What is the ETA on being able to import separate files into algorithms so we can split code off and share it between our algorithms? For example, I would like to put all my custom factors into a separate file instead of trying to copy/paste all of them between all my test algorithms and trying to keep the changes in sync. It is a management nightmare! The lack of this functionality dramatically reduces the usefulness of the Pipeline API and custom factors.
Going beyond custom factors, I would like to put together a basic class-based algorithm structure I can reuse to quickly write new test algorithms while recycling all my position management and entry/exit code. Again, this is essentially not possible at the moment simply due to the code management nightmare of trying to manually sync up any changes between all the different test algorithms.
In short, the lack of the most basic code management functionality of splitting it out into separate files for inclusion dramatically hinders the use of this platform and basically makes it look like more of a toy / hobby than a real, professional trading algorithm development platform. If I write all the custom factors I use in my algorithms elsewhere and I write the framework for how I execute trades and balance my portfolios, it will easily be around 10,000 lines of code. Do you really expect us to copy 10,000 lines of code from one algorithm to the next every time we want to test something? Do you really expect us to try to keep 10,000 lines of code in sync between multiple test and production algorithms every time we make a change to that framework code or those custom factors? The #1 biggest reason I have not yet transferred my trading algorithms over here and away from the paid service I currently use is because of this issue. The #2 issue is because of the complete lack of almost all useful factors for filtering stocks. Of course #2 can be fixed by me writing all of them myself, which I don't have an issue with. Except, that then requires issue #1 to be fixed.
I would really, really, like to use Quantopian because writing actual code to manage my algorithms would provide far more flexibility than the current system I use. In addition, it's free, where the current system I use is costing me over $2k/yr. But this lack of functionality makes it almost impossible to use it effectively.