Python Logging

Use deltafunnel.init_logging() to route Delta Funnel's Rust tracing events through standard-library logging before running a workflow.

Enable the logging bridge

Configure Python logging first, then initialize the bridge:

import logging
import deltafunnel

logging.basicConfig(level=logging.INFO)
installed = deltafunnel.init_logging()

init_logging() installs a process-global bridge to logging.getLogger("deltafunnel"). It returns True when the bridge is installed and False when a global Rust tracing subscriber is already installed.

The first successful call selects the bridge configuration for the process. Later calls that return False do not replace it. Pass logger="name" on the first call to use a different Python logger.

Select diagnostic events

Pass a tracing filter on the first call when you need more detail:

deltafunnel.init_logging(
    "delta_funnel=debug,delta_kernel=debug,object_store=debug,arrow_sql_server=debug"
)

You can also set the DELTAFUNNEL_LOG environment variable instead of passing a filter. An explicit filter argument takes precedence. When neither is set, Delta Funnel uses delta_funnel=info,arrow_sql_server=info.

DEBUG events must pass both the Rust tracing filter and Python logging levels. For example, this configuration enables terminal Parquet I/O and execution profile summaries:

import logging
import deltafunnel

handler = logging.StreamHandler()
handler.setLevel(logging.DEBUG)

logger = logging.getLogger("deltafunnel")
logger.setLevel(logging.DEBUG)
logger.addHandler(handler)
logger.propagate = False

deltafunnel.init_logging("delta_funnel=debug")

DELTAFUNNEL_LOG and the filter passed to init_logging() control only the Rust tracing filter. They do not lower the selected Python logger or handler level. A Python INFO threshold still discards forwarded DEBUG records.

For event fields and terminal outcome rules, see Inspect terminal Parquet I/O and Inspect terminal execution profiles.

Keep Python in control

Delta Funnel does not configure Python handlers, formatters, levels, files, or external exporters. Existing Datadog, OpenTelemetry, JSON logging, file logging, pytest capture, notebook, and framework integrations continue to own Python logging output.

For report fields and a safe bug-report checklist, see Troubleshoot a failed run.