flowdump

用于与flowview一起使用的NiPype-1工作流序列化器。

安装

flowdump可在PyPI上找到，并且可以使用pip进行安装

pip install flowdump

用法

使用flowdump就像在NiPype-1工作流对象上调用run_and_save_workflow一样简单。这将执行工作流并将执行前后的工作流数据保存到JSON文件中。

import nipype.pipeline.engine as pe  # pypeline engine
from flowdump import run_and_save_workflow

# Typical NiPype workflow creation
amazing_workflow = pe.Workflow(name="main_workflow")
# amazing_workflow.connect(...)
# amazing_workflow.connect(...)
# amazing_workflow.connect(...)

# Let `flowdump` execute and save pre- and post-execution data.
run_and_save_workflow(
    amazing_workflow,
    out_dir='my/target/dir'
)

高级

如果需要对工作流执行有更多控制，可以手动序列化工作流。

import nipype.pipeline.engine as pe  # pypeline engine
import os.path
from flowdump import WorkflowJSONMeta, save_workflow_json

# Typical NiPype workflow creation
amazing_workflow = pe.Workflow(name="main_workflow")
# amazing_workflow.connect(...)
# amazing_workflow.connect(...)
# amazing_workflow.connect(...)

# Create workflow metadata object (traces execution time and stage)
workflow_meta = WorkflowJSONMeta(
    pipeline_name='My amazing pipeline',
    stage='pre'
)
# Dump pre-execution workflow
save_workflow_json(
    filename=os.path.join('my/target/dir', workflow_meta.filename()),
    workflow=amazing_workflow,
    meta=workflow_meta
)

# Execute NiPype workflow
amazing_workflow_result = amazing_workflow.run()

# Update metadata
workflow_meta.stage = 'post'
# Dump post-execution workflow
save_workflow_json(
    filename=os.path.join('my/target/dir', workflow_meta.filename()),
    workflow=amazing_workflow_result,
    meta=workflow_meta
)

自定义字段序列化

可以为具有自定义NiPype节点类型的项目实现自定义序列化器。序列化器是一个函数，它接受默认的flowdump序列化函数（可选地回退到）以及一个对象，并返回一个可序列化为JSON的对象。

def my_custom_serializer(
        flowdump_serializer: Callable[[object], object],
        obj: object
):
    if isinstance(obj, MyType):
        return my_make_string(obj)
    return flowdump_serializer(obj)

save_workflow_json(
    filename=os.path.join('my/target/dir', workflow_meta.filename()),
    workflow=amazing_workflow_result,
    meta=workflow_meta
)