DataLad NEXT extension
This DataLad extension can be thought of as a staging area for additional functionality, or for improved performance and user experience. Unlike other topical or more experimental extensions, the focus here is on functionality with broad applicability. This extension is a suitable dependency for other software packages that intend to build on this improved set of functionality.
Installation
# create and enter a new virtual environment (optional)
$ virtualenv --python=python3 ~/env/dl-next
$ . ~/env/dl-next/bin/activate
# install from PyPi
$ python -m pip install datalad-nextHow to use
Additional commands provided by this extension are immediately available after installation. However, in order to fully benefit from all improvements, the extension has to be enabled for auto-loading by executing:
git config --global --add datalad.extensions.load nextDoing so will enable the extension to also alter the behavior the core DataLad package and its commands.
Summary of functionality provided by this extension
- A replacement sub-system for credential handling that is able to handle arbitrary properties for annotating a secret, and facilitates determining suitable credentials while minimizing avoidable user interaction, without compromising configurability. A convenience method is provided that implements a standard workflow for obtaining a credential.
- A user-facing
credentialscommand to set, remove, and query credentials. - The
create-sibling-...commands for the platforms GitHub, GIN, GOGS, Gitea are equipped with improved credential handling that, for example, only stores entered credentials after they were confirmed to work, or auto-selects the most recently used, matching credentials, when none are specified. - A
create-sibling-webdavcommand for hosting datasets on a WebDAV server via a sibling tandem for Git history and file storage. Datasets hosted on WebDAV in this fashion are cloneable withdatalad-clone. A full annex setup for storing complete datasets with historical file content version, and an additional mode for depositing single-version dataset snapshot are supported. The latter enables convenient collaboration with audiences that are not using DataLad, because all files are browsable via a WebDAV server's point-and-click user interface. - Enhance
datalad-pushto automatically export files to git-annex special remotes configured withexporttree=yes. - Speed-up
datalad-pushwhen processing non-git special remotes. This particularly benefits less efficient hosting scenarios like WebDAV. - Enhance
datalad-siblings enable(AnnexRepo.enable_remote()) to automatically deploy credentials for git-annex special remotes that require them. git-remote-datalad-annexis a Git remote helper to push/fetch to any location accessible by any git-annex special remote.git-annex-backend-XDLRA(originally available from themihextrasextension) is a custom external git-annex backend used bygit-remote-datalad-annex. A base class to facilitate development of external backends in Python is also provided.- Enhance
datalad-configurationto support getting configuration from "global" scope without a dataset being present. - New modular framework for URL operations. This framework directly supports operation
on
http(s),ssh, andfileURLs, and can be extended with custom functionality for additional protocols or even interaction with specific individual servers. The basic operationsdownload,upload,delete, andstatare recognized, and can be implemented. The framework offers uniform progress reporting and simultaneous content has computation. This framework is meant to replace and extend the downloader/provide framework in the DataLad core package. In contrast to its predecessor it is integrated with the new credential framework, and operations beyond downloading. git-annex-remote-uncurlis a special remote that exposes the new URL operations framework via git-annex. It provides flexible means to compose and rewrite URLs (e.g., to compensate for storage infrastructure changes) without having to modify individual URLs recorded in datasets. It enables seamless transitions between any services and protocols supported by the framework. This special remote can replace thedataladspecial remote provided by the DataLad core package.- A
downloadcommand is provided as a front-end for the new modular URL operations framework. - A
python-requestscompatible authentication handler (DataladAuth) that interfaces DataLad's credential system. - Boosted throughput of DataLad's
runnercomponent for command execution. - Substantially more comprehensive replacement for DataLad's
constraintssystem for type conversion and parameter validation. - Windows and Mac client support for RIA store access.
- A
next-statuscommand that is A LOT faster thanstatus, and offers amonorecursion mode that shows modifications of nested dataset hierarchies relative to the state of the root dataset. Requires Git v2.31 (or later).
Summary of additional features for DataLad extension development
- Framework for uniform command parameter validation. Regardless of the used API (Python, CLI, or GUI), command parameters are uniformly validated. This facilitates a stricter separation of parameter specification (and validation) from the actual implementation of a command. The latter can now focus on a command's logic only, while the former enables more uniform and more comprehensive validation and error reporting. Beyond per-parameter validation and type-conversion also inter-parameter dependency validation and value transformations are supported.
- Improved composition of importable functionality. Key components for
commands,annexremotes,datasets(etc) are collected in topical top-level modules that provide "all" necessary pieces in a single place. webdav_serverfixture that automatically deploys a local WebDAV server.- Utilities for HTTP handling
probe_url()discovers redirects and authentication requirements for an HTTP URLget_auth_realm()returns a label for an authentication realm that can be used to query for matching credentials
- Utilities for special remote credential management:
get_specialremote_credential_properties()inspects a special remote and returns properties for querying a credential store for matching credentialsupdate_specialremote_credential()updates a credential in a store after successful useget_specialremote_credential_envpatch()returns a suitable environment "patch" from a credential for a particular special remote type
- Helper for runtime-patching other datalad code (
datalad_next.utils.patch) - Base class for implementing custom
git-annexbackends. - A set of
pytestfixtures to:- check that no global configuration side-effects are left behind by a test
- check that no secrets are left behind by a test
- provide a temporary configuration that is isolated from a user environment and from other tests
- provide a temporary secret store that is isolated from a user environment and from other tests
- provide a temporary credential manager to perform credential deployment and manipulation isolated from a user environment and from other tests
- An
iter_subproc()helper that enable communication with subprocesses via input/output iterables. - A
shellcontext manager that enables interaction with (remote) shells, including support for input/output iterables for each shell-command execution within the context.
Patching the DataLad core package.
Some of the features described above rely on a modification of the DataLad core
package itself, rather than coming in the form of additional commands. Loading
this extension causes a range of patches to be applied to the datalad package
to enable them. A comprehensive description of the current set of patch is
available at http://docs.datalad.org/projects/next/en/latest/#datalad-patches
Developing with DataLad NEXT
This extension package moves fast in comparison to the core package. Nevertheless, attention is paid to API stability, adequate semantic versioning, and informative changelogs.
Public vs internal API
Anything that can be imported directly from any of the sub-packages in
datalad_next is considered to be part of the public API. Changes to this API
determine the versioning, and development is done with the aim to keep this API
as stable as possible. This includes signatures and return value behavior.
As an example: from datalad_next.runners import iter_git_subproc imports a
part of the public API, but from datalad_next.runners.git import iter_git_subproc does not.
Use of the internal API
Developers can obviously use parts of the non-public API. However, this should only be done with the understanding that these components may change from one release to another, with no guarantee of transition periods, deprecation warnings, etc.
Developers are advised to never reuse any components with names starting with
_ (underscore). Their use should be limited to their individual subpackage.
Acknowledgements
This DataLad extension was developed with funding from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under grant SFB 1451 (431549029, INF project).
