A CLI for ingesting data into the IOExplorer database.

## Project description

This repository contains the code for a command line tool to manage and ingest datasets into a Postgres database, for later use by the IOExplorer web application.

## Pre-Requisites

This CLI has three main depencies: docker and node, and python (specifically Python 3).

Installation instructions for each can be found below:

Once node is installed, you will also need to globally install some packages which are used to interact with the database.

npm i -g sequelize sequelize-cli pg


With some environments, you will get a permission error when you attempt to install these packages. There is a good article on how to fix your environment to avoid these errors.

To do a final check to make sure all software is installed, run the following:

docker --version
node --version
npm --version
sequelize --version
python --version


Note that python --version should return something starting with 3.

## Installing the CLI

Installing the CLI: simply,

pip install ioexplorer-dataloader
iodl --help


Make sure you have the most up to date version! Run the following if you have downloaded in the past:

pip install ioexplorer-dataloader --upgrade


## Example Workflows

### 1. Basics

##### Setting environment variables

To start, we need to set environment variables so that we can interact with a database. For example, to set up a development database, we can create a file called development.env with the following contents:

NODE_ENV=development
IOEXPLORER_MODE=development
IOEXPLORER_DEVELOPMENT_DATABASE_NAME=ioexplorerdb
IOEXPLORER_DEVELOPMENT_DATABASE_HOST=127.0.0.1
IOEXPLORER_DEVELOPMENT_DATABASE_PORT=5432



Then call set -a && source development.env. Note: set -a will cause all bash variables which are modified to be exported. This is the equivalent of calling export <line> for each <line> in development.env.

In order to use a production database, you would instead work with a file called production.env which would look slightly different

NODE_ENV=production
IOEXPLORER_MODE=production
IOEXPLORER_PRODUCTION_DATABASE_NAME=ioexplorerdb
IOEXPLORER_PRODUCTION_DATABASE_HOST=127.0.0.1
IOEXPLORER_PRODUCTION_DATABASE_PORT=5432
IOEXPLORER_GRAPHQL_URL=http://api:4000/graphql



Note: The differences here are that DEVELOPMENT is replaced with PRODUCTION for many of the environment variables. This is so you can have production and development variables loaded at the same time and easily toggle between the two contexts by setting IOEXPLORER_MODE and NODE_ENV. Also, the environment variable IOEXPLORER_GRAPHQL_URL is added in production, since the processes will be interconnected via a docker network in production rather than via localhost in development. Be sure to unset IOEXPLORER_GRAPHQL_URL in development, or else your graphql client will attempt to connet to the graphql api at the wrong url.

##### Starting a database.

Note that the database is not live yet, we only set our environment properly to connect to the database. Start the database by running

$iodl database start  The database should now be started. The database just a docker container running the postgres image, so you can see it being run with docker ps. ##### Opening a psql shell into a database. Now lets open a psql shell connected to our newly created database: $ iodl database shell
psql (11.0 (Debian 11.0-1.pgdg90+2))
Type "help" for help.

ioexplorerdb=# \dt
Did not find any relations.


The message Did not find any relations. lets us know that this database is completely empty and schemaless.

##### Applying migrations to our database.

The iodl CLI has a copy of all migrations used to produce the current production version of the IOExplorer database schema. To apply all these migrations run:

$iodl database migrate  Now if you open another psql shell and list the relations, you get the expected: $ iodl database shell
psql (11.0 (Debian 11.0-1.pgdg90+2))
Type "help" for help.

ioepxlorerdb=# \dt
List of relations
Schema |     Name      | Type  | Owner
--------+---------------+-------+-------
public | SequelizeMeta | table | root
public | cnas          | table | root
public | datasets      | table | root
public | fusions       | table | root
public | mutations     | table | root
public | samples       | table | root
public | subjects      | table | root
public | svs           | table | root
public | timelines     | table | root
(9 rows)


Now our database is ready to get some data.

##### Initializing a dataset

cd into a dataset to upload. Ask Ryan for one if you do not have any. TODO: upload example dataset.

The dataset should have the following directory structure:

.
├── data_clinical_patient.txt   (R)
├── data_clinical_sample.txt    (R)
├── data_CNA.txt
├── data_fusions.txt
├── data_expression.fpkm.txt
├── data_expression.rld.txt
├── data_expression.raw.txt
├── data_mutations_extended.txt
├── data_SV.txt
└── data_timeline.txt


Note: Only the files denoted with an (R) are actually required.

We now want to initialize the dataset. This step will

1. Run some quick validations to make sure the data structure is correct.
2. Collect some meta-information from the user.
3. Write a config.yaml file which stores information about this dataset and helps with ingestion.

Run:

(dataloader) ryan@galliumos:~/MSK/data/Hugo$iodl dataset init INFO: Initializing new dataset! ... Some success messages will appear here, or a prompt will ask you if you would like to continue with missing data. ... ? What is the dataset name? my-dataset ? What is a description of the dataset? this is a test dataset... ? Enter link to paper. http://google.com ? Who are you (person uploading data)? Ryan SUCCESS: Thanks! I made a file called config.yaml in this directory! Check it out and make sure everything looks OK!  ##### Ingesting a dataset With the config.yaml file already formed, ingesting the database is very simple. $ iodl dataset ingest


If there are any problems during ingestion, an error will be thrown and the data that already made it into the database (before the error) will be deleted. This will let you diagnose any problems with the data ingestion and re-attempt ingestion without messing with the state of the database.

### 2. Production on AWS

There are some subtle differences in the above when running on the production AWS server:

1. Instead of pip, you should use pip-3.6.
2. The environment variables are located in ~/ioexplorer/production.env.
3. iodl database shell will no longer work since the production database is running in a docker swarm. If you would like to shell into the datbase, you can find the id of the database container with docker container ls. Look for the line where the IMAGE is postgres:latest and the NAMES is something starting with ioexplorer_database. Copy the name string, which should look like ioexplorer_database.1.<a bunch of random characters>. Then, execute docker container exec -it ioexplorer_database.1.<a bunch of random characters> psql -d \$IOEXPLORER_PRODUCTION_DATABASE_NAME

## Project details

Uploaded source