ISIMIP publisher

A command line tool to publish climate impact data from the ISIMIP project. This tool is used for the ISIMIP repository.

Setup

First create a virtual environment in the directory env using:

python3 -m venv env

Next, install isimip-publisher directly from GitHub using

pip install git+https://github.com/ISI-MIP/isimip-publisher

If you want to make changes to the source code, clone the repository and use pip install -e instead:

git clone git@github.com:ISI-MIP/isimip-publisher
pip install -e isimip-publisher

PostgreSQL has to be available and a database user and a database has to be created, and the pg_trgm extension needs to be activated:

CREATE USER "isimip_metadata" WITH PASSWORD 'supersecretpassword';
CREATE DATABASE "isimip_metadata" WITH OWNER "isimip_metadata";
\c isimip_metadata
CREATE EXTENSION pg_trgm;

Usage

The publisher has several options which can be inspected using the help option -h, --help:

usage: isimip-publisher [-h] [--config-file CONFIG_FILE] [-i INCLUDE_FILE] [-e EXCLUDE_FILE]
                        [-v VERSION] [--remote-dest REMOTE_DEST] [--remote-dir REMOTE_DIR]
                        [--local-dir LOCAL_DIR] [--public-dir PUBLIC_DIR]
                        [--archive-dir ARCHIVE_DIR] [--database DATABASE] [--mock MOCK]
                        [--restricted RESTRICTED] [--protocol-location PROTOCOL_LOCATIONS]
                        [--datacite-username DATACITE_USERNAME]
                        [--datacite-password DATACITE_PASSWORD]
                        [--datacite-prefix DATACITE_PREFIX]
                        [--datacite-test-mode DATACITE_TEST_MODE]
                        [--isimip-data-url ISIMIP_DATA_URL]
                        [--rights {None,CC0,BY,BY-SA,BY-NC,BY-NC-SA}] [--log-level LOG_LEVEL]
                        [--log-file LOG_FILE]
                        {list_remote,list_remote_links,list_local,list_public,list_public_links,match_remote,match_remote_links,match_local,match_public,match_public_links,fetch_files,write_local_jsons,write_public_jsons,write_link_jsons,insert_datasets,update_datasets,publish_datasets,archive_datasets,check,clean,update_search,update_tree,run,insert_doi,update_doi,register_doi,link_links,link_files,link_datasets,link,init,update_views}
                        ...

optional arguments:
  -h, --help            show this help message and exit
  --config-file CONFIG_FILE
                        File path of the config file
  -i INCLUDE_FILE, --include INCLUDE_FILE
                        Path to a file containing a list of files to include
  -e EXCLUDE_FILE, --exclude EXCLUDE_FILE
                        Path to a file containing a list of files to exclude
  -v VERSION, --version VERSION
                        Version date override [default: today]
  --remote-dest REMOTE_DEST
                        Remote destination to fetch files from, e.g. user@example.com
  --remote-dir REMOTE_DIR
                        Remote directory to fetch files from
  --local-dir LOCAL_DIR
                        Local work directory
  --public-dir PUBLIC_DIR
                        Public directory
  --archive-dir ARCHIVE_DIR
                        Archive directory
  --database DATABASE   Database connection string, e.g.
                        postgresql+psycopg2://username:password@host:port/dbname
  --mock MOCK           If set to True, no files are actually copied. Empty mock files are used
                        instead
  --restricted RESTRICTED
                        If set to True, the files are flaged as restricted in the database.
  --protocol-location PROTOCOL_LOCATIONS
                        URL or file path to the protocol
  --datacite-username DATACITE_USERNAME
                        Username for DataCite
  --datacite-password DATACITE_PASSWORD
                        Password for DataCite
  --datacite-prefix DATACITE_PREFIX
                        Prefix for DataCite
  --datacite-test-mode DATACITE_TEST_MODE
                        If set to True, the test version of DataCite is used
  --isimip-data-url ISIMIP_DATA_URL
                        URL of the ISIMIP repository [default: https://data.isimip.org/]
  --rights {None,CC0,BY,BY-SA,BY-NC,BY-NC-SA}
                        Rights/license for the files [default: None]
  --log-level LOG_LEVEL
                        Log level (ERROR, WARN, INFO, or DEBUG)
  --log-file LOG_FILE   Path to the log file

subcommands:
  valid subcommands

  {list_remote,list_remote_links,list_local,list_public,list_public_links,match_remote,match_remote_links,match_local,match_public,match_public_links,fetch_files,write_local_jsons,write_public_jsons,write_link_jsons,insert_datasets,update_datasets,publish_datasets,archive_datasets,check,clean,update_search,update_tree,run,insert_doi,update_doi,register_doi,link_links,link_files,link_datasets,link,init,update_views}

The different steps of the publication process are covered by subcommands, which can be invoked separately.

# list remote files
isimip-publisher list_remote <path>

# match remote datasets
isimip-publisher match_remote <path>

# copy remote files to LOCAL_DIR
isimip-publisher fetch_files <path>

# create a JSON file with metadata for each dataset and file
isimip-publisher write_local_jsons <path>

# finds dataset and file and ingest their metadata into the database
isimip-publisher ingest_datasets <path>

# copy files from LOCAL_DIR to PUPLIC_DIR
isimip-publisher publish_datasets <path>

# copy files from PUBLIC_DIR to ARCHIVE_DIR
isimip-publisher archive_datasets <path>

# insert a new doi resource
isimip-publisher ingest_doi <resource-path>

# register a DOI resource with datacite
isimip-publisher register_doi <DOI>

<path> starts from REMOTE_DIR, LOCAL_DIR, etc., and must start with <simulation_round>/<product>/<sector>. After that more levels can follow to restrict the files to be processed further.

<resource-path> is the path to a JSON file containing metadata on the local disk.

match_remote, fetch_files, write_jsons, ingest_datasets, and publish_datasets can be combined using run:

isimip-publisher run <path>

For all commands a list of files with absolute pathes (as line separated txt file) can be provided to restrict the files processed, e.g.:

isimip-publisher -e exclude.txt -i include.txt run <path>

Default values for the optional arguments are set in the code, but can also be provided via:

• a config file given by --config-file, or located at isimip.conf, ~/.isimip.conf, or /etc/isimip.conf. The config file needs to have a section isimip-publisher and uses lower case variables and underscores, e.g.:

  [isimip-publisher]
  log_level = ERROR
  mock = false
  
  remote_dest = localhost
  remote_dir = /path/to/remote/
  local_dir = /path/to/local/
  public_dir = /path/to/public/
  archive_dir = /path/to/public/
  database = postgresql+psycopg2://USER:PASSWORD@host:port/DBNAME
  
  protocol_locations = '/path/to/isimip-protocol-3/output/ /path/to/isimip-protocol-3/output/'
  

• environment variables (in caps and with underscores, e.g. MOCK).

Scripts/Notebooks

The different functions of the tool can also be used in Python scripts or Jupyter Notebooks. Before any functions are called, the global settings object needs to be initialized, e.g.:

from isimip_publisher.main import init_settings
from isimip_publisher.utils.database import (init_database_session, retrieve_datasets)

path = 'ISIMIP3b/OutputData/marine-fishery_global'

settings = init_settings(config_file='~/data/isimip/isimip.conf')

session = init_database_session(settings.DATABASE)

datasets = retrieve_datasets(session, path)

...

Test

Install test dependencies:

pip install -r requirements/pytest.txt

Copy .env.pytest to .env. This sets the environment variables to the directories in testing.

Run:

pytest

Run a specific test, e.g.:

pytest isimip_publisher/tests/test_commands.py::test_empty

Run tests with coverage:

pytest --cov=isimip_publisher

Database schema

The database schema is automatically created when insert_datasets or init is used the first time. The tool creates 3 main tables, one for the datasets, one for the files (in each dataset), and one for the resources, for which DOI are created.:

                          Table "public.datasets"
   Column    |            Type             | Collation | Nullable | Default
-------------+-----------------------------+-----------+----------+---------
 id          | uuid                        |           | not null |
 target_id   | uuid                        |           |          |
 name        | text                        |           | not null |
 path        | text                        |           | not null |
 version     | character varying(8)        |           | not null |
 size        | bigint                      |           | not null |
 specifiers  | jsonb                       |           | not null |
 identifiers | text[]                      |           | not null |
 public      | boolean                     |           | not null |
 tree_path   | text                        |           |          |
 rights      | text                        |           |          |
 created     | timestamp without time zone |           |          |
 updated     | timestamp without time zone |           |          |
 published   | timestamp without time zone |           |          |
 archived    | timestamp without time zone |           |          |

                             Table "public.files"
    Column     |            Type             | Collation | Nullable | Default
---------------+-----------------------------+-----------+----------+---------
 id            | uuid                        |           | not null |
 dataset_id    | uuid                        |           |          |
 target_id     | uuid                        |           |          |
 name          | text                        |           | not null |
 path          | text                        |           | not null |
 version       | character varying(8)        |           | not null |
 size          | bigint                      |           | not null |
 checksum      | text                        |           | not null |
 checksum_type | text                        |           | not null |
 netcdf_header | jsonb                       |           |          |
 specifiers    | jsonb                       |           | not null |
 identifiers   | text[]                      |           | not null |
 created       | timestamp without time zone |           |          |
 updated       | timestamp without time zone |           |          |

                        Table "public.resources"
  Column  |            Type             | Collation | Nullable | Default
----------+-----------------------------+-----------+----------+---------
 id       | uuid                        |           | not null |
 doi      | text                        |           | not null |
 title    | text                        |           | not null |
 version  | text                        |           |          |
 paths    | text[]                      |           | not null |
 datacite | jsonb                       |           | not null |
 created  | timestamp without time zone |           |          |
 updated  | timestamp without time zone |           |          |

The many-to-many relation between datasets and resources is implemented using a seperate table:

          Table "public.resources_datasets"
   Column    | Type | Collation | Nullable | Default
-------------+------+-----------+----------+---------
 resource_id | uuid |           |          |
 dataset_id  | uuid |           |          |

Additional tables are created for the search and tree functionality of the repository.

                           Table "public.search"
   Column   |            Type             | Collation | Nullable | Default
------------+-----------------------------+-----------+----------+---------
 dataset_id | uuid                        |           | not null |
 vector     | tsvector                    |           | not null |
 created    | timestamp without time zone |           |          |
 updated    | timestamp without time zone |           |          |

                           Table "public.trees"
  Column   |            Type             | Collation | Nullable | Default
-----------+-----------------------------+-----------+----------+---------
 id        | uuid                        |           | not null |
 tree_dict | jsonb                       |           | not null |
 created   | timestamp without time zone |           |          |
 updated   | timestamp without time zone |           |          |

Two materialized views are used to allow a fast lookup to all identifiers (with the list of corresponding specifiers), as well as all words (the list of tokens for the search):

       Materialized view "public.identifiers"
   Column   | Type | Collation | Nullable | Default
------------+------+-----------+----------+---------
 identifier | text |           |          |
 specifiers | json |           |          |

        Materialized view "public.words"
 Column | Type | Collation | Nullable | Default
--------+------+-----------+----------+---------
 word   | text |           |          |