ABCD

Database storage and discovery of atomistic data.

Take a look at the examples.md file for.. examples!

Main features:

• Configurations that consist of atom positions, elements, forces, and various metadata are stored as a dictionary by a MongoDB backend.
• There is no predefined schema, any combination of keys are allowed for all configurations.
• Two modes: "discovery" and "download". Both use filter-type queries, but in "discovery" mode, summary statistics of the configurations that pass the filter are reported. In "download" mode, the matching configurations are downloaded and exported to a file.
• The "discovery" mode can be used to learn what keys exist in the set of configurations that have passed the current quiery filter. The user can use this to refine the query.
• Complex queries on dictionary key-value pairs are allowed, and their logical combinations.

Installation

creating tables and views

$ pip install git+https://github.com/libAtoms/abcd.git

Setup

If you have an already running mongo server, or install your own, they you are ready to go. Alternatively,

docker run -d --rm --name abcd-mongodb -v <path-on-your-machine-to-store-database>:/data/db -p 27017:27017 mongo

will download and install a docker and run a database in it.

To connect to a mongodb that is already running, use

abcd login mongodb://localhost

If you are running abcd inside a docker, and want to connect to a mongodb outside that docker use something like this (example is for Mac OS):

abcd login mongodb://docker.for.mac.localhost

The above login command will place create an ~/.abcd file with the following contents:

{"url": "mongodb://localhost"}

Remote access

You can set up an abcd user on your machine where the database is running, and then access it remotely for discovering data. Make sure you have the ~/.abcd file created for this user, then put this in the .ssh/authorized_keys file (substituting your public key for the last part):

command="/path/to/abcd --remote  ${SSH_ORIGINAL_COMMAND}",no-port-forwarding,no-X11-forwarding,no-agent-forwarding,no-pty ssh-rsa <public-key> your@email

Then you'll be able to access the database remotely using, e.g.

ssh [email protected] summary

GUI through a browser + visualisation

The database has a simple GUI, coupled with a visualiser. Data for now needs to be uploaded on the command line, but query can be done through the browsers. Instructions below (they include running abcd from a docker too, but of course you can run it outside the docker as well. )

Usage in docker

Currently a manual uploaded image is available, that was built on 7/2/2020 by Tamas K. Stenczel. To access it:

1. pull the image

   docker pull stenczelt/projection-abcd:latest
   

2. create a docker network, which enables the containers to communicate with each other and the outside world as well

   docker network create --driver bridge abcd-network
   

3. run the mongo (ABCD) and the visualiser as well

   docker run -d --rm --name abcd-mongodb-net -v <path-on-your-machine-to-store-database>:/data/db -p 27017:27017 --network abcd-network mongo
   
   docker run -it --rm --name visualiser-dev -p 9999:9999 --network abcd-network stenczelt/projection-abcd
   

   NB: You need a a directory where the database files are kept locally and you need to connect this to the mongo container. More info about this can be found in the original ABCD repo

This will start the visualiser with ABCD integration! Have fun!

After usage, for cleanup:

docker stop visualiser-dev abcd-mongodb-net         # stop the containers
docker rm visualiser-dev abcd-mongodb-net           # remove them if --rm did not
docker network rm abcd-network                      # remove the docker network