Web3.isAddress doesn't work for non prefixed checksumed values

[Uxio0]

• Version: 4.0.0b11
• Python: 3.6
• OS: linux

What was wrong?

As stated in the docs http://web3py.readthedocs.io/en/latest/overview.html#Web3.isAddress the function Web3.isAddress(value) should allow both 0x prefixed and non prefixed values.

If the address is not checksumed, it's ok not to have the 0x:

>>> Web3.isAddress('d3cda913deb6f67967b99d67acdfa1712c293601')
>>> True

But if it's checksumed

>>> Web3.isAddress('d3CdA913deB6f67967B99D67aCDFa1712C293601')
>>> False

No problem if we add the 0x:

>>> Web3.isAddress('0xd3CdA913deB6f67967B99D67aCDFa1712C293601')
>>> True

How can it be fixed?

Changing the documentation to state that checksumed addresses must have 0x or changing the function to accept checksumed addresses with 0x. I would just remove 0x at the beginning of the function (if found) and work with the address as that.

[pipermerriam]

This is due to a discrepancy between is_checksum_formatted_address and is_checksum_address. The former returns True but the later False

In [79]: is_checksum_formatted_address('d3CdA913deB6f67967B99D67aCDFa1712C293601')
Out[79]: True

In [80]: is_checksum_address('d3CdA913deB6f67967B99D67aCDFa1712C293601')
Out[80]: False

We could:

• Special case is_address to allow non-prefixed checksummed addresses.
• Be more forgiving with is_checksum_address to allow non-prefixed values.
• Remove the checksum checks since is_hex_address('d3CdA913deB6f67967B99D67aCDFa1712C293601') -> True

I'm torn. I like having a very strict representation of checksummed addresses (as well as an unambiguous requirement that they have 0x prefixes to remove ambiguity of whether it's a hex value). However, I can see a decent argument for being more forgiving with is_checksum_address and allowing non-prefixed values.

[carver]

The reason it works this way is because of how EIP-55 defines the checksum addresses. According to spec, it's not a valid checksum address if it does not have the "0x" prefix. Note that the test() method defined in the spec is only valid on prefixed addresses.

So this seems like the right solution:

Changing the documentation to state that checksumed addresses must have 0x

[pipermerriam]

Ok, but the real question here is *what is the is_address function intended to be used for?

• is this a valid address
• is this an address

If it's either of the valid variations then I think the function is poorly named, so I'd posture that the right fix is to update the docs as you specified, but to remove the checksum checking from the is_address function and have it simply return something like return is_hex_address(v) or is_binary_address(v)

[carver]

the real question here is *what is the is_address function intended to be used for?

Good point. It used to be that isAddress was a test for "is this value valid as an input, where addresses are required in web3.py?" I don't think we've re-evaluated the use case of this method, given the rule "only checksum addresses are accepted" in web3 v4.

I suppose it should be: "is this address valid as an input to: toChecksumAddress". Something seems a bit wrong about accepting invalid checksums as input to toChecksumAddress, but that's not a check we're doing now AFAIK, and I don't have a well-formed opinion that we should add that check. So I guess that leads us to the same place @pipermerriam described:

def is_address(v):
    return is_hex_address(v) or is_binary_address(v)

[fselmo]

Once this is merged in eth-utils, major version bump for eth-utils in web3.py would be needed since it is a breaking change. Basically, for v7, we would need eth-utils>={relevant major version with this change} in setup.py.