- Requires a new PythonNet 1.0.5.15 package where the different `.dll` are in a
specific folder: `\win` `\linux` and `\osx`
- Removed not present `decimal.py` from `Algorithm.Python` project. It
was moved into `Common`.
- Replace `xbuild` for `msbuild` required for using the `System.Runtime.InteropServices`.
Also note the `xbuild` on travis prints:
> >>>> xbuild tool is deprecated and will be removed in future updates, use msbuild instead <<<<
In the new package:
- C# decimal conversion will use C# double and python float due to the big performance impact of converting C# decimal to python decimal;
- This commit is related to PR 19 in QC/pythonnet
- C# decimal will be cast to C# double and converted into python
float
- Adding new `decimal.py` into the python algorithm project. This is
required for backwards compatibility with users performing operations
over expected decimal types (like `Price`)
- Updating two python regression test algorithms using custom python
execution models to be aware and ignore floating point precision errors
when handling order sizing.
- The HistoryRequestBenchmark will now perform a daily and minute history
request at end of day.
- C# performance is at ~15k data points per second
- Python performance is at ~11k data points per second
Explictly implements the indexert `this[string]` to all classed that inherit from `DataDictionary` since pythonnet was not able to access the indexer from the parent class.
- Changes DividentAlgorithm.py to test the fix.
- Moving `UniverseSelection.EnsureCurrencyDataFeeds` call into the
`IResultHandler` implementation through usage of the new `SetupHandlerHelper`
class, that will also set an initial conversion rate if none present.
- Adding regression test, that reproduces original issue
- Disabling C# and Py `BasicTemplateIntrinioEconomicData` regression
test. Free user credentials are invalid because Intrinio has now a 30 day trial for free users
- `CashBook[NullCurrency] { get; }` will throw an exception
- Revert `Currencies.USD` changes in user facing algorithms
- Improve some documentation
- Revert some format changes
- Adding more asserts for regression test
- Adding new regression tests using a custom fee model which returns
`OrderFee.Zero`
- Adding a non-usd account currency test to the cash book tests
- Adding some unit tests for `NullCurrency` and `OrderFee.Zero`
- Removing `AccountCurrency` from `Cash` and `Brokerage` classes.
`ICurrencyConverter` will now provide the `AccountCurrency`
- Adding new static `OrderFee.Zero` which will return a 0 order fee in
`NullCurrency`
- Adding static `Currencies.USD` value, replacing all "USD".
- Addin new static `Currencies.NullCurrency`
- Updating Bitfinex `FeeModel` so it return fees in quote currency.
Adding unit tests
- Refactoring `IFeeModel`. *This is a breaking change* for implementations
inheriting directly from the interface. Deleting old and adding a new method
`OrderFee GetOrderFee(OrderFeeParameters parameters)` that will use a parameter
and a result object.
- Refactoring `CashAmount` so it does not embed a `ICurrencyConverter`
instance.
- Updating unit tests
- The `Security.QuoteCurrency`, a `Cash` instance, will provide access
to the `AccountCurrency` as a property.
- Will maintain backwards compatibility with old python custom
FeeModels, Adding unit test.
> Note that for now, consumers will ignore the currency, as before, and
directly consume the amount
- Modifying `IFillModel` interface removing old methods and adding new
method `Fill Fill(FillModelParameters)`. This is a breaking change.
- Adding new `PythonWrapper` property for the `FillModel` base class.
This is required due to a limitation in PythonNet:
- Given C# class T has `virtual` methods A and B. Where method A
calls method B. And given custom python class L inherits class T.
And overrides method B. When class L calls
base method A (of class T). And when method A internally calls method B.
It will call C# implementation, not the python override. This issue
is solved going back to the `PythonWrapper`. Adding unit tests.
- Adding new `Parameters` property for the `FillModel` base class that will
be set by the call to `Fill()`. The `Parameters` property will be used by
the modified `XxxxFill()` implementations
- Adding new `Fill` result object for the `Fill(FillModelParameters)`
method
- Adding new check before removing a `SubscriptionDataConfig` due to the FillModels consuming the configuration collection when determining which Price to use. WIll now only remove the `SDC` if the symbol was removed from the selecting `universe`, this will avoid the case where the symbol is never deselected and the subscription ends, which happens at the end of all executions.
- Adding unit tests showcasing retro compatibility.
- Enabling C# `CustomModelsAlgorithm` as a regression test. Python
version returns a different result due to random number generation.
By using the python object parant class, which is either `PythonQuandl` or `PythonData`, instead of `DynamicData`, the `AlgorithmManager.Stream` method can find a matching subcription data configuration used to create a data feed packet.
Closes#2694
The C# version was supposed to handle python modules, but when they inherit from a C# module, pythonnet send them as C# objects. Consequently, they are not wrapped and cannot be used. The python version of `CompositeRiskManagementModel` solves the issue.
- Implements python version of `MaximumUnrealizedProfitPercentPerSecurity`
- Updates `CompositeRiskManagementModelFrameworkAlgorithm` in order to use python risk model.
`DataDictionary.GetValue` is meant to be used as an alternative to `DataDictionary.TryGetValue`. It was created due to limitations to the python implementation.
- Removed usages of algorithm.Securities.key as a parameter for the
`ManualUniverseSelectionModel()` since those securities, added through
`AddXXXX` calls will be managed by the `UserDefinedUniverse`. This was
causing for Universes to try to add the same subscription requests
- Adding new empty constructor for ManualUniverseSelectionModel,
required for Python
- ManualUniverse will return any existing SDC for the
symbol. This is for maintaining existing behavior and
preventing breaking changes: Specifically motivated by usages of
Algorithm.Securities.Keys as constructor parameter of the
ManualUniverseSelectionModel, since those Symbols added by Addxxx()
calls will already be managed by the UserDefinedUniverse
- Making some format modifications to aling with used Lean formatting
We didn't experience the expected performance improvements. Locally under
unit test there was aboout an order of magnitude throughput increase, but
when run against the history benchmark, this new approach was 60% slower.
We're reverting this for now to perform further analysis and better
understand the performance profiling of the python history stack.
1. Apply the pattern used in `EqualWeightingPortfolioConstructionModel`
2. Change the logic to compute the views from the insights.
3. Change the logis to compute the posterior mean and covariance
Use `UnconstrainedMeanVariancePortfolioOptimizer` in `BlackLittermanPortfolioOptimizationFrameworkAlgorithm` to bypass the difference in regression tests with `IPortfolioOptimizer` that rely on different algorithms in C# and python.
Adds unit tests for BLOPCV to test the implementation against Black and Litterman 1999 paper.
Martin Molinero