Skip to content

gdcc/python-dvuploader

Repository files navigation

Dataverse Uploader
PyPI version Build Badge Build Badge

Python equivalent to the DVUploader written in Java. Complements other libraries written in Python and facilitates the upload of files to a Dataverse instance via Direct Upload.

Features

  • Parallel direct upload to a Dataverse backend storage
  • Files are streamed directly instead of being buffered in memory
  • Supports multipart uploads and chunks data accordingly

DVUploader.mov

Getting started

To get started with DVUploader, you can install it via PyPI

python3 -m pip install dvuploader

or by source

git clone https://github.com/gdcc/python-dvuploader.git
cd python-dvuploader
python3 -m pip install .

Quickstart

Programmatic usage

In order to perform a direct upload, you need to have a Dataverse instance running and a cloud storage provider. The following example shows how to upload files to a Dataverse instance. Simply provide the files of interest and utilize the upload method of a DVUploader instance.

import dvuploader as dv


# Add file individually
files = [
    dv.File(filepath="./small.txt"),
    dv.File(directory_label="some/dir", filepath="./medium.txt"),
    dv.File(directory_label="some/dir", filepath="./big.txt"),
    *dv.add_directory("./data"), # Add an entire directory
]

DV_URL = "https://demo.dataverse.org/"
API_TOKEN = "XXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX"
PID = "doi:10.70122/XXX/XXXXX"

dvuploader = dv.DVUploader(files=files)
dvuploader.upload(
    api_token=API_TOKEN,
    dataverse_url=DV_URL,
    persistent_id=PID,
    n_parallel_uploads=2, # Whatever your instance can handle
)

Command Line Interface

DVUploader ships with a CLI ready to use outside scripts. In order to upload files to a Dataverse instance, simply provide the files of interest, persistent identifier and credentials.

Using arguments

dvuploader my_file.txt my_other_file.txt \
           --pid doi:10.70122/XXX/XXXXX \
           --api-token XXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX \
           --dataverse-url https://demo.dataverse.org/ \

Using a config file

Alternatively, you can also supply a config file that contains all necessary information for the uploader. The config file is a JSON/YAML file that contains the following keys:

  • persistent_id: Persistent identifier of the dataset to upload to.
  • dataverse_url: URL of the Dataverse instance.
  • api_token: API token of the Dataverse instance.
  • files: List of files to upload. Each file is a dictionary with the following keys:
    • filepath: Path to the file to upload.
    • directory_label: Optional directory label to upload the file to.
    • description: Optional description of the file.
    • mimetype: Mimetype of the file.
    • categories: Optional list of categories to assign to the file.
    • restrict: Boolean to indicate that this is a restricted file. Defaults to False.

In the following example, we upload three files to a Dataverse instance. The first file is uploaded to the root directory of the dataset, while the other two files are uploaded to the directory some/dir.

# config.yml
persistent_id: doi:10.70122/XXX/XXXXX
dataverse_url: https://demo.dataverse.org/
api_token: XXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX
files:
    - filepath: ./small.txt
    - filepath: ./medium.txt
      directory_label: some/dir
    - filepath: ./big.txt
      directory_label: some/dir

The config file can then be used as follows:

dvuploader --config-path config.yml

Development

To install the development dependencies, run the following command:

pip install poetry
poetry install --with test

Running tests locally

In order to test the DVUploader, you need to have a Dataverse instance running. You can start a local Dataverse instance by following these steps:

1. Start the Dataverse instance

docker compose \
    -f ./docker/docker-compose-base.yml \
    --env-file local-test.env \
    up -d

2. Set up the environment variables

export BASE_URL=http://localhost:8080
export $(grep "API_TOKEN" "dv/bootstrap.exposed.env")
export DVUPLOADER_TESTING=true

3. Run the test(s) with pytest

python -m pytest -v

Linting

This repository uses ruff to lint the code and codespell to check for spelling mistakes. You can run the linters with the following command:

python -m ruff check
python -m codespell --check-filenames