Runners
... or how secator's internals work.
Last updated
... or how secator's internals work.
Last updated
A runner is at the core of secator
live processing capabilities. It handles the parsing, converting and processing of input options (CLI and library) and output items.
All runners inherit from secator.runners._base.Runner
.
Some built-in runners are available out-of-the-box:
Command
Run an external command and stream it's output.
Automatic command install.
Priviledged mode (sudo
).
Task
Run a task.
Remote mode (Celery).
Chunking on big inputs.
Direct calling from library.
Workflow
Run a DAG of tasks, defined in a YAML config file.
Remote mode (Celery).
Distributed (Celery).
Task chaining and parallel.
Re-use previous results as task inputs.
Scan
Run a DAG of workflows, defined in a YAML config file.
Distributed (Celery).
Workflow chaining.
Re-use previous results as workflow inputs.
Here is an overview of how a runner's lifecycle:
The Runner
lifecycle contains hooks that a user can plug into:
Base hooks:
before_init
: executed before the base runner's init starts.
on_init
: executed when the base runner's init is completed.
on_start
: executed when the runner has started running.
on_iter
: executed when the runner iterates.
on_end
: executed when the runner has finished running.
on_cmd
: runs when the mapped command is built [ Command
runner only ].
on_cmd_done
: runs when the command has finished running [ Command
runner only ].
on_line
: executed when a line is output to stdout or stderr [ Command
runner only ].
Item hooks:
on_item_pre_convert
: executed before an item is converted to an output type.
on_item
: executed when the runner emits an item.
on_duplicate
: runs after an item has been marked as a duplicate.
on_line
: executed when a line is emitted to stdout
[ Command
runner only ].
on_error
: executed when an error is emitted by the command [ Command
runner only ] .
on_
{serializer}
_loaded
: executed when a serializer has finished running. For instance, on_json_loaded
after the JSONSerializer
has finished running [ Command
runner only ] .
All hooks are defined with the @staticmethod
decorator and take self
as the first argument so that you can use the runner data in your hook implementation.
Item hooks take item
as the second argument and expect you to return the modified item.
There are two different ways of specifying hooks: static hooks (in the task definition class), dynamic hooks (passed to a runner at runtime), or drivers.
Static hooks are specified in the task specification class as staticmethod
s:
Dynamic hooks are specified at runtime by passing them to a runner.
Dynamic hooks are a library-only feature, they are not available in the CLI.
Here are examples of specifying dynamic hooks:
See Drivers.