eedfc64a8ceb38c3743165405ab40a7ee1daa3a8
5 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
F. Eugene Aumson
|
f51c80adb2 |
Change all instances of networkId to chainId (#2313)
* abi-gen/test: recompile contract fixtures for 3.0 It seems this hadn't been done since the merge with the 3.0 branch. * Sync `monorepo$ yarn test` exclusions to CI config * sra-spec: correct typo * contract-wrappers: TODO after coord.-server update * utils: fix typo in comment * Refactor networkId to chainId everywhere * Update CHANGELOGs |
||
|
F. Eugene Aumson
|
e61f23d001 |
Migrate Python libraries to v3 (#2284)
* .gitignore migrations/0x_ganache_snapshot
* .gitignore new-ish Python contract wrappers
These should have been added back when we started generating these
wrappers.
* rm superfluous contract artifact in Python package
All of the contract artifacts were removed from the Python package
recently, because now they're copied from the monorepo/packages area as
an automated build step. Somehow this one artifact slipped through the
cracks.
* Eliminate circular dependency
This was preventing the Exchange wrapper from ever importing its
validator!
* Improve output of monorepo-level parallel script
- Capture stderr (and have it included in stdout) so that it doesn't
leak onto the console for commands that didn't actually fail.
- Include all error output in the Exception object (eliminate print
statement).
* Silence new versions of linters
Newer versions care about this stuff. Old versions didn't, and we don't
either.
* Support Rich Reverts via Web3.py middleware
* Fix bug in generated wrappers' bytes handling
`bytes.fromhex(bytes.decode('utf-8')` is just plain wrong. It would
work for some cases, but is not working when trying to fill orders with
the latest Exchange contract.
* Migrate to Exchange v3
* Fix typo in DevUtils documentation
* Include new contracts in docs
* Re-enable Python checks in CI
* Accept strings for bytes
* Fix CircleCI build artifacts for gen'd python
I swear the previous way was working before, but it wasn't working now,
so this fixes it.
* Accept a provider OR a Web3 object
In various places. This allows the caller to install middleware (which
in web3.py is installed on a Web3 object, not on a provider) before
executing any RPC calls, which is important for the case where one wants
to produce signatures locally before submitting to a remote node.
* wrapper base: don't assume there are accounts
* Eliminate some inline linter directives
* make CHANGELOGs be REVERSE chronological
* Update CHANGELOG entries and bump version numbers
* @0x/contract-addresses: Put addr's in JSON, not TS
This allows easier consumption by other languages. (Specifically, it
eliminates the overhead of keeping the Python addresses package in sync
with the TypeScript one.)
* sra_client.py: incl. docker in `./setup.py clean`
* sra_client.py: Migrate to protocol v3
Removed script that existed only to exclude runs of sra_client builds
(parallel_without_sra_client). Now `parallel` is used by CI,
re-including sra_client in CI checks.
* abi-gen/templates/Py: clarify if/else logic
In response to
https://github.com/0xProject/0x-monorepo/pull/2284#discussion_r342200906
* sra_client.py: Update CHANGELOG and bump version
* contract_addresses/setup.py: rm unnecessary rm
* json_schemas.py: corrections to dev dependencies
* In tests against deployment, also run doctests
* contract_wrappers example: rm xtra Order attribute
Thanks to @steveklebanoff for catching this.
https://github.com/0xProject/0x-monorepo/pull/2284#pullrequestreview-312065368
|
||
|
F. Eugene Aumson
|
d9378e9a8f |
First auto-gen'd Python contract wrapper: ERC20 (#1878)
* abi-gen-wrappers: fix half-baked folder rename
* .circleci: give cache more descriptive name
* abi-gen: rename type field. tsName->langLocalName
* contract-artifacts: add devdoc to ERC20Token.json
Artifact changes crafted manually: re-generated artifacts from within
@0x/contracts-erc20, and then copied them into @0x/contract-artifacts,
but only committed the changes that added devdoc.
Modified @0x/contracts-erc20/compiler.json to include devdoc in the
hopes that next time contract-artifacts are re-generated en masse, the
devdoc output will just come along for the ride.
Compiling ERC20 TypeScript wrappers after adding devdoc to the artifacts
revealed some inconsistencies in the types. ethereum-types'
DevdocOutput REQUIRED many fields which are not always present,
depending the devdoc contents itself. Relaxing the requirement had some
ramifications.
* abi-gen: Python!
Lots more to come, but so far generating typed methods with some
parameter validation and with reasonable docstrings; enough to make
ERC20 work.
* existing erc20 python wrapper: re-order methods
In order to minimize the diff in the upcoming commit of the
auto-generated version.
* existing erc20 python wrapper: rename method
Rename method _erc20 to _get_contract_instance.
* existing erc20 python wrapper: rename vars
Rename method parameters to match names used in contract methods.
* existing erc20 python wrapper: correction
Contract method `allowance` previously was returning a transaction hash,
but it's a const method, so changed it to return the int that the
underlying method returns.
* contract_wrappers.py: pull in generated code
Custom setup.py command to pull in code previously generated into
../../packages/abi-gen-wrappers/src/generated-wrappers/py.
Changes to existing wrapper code reflect differences between the old,
manually-written wrapper and the new, auto-generated wrapper.
Reconfigured Circle CI to give Python build access to the output of the
TypeScript build, in order to import the generated wrappers from there.
* abi-gen: sanitize docstrings for pycodestyle lint
* abi-gen-templates.py: expose ABI from template
Use ABI given by template, not gotten by name from
zero_ex.contract_artifacts.
Also, expose ABI as a static method.
* py wrappers: contract addr to ctor, not methods
In all Python wrappers (old, manual ones; and new, generated ones), have
client pass the contract address to the wrapper's constructor, rather
than to the individual method calls.
* py wrappers: remove unused param `account_address`
* py wrappers: document use of `private_key`
* Rename file erc20_wrapper.py to erc20_token.py
To match the name of the underlying contract.
* Update CHANGELOG.json's
* git rm erc20_token.py
No need to keep it checked in. Doing so would require manual overhead
of keeping the generated copy in sync with the generation code, which
may get overlooked and cause confusion for others. Authoritative source
will be the published package on PyPI.
* abi-gen-templates: tweak CHANGELOG wording
Co-Authored-By: Fabio B <kandinsky454@protonmail.ch>
* Include transaction parameters in const calls too
* Doc contract_address param to gen'd wrapper ctor
* make myself a CODEOWNER of abi-gen*
* rename ids: langLocalName -> languageSpecificName
* Move Python generation to its own packages/ folder
* Stop duplicating contract artifacts in Python pkg!
Thanks to the way we're now using the `./setup.py pre_install`
convention, there's no longer a need to check the artifacts in to the
Python package.
* move py templates BACK to abi-gen-templates
I got a little overzealous in the previous commit that moved ALL the
python stuff into the new packages/python-contract-wrappers folder.
* Update known-good test output: prettify
* add getABIEncodedTransationData to gen'd code
It was added into the template in the following commit, but the
corresponding checked-in generated code was not updated accordingly.
|
||
|
F. Eugene Aumson
|
0564ac1530 |
Python doc polish (#1757)
* Exercise doctests as a test not as a linter * Add a contract artifact doctest, and exercise it * Clean up linter issues * Change asset data decoding output type Previously, it was a TypedDict, but that was causing problems. Sphinx seems to be broken, such that none of the fields of the class were being rendered into the doc. Thinking on it further, I decided that a NamedTuple makes more sense here anyways, since tuples are immutable and this output value isn't something someone should ever build or modify. And, NamedTuple is getting its fields properly rendered by Sphinx. * Add type annotations to JSON schemas docs * Add doc publish metadata file for middlewares pkg * Improve documentation Note that none of the changes to .py files impact functionality in any way, because the changes are restricted to "docstrings", which to the Python interpreter are simply no-op statements. However, one caveat to that is that much of these docstring changes DO affect the functionality of automated test runs, because all of the code examples (blocks beginning with `>>> `) are "doctests", which are exercised via the test framework. The index.rst files are the top-level templates for generating the documentation, and the "automodule"/"autoclass"/etc statements pull in the docstrings from the source code. * correct package name in doc URL * Move sra_client module into zero_ex namespace * Add functions to encode asset data to bytes * Fix: SRA client was deserializing orders weirdly The generated code was transforming the order structure, from the camel case field name format in the spec, into the snake case field name format expected by Python convention. With this problem in place, the only way to take an order from a relayer and send it to a contract (for fill, cancel, etc) was to manually transform the field names, one by one, into a new structure. * Fix problem with Web3/JSON order conversion utils * doctest: maker, trade ZRX for WETH, not vice versa * Remove redundant test * Construct order in native Python, not JSON Then convert it to JSON before sending it to the relayer. * doctest: simplify asset units * Add doctests for filling and cancelling * Minor doctetst copy edits; whitespace * Rename function, and add optional parameter * Tweak docstrings on JSON conversion functions. * Demo asset data decoding to view asset pairs * Demo selecting an order from the order book And have taker take it. * Rename variable * Abstract ganache from examples Doing that exposed excessive use of the verbose NETWORK_TO_ADDRESSES[NetworkId.Ganache] construct, so simplified that, which ripped into eliminating other temporary variables that had been used to hold specific contract addresses. Also cleaned up some misplaced import statements. * Add missing SRA client doc publication metadata * Ran prettier on new SRA client doc pub metadata * Remove local env customizations in doc metadata * Eliminate temporary variable * Rename variable * Show `pip install` in every package's doc * Doc NetorkID & pagination params as int, not float * Clean up unmatched parenthesis in docs |
||
Michael Huang
|
a256494ec8 | Python contract wrappers (#1721) |