- `PortfolioTargetCollection` avoid calling `Count` on
ConcurrentDictionary directly -> has to take all locks
- `SecurityChanges` change Union for Concat since constructor will call
HashSet
- Make `DynamicSecurityData` hold lazy data objects
- `RegisteredSecurityDataTypesProvider` avoid looping over all
registered types, adding `TryGetType`
- `Security.Update()` will no call group by on data since this data is
already grouped by type. Adding `ContainsFillForwardData` will allows to
be lazy and not re loop through the data unless necessary
- `DefaultAlphaHandler` will use the `static`
`Enumerable.Empty<Insight>` instance when possible
- `SubscriptionSynchronizer` will be lazy to construct the
`universeData` dictionary which is not used in most of the times. Will
use `Count` vs `Any` -> `Count` is known by the dictionary
- For python algorithms `JobQueue` will respect `AlgorithmLocation`, was
using unexisting `"algorithm-path-python"`
- Renaming files from `Schedule...` to `Scheduled...` to match algorithm
name.
- Removed `ScheduledEvents` set every 5 seconds
- Increased the every day `ScheduledEvents` from 200 to 600
- 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
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.
Martin Molinero