pyworkflow.utils.retry_streaming module

pyworkflow.utils.retry_streaming.is_sqlite_lock_error(exc: Exception) → bool

Return True if exc is a SQLite OperationalError indicating locked/busy DB.

pyworkflow.utils.retry_streaming.refreshStreamState(setObj) → None

Update the in-memory stream state of a Set from the database.

Uses safeIsStreamOpen() to read the current state from an independent connection, then patches the in-memory _streamState attribute so that subsequent setObj.isStreamOpen() calls return the fresh value without needing loadAllProperties().

pyworkflow.utils.retry_streaming.retry_on_sqlite_lock(max_attempts: int = 15, initial_delay: float = 0.25, backoff_factor: float = 1.7, max_delay: float = 10, jitter: float = 0.05, log=None, predicate=<function is_sqlite_lock_error>)

Decorator that retries the wrapped function when SQLite signals lock/busy.

Behavior:

• Exponential backoff with small jitter to avoid retry synchronization.

• Only retries when predicate(exc) is True (i.e., lock/busy by default).

• Propagates any non-retriable exceptions immediately.

• If max attempts are exhausted, re-raises the original exception.

Parameters:

max_attempts Max number of tries (including the first call). initial_delay Initial sleep before the first retry (seconds). backoff_factor Multiply delay on each retry. max_delay Cap the delay (seconds). jitter Random noise added to delay (seconds). log Optional logger (e.g., logging.getLogger(__name__)). predicate Function deciding whether the exception should be retried.

pyworkflow.utils.retry_streaming.safeIsStreamOpen(setObj) → bool

Check if a Set’s stream is open using an independent SQLite connection.

Opens a short-lived, read-only connection to the Set’s backing SQLite file and queries ONLY the _streamState property. This avoids:

• Reusing the Set’s internal mapper connection (no mapper contention)

• Holding protocol-level locks during the check

• The heavyweight loadAllProperties() call that reads ALL properties

Safe to call from any thread while the producer may be writing concurrently. Returns False (stream closed) on any error, since a missing/corrupt DB means no more data is coming.