Remove utf-8 string decoding from normalizers #974
Conversation
dylanjw
force-pushed
the
eth_abi_string_decode
branch
from
f4dafc1
to
c4bb22c
Compare
There was a problem hiding this comment.
I'm a litlte uncomfortable with this change. I think we should instead support the 1x and 2x line concurrently before dropping 1x support entirely. I'm up for being convinced otherwise and I understand that this adds some testing overhead.
My concern is that eth-abi>=2 has a breaking change and anyone who relies on both the web3 library and any low level functionality in the eth-abi library will be forced to abruptly upgrade their code without us having issued any deprecation warnings.
dylanjw
force-pushed
the
eth_abi_string_decode
branch
3 times, most recently
from
59878f8
to
a2533a7
Compare
dylanjw
force-pushed
the
eth_abi_string_decode
branch
from
a2533a7
to
2041b5d
Compare
|
@pipermerriam Updated to add support for eth-abi >=2 without removing support for <2. |
web3/utils/normalizers.py
Outdated
| @@ -140,11 +144,14 @@ def abi_ens_resolver(w3, abi_type, val): | |||
|
|
|||
|
|
|||
| BASE_RETURN_NORMALIZERS = [ | |||
| addresses_checksummed, | |||
| decode_abi_strings, | |||
| addresses_checksummed | |||
There was a problem hiding this comment.
style nitpick leave the trailing comma.
| @@ -184,6 +188,9 @@ def test_call_get_string_value(string_contract, call): | |||
| assert result == "Caqalai" | |||
|
|
|||
|
|
|||
| @pytest.mark.skipif( | |||
| LooseVersion(eth_abi.__version__) >= LooseVersion("2"), | |||
| reason="eth-abi >=2 does utf-8 string decoding") | |||
| def test_call_read_string_variable(string_contract, call): | |||
| result = call(contract=string_contract, | |||
| contract_function='constValue') | |||
There was a problem hiding this comment.
Maybe add a second test that shows what happens on an ABI decoding error. We should probably do something like capture the eth-abi error and re-raise a generic ValueError.
There was a problem hiding this comment.
It raises a 'BadFunctionCallOutput' error
|
@pipermerriam @carver Ready for another review |
setup.py
Outdated
| @@ -19,7 +19,7 @@ | |||
| install_requires=[ | |||
| "toolz>=0.9.0,<1.0.0;implementation_name=='pypy'", | |||
| "cytoolz>=0.9.0,<1.0.0;implementation_name=='cpython'", | |||
| "eth-abi>=1.1.1,<2", | |||
| "eth-abi==2.0.0a1", | |||
There was a problem hiding this comment.
We shouldn't require people to upgrade. We especially shouldn't require people to install an alpha dependency in a stable release. eth-abi>=1.1.1,<3 seems like a reasonable choice
dylanjw
force-pushed
the
eth_abi_string_decode
branch
from
9fb151a
to
2cc941e
Compare
|
@pipermerriam @carver Is this good to merge? |
What was wrong?
The eth-abi v2 alpha release adds utf-8 string decoding to the string type decoder. String decoding should therefore be removed from web3 return data normalizers.
Cute Animal Picture