pymatgen-db 2022.5.2

Major change

From v2021.5.13, pymatgen-db is now a proper namespace add-on to pymatgen. In other words, you no longer import from matgendb but rather pymatgen.db.

Stable version

The easiest way to install pymatgen-db on any system is to use pip, as follows:

pip install pymatgen-db

mgdb --help

Initial setup

The first step is to install and setup MongoDB on a server of your choice. The MongoDB manual is an excellent place to start. For the purposes of testing out the tools here, you may simply download the binary distributions corresponding to your OS from the MongoDB website, and then running the following commands:

# For Mac and Linux OS.
mkdir test_db && mongod --dbpath test_db

This will create a test database and start the Mongo daemon. Once you are done with testing, you can simply press Ctrl-C to stop the server and delete the “test_db” folder. Running a Mongo server this way is insecure as Mongo does not enable authentication by default. Please refer to the MongoDB manual when setting up your production database.

After your server is up, you should create a database config file by running the following command:

mgdb init -c db.json

This will prompt you for a few parameters to create a database config file, which will make it much easier to use mgdb in future. Note that the config file name can be anything of your choice, but using “db.json” will allow you to use mgdb without explicitly specifying the filename in future. If you are just testing using the test database, simply hit Enter to accept the defaults for all settings.

For more advanced use of the “db.json” config file (e.g., specifying aliases, defaults, etc., please refer to the following sample.

Inserting calculations

To insert an entire directory of runs (where the topmost directory is “dir_name”) into the database, use the following command:

# Note that "-c db.json" may be omitted if the config filename is the
# current directory under the default filename of db.json.

mgdb insert -c db.json dir_name

A sample run has been provided for download for testing purposes. Unzip the file and run the above command in the directory.

Querying a database

Sometimes, more fine-grained querying is needed (e.g., for subsequent postprocessing and analysis).

The mgdb script allows you to make simple queries from the command line:

# Query for the task id and energy per atom of all calculations with
# formula Li2O. Note that the criteria has to be specified in the form of
# a json string. Note that "-c db.json" may be omitted if the config
# filename is the current directory under the default filename of db.json.

mgdb query -c db.json --crit '{"pretty_formula": "Li2O"}' --props task_id energy_per_atom

For more advanced queries, you can use the QueryEngine class for which an alias is provided at the root package. Some examples are as follows:

>>> from pymatgen.db import QueryEngine
# Depending on your db.json, you may need to supply keyword args below
# for `port`, `database`, `collection`, etc.
>>> qe = QueryEngine()

#Print the task id and formula of all entries in the database.
>>> for r in qe.query(properties=["pretty_formula", "task_id"]):
...     print "{task_id} - {pretty_formula}".format(**r)
...
12 - Li2O

# Get a pymatgen Structure from the task_id.
>>> structure = qe.get_structure_from_id(12)

# Get pymatgen ComputedEntries using a criteria.
>>> entries = qe.get_entries({})

The language follows very closely to pymongo/MongoDB syntax, except that QueryEngine provides useful aliases for commonly used fields as well as translation to commonly used pymatgen objects like Structure and ComputedEntries.

Extending pymatgen-db

Currently, pymatgen-db is written with standard VASP runs in mind. However, it is perfectly extensible to any kind of data, e.g., other kinds of VASP runs (bandstructure, NEB, etc.) or just any form of data in general. Developers looking to adapt pymatgen-db for other purposes should look at the VaspToDbTaskDrone class as an example and write similar drones for their needs. The QueryEngine can generally be applied to any Mongo collection, with suitable specification of aliases if desired.