > 2**16:
            raise ValueError('port overflow')
        return v

[flapili]

Hi,
rfc 793 explain ports are 16 unsigned bits

Your validator will fail if port is missing, with tcp protocol that doesn't really make sense but with http(s) protocol is does:
example

http://example.com the validator will fail because v.port is None, int will raise a TypeError, but thanks for the tips I totally forgot port attribute 🤦

class Test(BaseModel):
    url: stricturl(allowed_schemes=['tcp'])

    @validator('url')
    def validate_port(cls, v):
        if v.port is None:
            return v
        elif int(v.port) > 2**16:
            raise ValueError('port overflow')
        return v

(the tcp protocol is a bad showcase about default port)

[samuelcolvin]

Good point, if v.port is not None and int(v.port) > 2**16 would be better. (is port=0 allowed?)

rfc 793 explain ports are 16 unsigned bits

PR welcome to add this check to all URL validation, I know technically it's a breaking chance, but I think a reasonable one, given that a higher port is invalid anyway.

[flapili]

the port 0 is tagged as reserved and should not be usable as such since most routers will reject it.
In a other hand in a lot of APIs port 0 mean "use the first avaiable socket" so I would say yes.

I'll work on a PR within a few days (this weekend I hope if I have time)

[samuelcolvin]

makes sense to allow 0, also makes the check easier.

I'll work on a PR within a few days (this weekend I hope if I have time)

Thanks so much

[flapili]

I'm currently working on a PR,
looking at the sources of networks I have few questions about some design choices

Should I open news issues for each or asking here is fine ?

[Skeen]

Would it make sense to separate Port into a seperate type, and to utilize it when checking urls? - I'm asking as I'm currently in need of port validation and I thought it was provided out of the box, or atleast as an internal type utilized by url parsing / validation.

[flapili]

Would it make sense to separate Port into a seperate type, and to utilize it when checking urls? - I'm asking as I'm currently in need of port validation and I thought it was provided out of the box, or atleast as an internal type utilized by url parsing / validation.

You can use conint for that @Skeen