Skip to content

Organise data in AiiDA with a familiar file-system like way

License

Notifications You must be signed in to change notification settings

zhubonan/aiida-grouppathx

Repository files navigation

Build Status Coverage Status PyPI version

aiida-grouppathx

AiiDA plugin provides the GroupPathX class.

This plugin was kickstarted using AiiDA plugin cutter, intended to help developers get started with their AiiDA plugins.

Features and usage

Interactive example at: Binder

Organize AiiDA data with a file-system-like interface

This package provides a enhanced version of GroupPath - GroupPathX. The main feature is that it allows nodes stored under a group to be named by an alias. This way, one can address a specific Node as GroupPath('mygroup/structure1'). In addition, a show_tree method is provided for visualising the content of a specific GroupPathX, similar to the command line tool tree that works on the file system. The goal is to provide a way for managing data with an interface what is similar to a file system based approach.

tree aiida_grouppathx

aiida_grouppathx
├── __init__.py
├── pathx.py
└── __pycache__
    ├── __init__.cpython-38.pyc
    └── pathx.cpython-38.pyc

In analogy:

from aiida_grouppathx import GroupPathX
path = GroupPathX('group1')
path.get_or_create_group()
path['group2'].get_or_create_group()
path.add_node(Int(1).store(), 'int1')
path['group2'].add_node(Int(1).store(), 'int2')

path.show_tree()

gives

group1
├── group2
│   └── int2 *
└── int1 *

where the * highlights that a leaf is a Node rather than a group. This kind of mark up can be customised, for example, to show the status of workflow nodes.

def decorate_name(path):
    if path.is_node:
        return ' ' + str(path.get_node())
path.show_tree(decorate_name)

gives:

group1
├── group2
│   └── int2  uuid: de79d244-d3bb-4f61-9d3a-b3f09e1afb72 (pk: 7060) value: 1
└── int1  uuid: e2f70643-0c25-4ae5-929a-a3e055969d10 (pk: 7059) value: 1

Multiple decorators can be combined

from aiida_grouppathx import decorate_with_group_names, decorate_with_label decorate_with_uuid_first_n

path.show_tree(decorate_with_group_names, decorate_with_label, decorate_with_uuid_first_n())

output:

group1
├── group2
│   └── int2 group1/group2 |  | de79d244-d3b
└── int1 group1 |  | e2f70643-0c2

The stored nodes can be access through:

group1['group2/int2'].get_node()  # Gives node de89d2
group1.browse.group2.int2().get_node()  # Also gives node de89d2

and also

path.browse.<tab>
path.browse.int1()     # To access the `group1/int1` path
path.browse.int1().get_node()     # To access the `group1/int1` node

Please see the pathx.py for the extended methods, and the official documentation for the concept of GroupPath.

The package does not change how Group and Node operates in the AiiDA. It is only built on top of the existing system as an alternative way to access the underlying data.

The nodes under a GroupPathX can be be added using add_node or add_nodes methods. Existing node can be deleted or rename using unlink or rename methods. Note that unlink does not dissociate the node from the actual group, but only removes its alias stored inside the extras that is used by this package

Keeping a limited number of active jobs

When performing high-throughput workflows, it may be necessary to limit the number of active jobs to avoid overloading the remote computer. We provide a GroupLauncher method to achieve this. The idea is to use a group path of jobs inputs as inputs, and store the launched jobs in another GropuPathX, for example:

from aiida_grouppathx import GroupPathX, GroupLauncher
launcher = GroupLauncher(GroupPathX('target_group'), max_concurrent=200, callback=launch_relax, source_gp=GroupPathX('source_group'))
launcher.launch()

Here the a callback function launch_relax is called for each node in the source_group with the node and its alias as the input arguments. This callback function should launch the calculation/workchain and return a tuple of the launched process and its alias to store inside the target_group.

Instead of using a GroupPathX as the source, one can also use a list of nodes and alias pairs the source by passing them through the source_key_obj_pairs keyword argument.

Installation

pip install aiida-grouppathx
verdi quicksetup  # better to set up a new profile

Development

git clone https://github.com/zhubonan/aiida-grouppathx .
cd aiida-grouppathx
pip install --upgrade pip
pip install -e .[pre-commit,testing]  # install extra dependencies
pre-commit install  # install pre-commit hooks
pytest -v  # discover and run all tests

See the developer guide for more information.

License

MIT

Contact

zhubonan@outlook.com

About

Organise data in AiiDA with a familiar file-system like way

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages