Errors and Exceptions
=====================

Errors and exceptions, including from user strategy, are caught by 
Blueshift and handled based on their category and recoverability. If 
an error is recoverable, Blueshift logs the error details and continues 
to run. If it is irrecoverable, Blueshift executes the exit processes 
(graceful exit, saving data etc.). Errors are usually classified based 
on their types and sources, as described below.

- User Strategy Errors
    Any errors generated from user strategy code are considered 
    fatal and no attempt is made to recover.
    
- Recoverable Errors
    Errors due to a drop in connection, or from broker API (e.g. invalid 
    or illegal orders, bad data from broker etc.) are considered
    recoverable. If a recoverable error is raised, the event loop stops the 
    current iteration at that point, and starts the next iteration 
    as usual in the next cycle. If the error continues to repeat, after 
    a certain threshold, it is upgraded to irrecoverable error.

- Irrecoverable Errors
    Any errors during the Blueshift event loop initialization is considered
    irrecoverable in a fail-fast manner (including connection errors, API
    errors or validation failures). Once the event loop starts normal
    cycles, any unexpected error (e.g. disk full, or server crash) or 
    any error upgraded (as described above) is considered fatal.



Error Handling in Strategy
--------------------------

It is usually a good idea to handle possible errors in the strategy 
itself, rather than leaving it to Blueshift. Strategy code can use the 
Python try-except block to achieve the same. Below are a list of useful
errors that can be handled from within the strategy.

.. py:module:: blueshift.errors

.. autoclass:: ValidationError

.. autoclass:: SymbolNotFound

.. autoclass:: ServerError

.. autoclass:: IllegalRequest

.. autoclass:: BrokerError

.. autoclass:: AlgoOrderError

.. autoclass:: APIError

.. autoclass:: BrokerConnectionError

.. autoclass:: InsufficientFund

.. autoclass:: OrderAlreadyProcessed

.. autoclass:: TradingControlError

.. autoclass:: BadDataError

.. autoclass:: NoDataForAsset

.. autoclass:: HistoryWindowStartsBeforeData

.. autoclass:: NoSuchPipeline

In addition to error handling, it is highly recommended that strategy 
code also applies input data sanity checks.