IGA

IGA is the InvenioRDM GitHub Archiver, a standalone program as well as a GitHub Actions workflow that lets you automatically archive GitHub software releases in an InvenioRDM repository.

Table of contents

• Introduction
• Installation
  • IGA as a standalone program
  • IGA as a GitHub Actions workflow
• Quick start
• Usage
  • Identifying the InvenioRDM server
  • Providing an InvenioRDM access token
  • Providing a GitHub access token
  • Specifying a GitHub release
  • Gathering metadata for an InvenioRDM record
  • Specifying GitHub file assets
  • Handling communities
  • Indicating draft versus published records
  • Versioning records
  • Other options recognized by IGA
  • Summary of command-line options
  • Return values
  • Adding a DOI badge to your GitHub repository
• Known issues and limitations
• Getting help
• Contributing
• License
• Acknowledgments

Introduction

InvenioRDM is the basis for many institutional repositories such as CaltechDATA that enable users to preserve software and data sets in long-term archive. Though such repositories are critical resources, creating detailed records and uploading assets can be a tedious and error-prone process if done manually. This is where the InvenioRDM GitHub Archiver (IGA) comes in.

IGA creates metadata records and sends releases from GitHub to an InvenioRDM-based repository server. IGA can be invoked from the command line; it also can be set up as a GitHub Actions workflow to archive GitHub releases automatically for a repository each time they are made.

➜

IGA offers many notable features:

• Automatic metadata extraction from GitHub plus codemeta.json and CITATION.cff files
• Thorough coverage of InvenioRDM record metadata using painstaking procedures
• Recognition of identifiers in CodeMeta & CFF files: ORCID, DOI, PMCID, and more
• Automatic lookup of publication data in DOI.org, PubMed, Google, and other sources
• Automatic lookup of organization names in ROR (assuming ROR id's are provided)
• Automatic lookup of human names in ORCID.org (assuming ORCID id's are provided)
• Automatic splitting of human names into family & given names using ML methods
• Support for InvenioRDM communities
• Support for overriding the record that IGA creates, for complete control if you need it
• Support for using the GitHub API without a GitHub access token in simple cases
• Extensive use of logging so you can see what's going on under the hood

Installation

IGA can be installed as either (or both) a command-line program on your computer or a GitHub Action in a GitHub repository.

IGA as a standalone program

Please choose an approach that suits your situation and preferences.

Once you have installed iga, the next steps are (1) get an InvenioRDM token and (2) configure iga for running locally.

IGA as a GitHub Actions workflow

A GitHub Actions workflow is an automated process that runs on GitHub's servers under control of a file in your repository. Follow these steps to create the IGA workflow file:

1. In the main branch of your GitHub repository, create a .github/workflows directory

2. In the .github/workflows directory, create a file named (e.g.) iga.yml and copy the following contents into it:

   # ╭────────────────────────────────────────────╮
   # │ Configure this section                     │
   # ╰────────────────────────────────────────────╯
   
   env:
     # 👋🏻 Set the next variable to your InvenioRDM server address 👋🏻
     INVENIO_SERVER: https://your-invenio-server.org
   
     # Set to an InvenioRDM record ID to mark release as a new version.
     parent_record: none
   
     # The variables below are other IGA options. Please see the docs.
     community:     none
     draft:         false
     all_assets:    false
     all_metadata:  false
     debug:         false
   
   # ╭────────────────────────────────────────────╮
   # │ The rest of this file should be left as-is │
   # ╰────────────────────────────────────────────╯
   
   name: InvenioRDM GitHub Archiver
   on:
     release:
       types: [published]
     workflow_dispatch:
       inputs:
         release_tag:
           description: The release tag (empty = latest)
         parent_record:
           description: ID of parent record (for versioning)
         community:
           description: Name of InvenioRDM community (if any)
         draft:
           description: Mark the record as a draft
           type: boolean
         all_assets:
           description: Attach all GitHub assets
           type: boolean
         all_metadata:
           description: Include additional GitHub metadata
           type: boolean
         debug:
           description: Print debug info in the GitHub log
           type: boolean
   
   run-name: Archive ${{inputs.release_tag || 'latest release'}} in InvenioRDM
   jobs:
     run_iga:
       name: Send to ${{needs.get_repository.outputs.server}}
       runs-on: ubuntu-latest
       needs: get_repository
       steps:
         - uses: caltechlibrary/iga@v1
           with:
             INVENIO_SERVER: ${{env.INVENIO_SERVER}}
             INVENIO_TOKEN:  ${{secrets.INVENIO_TOKEN}}
             all_assets:     ${{github.event.inputs.all_assets || env.all_assets}}
             all_metadata:   ${{github.event.inputs.all_metadata || env.all_metadata}}
             debug:          ${{github.event.inputs.debug || env.debug}}
             draft:          ${{github.event.inputs.draft || env.draft}}
             community:      ${{github.event.inputs.community || env.community}}
             parent_record:  ${{github.event.inputs.parent_record || env.parent_record}}
             release_tag:    ${{github.event.inputs.release_tag || 'latest'}}
     get_repository:
       name: Get repository name
       runs-on: ubuntu-latest
       outputs:
         server: ${{steps.parse.outputs.host}}
       steps:
         - name: Extract name from INVENIO_SERVER
           id: parse
           run: echo "host=$(cut -d'/' -f3 <<< ${{env.INVENIO_SERVER}} | cut -d':' -f1)" >> $GITHUB_OUTPUT

3. Edit the value of the INVENIO_SERVER variable (line 7 above) ↑

4. Optionally, change the values of other options (parent_record, community, etc.)

5. Save the file, commit the changes to git, and push your changes to GitHub

Once you have installed the GitHub Action workflow for IGA, the next steps are (1) get an InvenioRDM token and (2) configure the GitHub Action workflow.

Quick start

No matter whether IGA is run locally on your computer or as a GitHub Actions workflow, in both cases it must be provided with a personal access token (PAT) for your InvenioRDM server. Getting one is the first step.

Getting an InvenioRDM token

1. Log in to your InvenioRDM account
2. Go to the Applications page in your account profile
3. Click the New token button next to "Personal access tokens"
4. On the page that is shown after you click that button, name your token (the name does not matter) and click the Create button
5. After InvenioRDM creates and shows you the token, copy it to a safe location because InvenioRDM will not show it again

Configuring and running IGA locally

To send a GitHub release to your InvenioRDM server, IGA needs this information:

1. (Required) The identity of the GitHub release to be archived
2. (Required) The address of the destination InvenioRDM server
3. (Required) A personal access token for InvenioRDM (from above)
4. (Optional) A personal access token for GitHub

The identity of the GitHub release is always given as an argument to IGA on the command line; the remaining values can be provided either via command-line options or environment variables. One approach is to set environment variables in shell scripts or your interactive shell. Here is an example using Bash shell syntax, with fake token values:

export INVENIO_SERVER=https://data.caltech.edu
export INVENIO_TOKEN=qKLoOH0KYf4D98PGYQGnC09hiuqw3Y1SZllYnonRVzGJbWz2
export GITHUB_TOKEN=ghp_wQXp6sy3AsKyyEo4l9esHNxOdo6T34Zsthz

Once these are set, use of IGA can be as simple as providing a URL for a release in GitHub. For example, the following command creates a draft record (the -d option is short for --draft) for another project in GitHub and tells IGA to open (the -o option is short for --open) the newly-created InvenioRDM entry in a web browser:

iga -d -o https://github.com/mhucka/taupe/releases/tag/v1.2.0

More options are described in the section on detailed usage information below.

Configuring and running IGA as a GitHub Actions workflow

After doing the GitHub Actions installation steps and obtaining an InvenioRDM token, one more step is needed: the token must be stored as a "secret" in your GitHub repository.

1. Go to the Settings page of your GitHub repository

   

2. In the left-hand sidebar, find Secrets and variables in the Security section, click on it to reveal Actions underneath, then click on Actions

   

3. In the next page, click the green New repository secret button

   

4. Name the variable INVENIO_TOKEN and paste in your InvenioRDM token
5. Finish by clicking the green Add secret button

Testing the workflow

After setting up the workflow and storing the InvenioRDM token in your repository on GitHub, it's a good idea to run the workflow manually to test that it works as expected.

1. Go to the Actions tab in your repository and click on the name of the workflow in the sidebar on the left

   

2. Click the Run workflow button in the right-hand side of the blue strip
3. In the pull-down, change the value of "Mark the record as a draft" to true

   

4. Click the green Run workflow button near the bottom
5. Refresh the web page and a new line will be shown named after your workflow file

   

6. Click the title of the workflow to see the IGA workflow progress and results

Running the workflow when releasing software

Once the personal access token from InvenioRDM is stored as a GitHub secret, the workflow should run automatically every time a new release is made on GitHub – no further action should be needed. You can check the results (and look for errors if something went wrong) by going to the Actions tab in your GitHub repository.

Usage

This section provides detailed information about IGA's operation and options to control it.

Identifying the InvenioRDM server

The server address must be provided either as the value of the option --invenio-server or in an environment variable named INVENIO_SERVER. If the server address does not begin with https://, IGA will prepend it automatically.

Providing an InvenioRDM access token

A personal access token (PAT) for making API calls to the InvenioRDM server must be also supplied when invoking IGA. The preferred method is to set the value of the environment variable INVENIO_TOKEN. Alternatively, you can use the option --invenio-token to pass the token on the command line, but you are strongly advised to avoid this practice because it is insecure.

To obtain a PAT from an InvenioRDM server, first log in to the server, then visit the page at /account/settings/applications and use the interface there to create a token. The token will be a long string of alphanumeric characters such as OH0KYf4PGYQGnCM4b53ejSGicOC4s4YnonRVzGJbWxY; set the value of the variable INVENIO_TOKEN to this string.

Providing a GitHub access token

It may be possible to run IGA without providing a GitHub access token. GitHub allows up to 60 API calls per minute when running without credentials, and though IGA makes several API calls to GitHub each time it runs, for some public repositories IGA will not hit the limit. However, if you are archiving a private repository, run IGA multiple times in a row, or the repository has many contributors, then you will need to supply a GitHub access token. The preferred way of doing that is to set the value of the environment variable GITHUB_TOKEN. Alternatively, you can use the option --github-token to pass the token on the command line, but you are strongly advised to avoid this practice because it is insecure. To obtain a PAT from GitHub, visit https://docs.github.com/en/authentication and follow the instructions for creating a "classic" personal access token.

Note that when you run IGA as a GitHub Actions workflow, you do not need to create or set a GitHub token because it is obtained automatically by the GitHub Actions workflow.

Specifying a GitHub release

A GitHub release can be specified to IGA in one of two mutually-exclusive ways:

1. The full URL of the web page on GitHub of a tagged release. In this case, the URL must be the final argument on the command line invocation of IGA and the options --account and --repo must be omitted.
2. A combination of account name, repository name, and tag. In this case, the final argument on the command line must be the tag, and in addition, values for the options --account and --repo must be provided.

Here's an example using approach #1 (assuming environment variables INVENIO_SERVER, INVENIO_TOKEN, and GITHUB_TOKEN have all been set):

iga https://github.com/mhucka/taupe/releases/tag/v1.2.0

and here's the equivalent using approach #2:

iga --github-account mhucka --github-repo taupe v1.2.0

Note that when using this form of the command, the release tag (v1.2.0 above) must be the last item given on the command line.

Gathering metadata for an InvenioRDM record

The record created in InvenioRDM is constructed using information obtained using GitHub's API as well as several other APIs as needed. The information includes the following:

• (if one exists) a codemeta.json file in the GitHub repository
• (if one exists) a CITATION.cff file in the GitHub repository
• data available from GitHub for the release
• data available from GitHub for the repository
• data available from GitHub for the account of the owner
• data available from GitHub for the accounts of repository contributors
• file assets associated with the GitHub release
• data available from ORCID.org for ORCID identifiers
• data available from ROR.org for Research Organization Registry identifiers
• data available from DOI.org, NCBI, Google Books, & others for publications
• data available from spdx.org for software licenses

IGA tries to use CodeMeta.json first and CITATION.cff second to fill out the fields of the InvenioRDM record. If neither of those files are present, IGA uses values from the GitHub repository instead. You can make it always use all sources of info with the option --all-metadata. Depending on how complete and up-to-date your CodeMeta.json and CITATION.cff are, this may or may not make the record more comprehensive and may or may not introduce redundancies or unwanted values.

To override the auto-created metadata, use the option --read-metadata followed by the path to a JSON file structured according to the InvenioRDM schema used by the destination server. When --read-metadata is provided, IGA does not extract the data above, but still obtains the file assets from GitHub.

Specifying GitHub file assets

By default, IGA attaches to the InvenioRDM record only the ZIP file asset created by GitHub for the release. To make IGA attach all assets associated with the GitHub release, use the option --all-assets.

To upload specific file assets and override the default selections made by IGA, you can use the option --file followed by a path to a file to be uploaded. You can repeat the option --file to upload multiple file assets. Note that if --file is provided, then IGA does not use any file assets from GitHub; it is the user's responsibility to supply all the files that should be uploaded.

If both --read-metadata and --file are used, then IGA does not actually contact GitHub for any information.

Handling communities

To submit your record to a community, use the --community option together with a community name. The option --list-communities can be used to get a list of communities supported by the InvenioRDM server. Note that submitting a record to a community means that the record will not be finalized and will not be publicly visible when IGA finishes; instead, the record URL that you receive will be for a draft version, pending review by the community moderators.

Indicating draft versus published records

If the --community option is not used, then by default, IGA will finalize and publish the record. To make it stop short and leave the record as a draft instead, use the option --draft. The draft option also takes precedence over the community option: if you use both --draft and --community, IGA will stop after creating the draft record and will not submit it to the community. (You can nevertheless submit the record to a community manually once the draft is created, by visiting the record's web page and using the InvenioRDM interface there.)

Versioning records

The option --parent-record can be used to indicate that the record being constructed is a new version of an existing record. This will make IGA use the InvenioRDM API for record versioning. The newly-created record will be linked to a parent record identified by the value passed to --parent-record. The value must be either an InvenioRDM record identifier (which is a sequence of alphanumeric characters of the form XXXXX-XXXXX, such as bknz4-bch35, generated by the InvenioRDM server), or a URL to the landing page of the record in the InvenioRDM server. (Note that such URLs end in the record identifier.) Here is an example of using this option:

iga --parent-record xbcd4-efgh5 https://github.com/mhucka/taupe/releases/tag/v1.2.0

Other options recognized by IGA

Running IGA with the option --save-metadata will make it create a metadata record, but instead of uploading the record (and any assets) to the InvenioRDM server, IGA will write the result to the given destination. This can be useful not only for debugging but also for creating a starting point for a custom metadata record: first run IGA with --save-metadata to save a record to a file, edit the result, then finally run IGA with the --read-metadata option to use the modified record to create a release in the InvenioRDM server.

The --mode option can be used to change the run mode. Four run modes are available: quiet, normal, verbose, and debug. The default mode is normal, in which IGA prints a few messages while it's working. The mode quiet will make it avoid printing anything unless an error occurs, the mode verbose will make it print a detailed trace of what it is doing, and the mode debug will make IGA even more verbose. In addition, in debug mode, IGA will drop into the pdb debugger if it encounters an exception during execution. On Linux and macOS, debug mode also installs a signal handler on signal USR1 that causes IGA to drop into the pdb debugger if the signal USR1 is received. (Use kill -USR1 NNN, where NNN is the IGA process id.)

By default, informational output is sent to the standard output (normally the terminal console). The option --log-dest can be used to send the output to the given destination instead. The value can be - (i.e., a dash) to indicate console output, or it can be a file path to send the output to the file. A special exception is that even if a log destination is given, IGA will still print the final record URL to stdout. This makes it possible to invoke IGA from scripts that capture the record URL while still saving diagnostic output in case debugging is needed.

By default, IGA prints only the record URL when done. The option --print-doi will make it also print the DOI of the record. (Note that this only works when publishing records; if options --draft or --community are used, then there will be no DOI. In those case, only the URL will be printed.)

Reading and writing large files may take a long time; on the other hand, IGA should not wait forever on network operations before reporting an error if a server or network becomes unresponsive. To balance these conflicting needs, IGA automatically scales its network timeout based on file sizes. To override its adaptive algorithm and set an explicit timeout value, use the option --timeout with a value in seconds.

If given the --version option, this program will print its version and other information, and exit without doing anything else.

Running IGA with the option --help will make it print help text and exit without doing anything else.

Summary of command-line options

As explain above, IGA takes one required argument on the command line: either (1) the full URL of a web page on GitHub of a tagged release, or (2) a release tag name which is to be used in combination with options --github-account and --github-repo. The following table summarizes all the command line options available.

⚑   Can repeat the option to specify multiple files.
⚐   To write to the console, use the character - as the value of OUT; otherwise, OUT must be the name of a file where the output should be written.
✯   When using --github-account and --github-repo, the last argument on the command line must be a release tag name.
❖   The record identifier must be given either as a sequence of alphanumeric characters of the form XXXXX-XXXXX (e.g., bknz4-bch35), or as a URL to the landing page of an existing record in the InvenioRDM server.

Return values

This program exits with a return status code of 0 if no problem is encountered. Otherwise, it returns a nonzero status code. The following table lists the possible values:

Adding a DOI badge to your GitHub repository

Once you have set up the IGA workflow in your GitHub repository, you may wish to add a DOI badge to your repository's README file. It would be a chore to keep updating the DOI value in this badge every time a new release is made, and thankfully, it's not necessary: it's possible to make the badge get the current DOI value dynamically. Here is how:

1. After you have at least one release archived in your InvenioRDM server, find out the DOI of that release in InvenioRDM, and extract the tail end of that DOI. The DOI assigned by InvenioRDM will be a string such as 10.22002/zsmem-2pg20; the tail end is the zsmem-2pg20 part. (Your DOI and tail portion will be different.)

2. Let SERVERURL stand for the URL for your InvenioRDM server, and let IDENTIFIER stand for the identifier portion of the DOI. In your README.md file, write the DOI badge as follows (without line breaks):

   [![DOI](https://img.shields.io/badge/dynamic/json.svg?label=DOI&query=$.pids.doi.identifier&uri=SERVERURL/api/records/IDENTIFIER/versions/latest)](SERVERURL/records/IDENTIFIER/latest)

For example, for CaltechDATA and the archived IGA releases there,

• SERVERURL = http://data.caltech.edu
• IDENTIFIER = zsmem-2pg20

which leads to the following badge string for IGA's README.md file:

[![DOI](https://img.shields.io/badge/dynamic/json.svg?label=DOI&query=$.pids.doi.identifier&uri=https://data.caltech.edu/api/records/zsmem-2pg20/versions/latest)](https://data.caltech.edu/records/zsmem-2pg20/latest)

The result looks like this:

You can change the look of the badge by using style parameters. Please refer to the Shields.io documentation for static badges.

Known issues and limitations

The following are known issues and limitations.

• As of mid-2023, InvenioRDM requires names of record creators and other contributors to be split into given (first) and family (surname). This is problematic for multiple reasons. The first is that mononyms are common in many countries: a person's name may legitimately be only a single word which is not conceptually a "given" or "family" name. To compound the difficulty for IGA, names are stored as single fields in GitHub account metadata, so unless a repository has a codemeta.json or CITATION.cff file (which allow authors more control over how they want their names represented), IGA is forced to try to split the single GitHub name string into two parts. A foolproof algorithm for doing this does not exist, so IGA will sometimes get it wrong. (That said, IGA goes to extraordinary lengths to try to do a good job.)
• InvenioRDM requires that identities (creators, contributors, etc.) to be labeled as personal or organizational. The nature of identities is usually made clear in codemeta.json and CITATION.cff files. GitHub also provides a flag that is meant to be used to label organizational accounts, but sometimes people don't set the GitHub account information correctly. Consequently, if IGA has to use GitHub data to get (e.g.) the list of contributors on a project, it may mislabel identities in the InvenioRDM record it produces.
• Some accounts on GitHub are software automation or "bot" accounts, but are not labeled as such. These accounts are generally indistinguishable from human accounts on GitHub, so if they're not labeled as bot or organizational accounts in GitHub, IGA can't recognize that they're humans. If such an account is the creator of a release in GitHub, and IGA tries to use its name-splitting algorithm on the name of the account, it may produce a nonsensical result. For example, it might turn "Travis CI" into an entry with a first name of "Travis" and last name of "CI".
• Funder and funding information can only be specified in codemeta.json files; neither GitHub nor CITATION.cff have provisions to store this kind of metadata. The CodeMeta specification defines two fields for this purpose: funder and funding. Unfortunately, these map imperfectly to the requirements of InvenioRDM's metadata format. In addition, people don't always follow the CodeMeta guidelines, and sometimes they write funding information as text strings (instead of structured objects), the interpretation of which would require software that can recognize grant and funding agency information from free-text descriptions. This combination of factors means IGA often can't fill in the funding metadata in InvenioRDM records even if there is some funding information in the codemeta.json file.

Getting help

If you find an issue, please submit it in the GitHub issue tracker for this repository.

Contributing

Your help and participation in enhancing IGA is welcome! Please visit the guidelines for contributing for some tips on getting started.

License

Software produced by the Caltech Library is Copyright © 2022–2024 California Institute of Technology. This software is freely distributed under a BSD-style license. Please see the LICENSE file for more information.

Acknowledgments

This work was funded by the California Institute of Technology Library.

IGA uses multiple other open-source packages, without which it would have taken much longer to write the software. I want to acknowledge this debt. In alphabetical order, the packages are:

• Aenum – package for advanced enumerations
• Arrow – a library for creating & manipulating dates
• Boltons – package of miscellaneous Python utilities
• caltechdata_api – package for using the CaltechDATA API
• CommonPy – a collection of commonly-useful Python functions
• dirtyjson – JSON decoder that copes with problematic JSON files and reports useful error messages
• flake8 – Python code linter and style analyzer
• Ginza – Japanese NLP Library
• httpx – HTTP client library that supports HTTP/2
• humanize – make numbers more easily readable by humans
• idutils – package for validating and normalizing various kinds of persistent identifiers
• ipdb – the IPython debugger
• iptools – utilities for dealing with IP addresses
• isbnlib – utilities for dealing with ISBNs
• Jamo – Hangul character analysis
• json5 – extended JSON format parser
• latexcodec – lexer and codec to work with LaTeX code in Python
• Lingua – language detection library
• linkify-it-py – a link recognition library with full unicode support
• lxml – an XML parsing library
• Markdown – Python package for working with Markdown
• markdown-checklist – GitHub-style checklist extension for Python Markdown package
• mdx-breakless-lists – GitHub-style Markdown lists that don't require a line break above them
• mdx_linkify – extension for Python Markdown will convert text that look like links to HTML anchors
• MyST-parser – A Sphinx and Docutils extension to parse an extended version of Markdown
• nameparser – package for parsing human names into their individual components
• probablepeople – package for parsing names into components using ML-based techniques
• pybtex – BibTeX parser and formatter
• pybtex-apa7-style – plugin for pybtex that provides APA7 style formatting
• pymdown-extensions – extensions for Python Markdown
• pytest – testing framework
• pytest-cov – coverage reports for use with pytest
• pytest-mock – wrapper around the mock package for use with pytest
• PyYAML – YAML parser
• Rich – library for writing styled text to the terminal
• rich-click – CLI interface built on top of Rich
• setuptools – library for setup.py
• Sidetrack – simple debug logging/tracing package
• spaCy – Natural Language Processing package
• spacy-alignments – alternate alignments for spaCy
• spacy-legacy – spaCy legacy functions and architectures for backwards compatibility
• spacy-loggers – loggers for spaCy
• spacy-pkuseg – Chinese word segmentation toolkit for spaCy
• spacy-transformers – pretrained Transformers for spaCy
• Sphinx – documentation generator for Python
• sphinx-autobuild – rebuild Sphinx docs automatically
• sphinx-material – a responsive Material Design theme for Sphinx
• sphinxcontrib-mermaid – support Mermaid diagrams in Sphinx docs
• StringDist – library for calculating string distances
• Twine – utilities for publishing Python packages on PyPI
• url-normalize – URI/URL normalization utilities
• validators – data validation package for Python
• wheel – setuptools extension for building wheels