Skip to content

Latest commit

 

History

History
146 lines (94 loc) · 5.12 KB

CHANGELOG.rst

File metadata and controls

146 lines (94 loc) · 5.12 KB
Raw

0.4.0 — 2023-05-19

Removed

  • Nomics market data is no longer available, so we have removed all nomics related functionality.
  • Removed pool lp token symbol lookup when creating pools from on-chain data.
  • Standalone functions for AMM logic are removed. This eases the testing burden without impacting performance (due to other changes such as __slot__ usage).
  • The matplotlib results plotter was replaced with an altair plotter.

Added

  • Curve pool types now use __slots__ for more performant attribute access.
  • Revamped pool hierarchy so that the implementations of SimPool derive from Curve pools and SimStableSwapBase, which better organizes common logic across stableswap sim pool types and decouples sim logic from the core AMM pools.
  • Pools now have a revert-to-snapshot functionality.
  • Add CurveCryptoPool, a python implementation of the vyper factory cryptoswap pool.
  • curvesim.pool.get can now be used to fetch and instantiate cryptopools.
  • curvesim now uses python logging with log levels. This allows for debug logging and saving logs to files.
  • Altair results plotter added to improve readability and support flexible plotting (e.g., across pool types)
  • Metrics submodule added to facilitate simulation metric development.

Changed

  • CurveMetaPool uses rate_multiplier as in the vyper logic instead of p
  • By default, pool instantiation will now create balances in native token units. Previously it had normalized to 18 decimals. This option still exists but must now be chosen explicitly.
  • Some internal objects used by curvesim were refactored for better maintainability, namely PoolData is now split into PoolDataCache and PoolMetaData.

Fixed

  • Use rates property consistently across stableswap types
  • Add error handling for convex subgraph
  • Timezone issue in subgraph queries.

0.3.0 — 2022-11-08

Added

  • Multi-chain pool data support via the Convex Community subgraph.
  • New network subpackage uses asyncio for concurrency.
  • Pool_data and price_data submodules added.
  • Single cpu-mode (ncpu=1) will run in a single process without the multiprocessing library. This makes it easier to profile using tools like cProfile.
  • python3 -m curvesim will do a demo sim run so users can check everything is setup properly.
  • Support use of environment variable (and loading from .env file) for NOMICS_API_KEY, ALCHEMY_API_KEY, and ETHERSCAN_API_KEY. The latter two are optional in that the package provides default keys, but it is recommended to use your own if you need the functionality (currently only for pulling coin data for lending pools).
  • Pipeline and iterators submodules added to support custom simulation pipelines. This will allow more complex arbitrage scenarios and let users create bespoke simulations.
  • Pool simulation interfaces added to decouple pool implementations from the simulation framework. The interfaces enable additional runtime optimizations.
  • Standard volume-limited arbitrage simulation re-implemented using the new pipeline framework.
  • Pools initiated from external data now store their metadata in pool.metadata for introspection and debugging.
  • Create versioning structure to bump versions which will reflect in the changelog and future package releases.
  • Added end-to-end tests for simulation runs that run in continuous integration. Unit tests added for pool calculations. Increasing test coverage with component-level tests will be a key part of getting to v1.

Removed

  • sim() and psim() replaced by pipeline framework.
  • PoolDF CSVs no longer used for pool data lookup.

Changed

  • Transitioned repo organization to reflect standard packaging style.
  • Frequently used calculations such as D, y, and dydxfee use the GNU Multiple Precision Arithmetic Library (GMP) to speed up big integer arithmetic.
  • The spot pricing function, dydxfee, uses a derivation from calculus instead of bumping a pool balance and recalculating, with the exception of a certain case for metapools.
  • The monolithic Pool class was split into a generic base class, with derived classes CurvePool, CurveMetaPool, and CurveRaiPool.
  • Bonding curve and order-book Pool methods changed to standalone functions.
  • "Price depth" metrics now report liquidity density (i.e., % change in holdings per % change in price).
  • Curvesim.autosim() now only accepts ints or iterables of ints for pool parameters (e.g., A, D, fee).
  • External pool data now referenced using pool address or LP token symbol and chain. This logic is used in autosim and pool_data.

Fixed

  • Subgraph volume query was updated due to a recent update.
  • Fixed bug in vol_mode=2 for non-meta-pools
  • Various updates to pool calculations to align the results with their on-chain equivalents.
  • Codebase is much more PEP8 compliant with consistent style and formatting due to the enforced usage of tools such as black, flake8, and pylint. This is particularly important as we onboard more contributors to the repo.