GitHub | PyPI | Issues | Changelog

gamdam is the Git-Annex Mass Downloader and Metadata-er. It takes a stream of JSON Lines describing what to download and what metadata each file has, downloads them in parallel to a git-annex repository, attaches the metadata using git-annex's metadata facilities, and commits the results.

This program was written as an experiment/proof-of-concept for a larger program and is no longer maintained. However, the author has also produced a Rust translation of this program at <https://github.com/jwodder/gamdam-rust> which is currently being maintained.

Installation

gamdam requires Python 3.8 or higher. Just use pip for Python 3 (You have pip, right?) to install gamdam and its dependencies:

python3 -m pip install gamdam

gamdam also requires git-annex v10.20220222 or higher to be installed separately in order to run.

Usage

gamdam [<options>] [<input-file>]

gamdam reads a series of JSON entries from a file (or from standard input if no file is specified) following the input format described below. It feeds the URLs and output paths to git-annex addurl, and once each file has finished downloading, it attaches any listed metadata and extra URLs using git-annex metadata and git-annex registerurl, respectively.

Note that the latter step can only be performed on files tracked by git-annex; if you, say, have configured git-annex to not track text files, then any text files downloaded will not have any metadata or alternative URLs registered.

Options

--addurl-opts OPTIONS Extra options to pass to the git-annex addurl command. Note that multiple options &

arguments need to be quoted as a single string, which must also use proper shell quoting internally; e.g., --addurl-opts="--user-agent 'gamdam via git-annex'".

-C DIR, --chdir DIR The directory in which to download files;

defaults to the current directory. If the directory does not exist, it will be created. If the directory does not belong to a Git or git-annex repository, it will be initialized as one.

-F FILE, --failures FILE If any files fail to download, write their

input records back out to FILE

-J INT, --jobs INT Number of parallel jobs for git-annex addurl to use; by default, the process is

instructed to use one job per CPU core.

-l LEVEL, --log-level LEVEL Set the log level to the given value. Possible

values are "CRITICAL", "ERROR", "WARNING", "INFO", "DEBUG" (all case-insensitive) and their Python integer equivalents. [default: INFO]

-m TEXT, --message TEXT The commit message to use when saving. This

may contain a {downloaded} placeholder which will be replaced with the number of files successfully downloaded.

--no-save-on-fail Don't commit the downloaded files if any files

failed to download

--save, --no-save Whether to commit the downloaded files once

they've all been downloaded [default: --save]

Input Format

Input is a series of JSON objects, one per line (a.k.a. "JSON Lines"). Each object has the following fields:

url

(required) A URL to download

path

(required) A relative path where the contents of the URL should be saved. If an entry with a given path is encountered while another entry with the same path is being downloaded, the later entry is discarded, and a warning is emitted.

If a file already exists at a given path, git-annex will try to register the URL as an additional location for the file, failing if the resource at the URL is not the same size as the extant file.

metadata

A collection of metadata in the form used by git-annex metadata, i.e., a dict mapping key names to lists of string values.

extra_urls

A list of alternative URLs for the resource, to be attached to the downloaded file with git-annex registerurl.

If a given input line is invalid, it is discarded, and an error message is emitted.

Library Usage

gamdam can also be used as a Python library. It exports the following:

async def download(
    repo: pathlib.Path,
    objects: AsyncIterator[Downloadable],
    jobs: Optional[int] = None,
    addurl_opts: Optional[List[str]] = None,
    subscriber: Optional[anyio.abc.ObjectSendStream[DownloadResult]] = None,
) -> Report

Download the items yielded by the async iterator objects to the directory repo (which must be part of a git-annex repository) and set their metadata. jobs is the number of parallel jobs for the git-annex addurl process to use; a value of None means to use one job per CPU core. addurl_opts contains any additional arguments to append to the git-annex addurl command.

If subscriber is supplied, it will be sent a DownloadResult (see below) for each completed download, both successful and failed. This can be used to implement custom post-processing of downloads.

class Downloadable(pydantic.BaseModel):
    path: pathlib.Path
    url: pydantic.AnyHttpUrl
    metadata: Optional[Dict[str, List[str]]] = None
    extra_urls: Optional[List[pydantic.AnyHttpUrl]] = None

Downloadable is a pydantic model used to represent files to download; see Input Format above for the meanings of the fields.

class DownloadResult(pydantic.BaseModel):
    downloadable: Downloadable
    success: bool
    key: Optional[str] = None
    error_messages: Optional[List[str]] = None

DownloadResult is a pydantic model used to represent a completed download. It contains the original Downloadable, a flag to indicate download success, the downloaded file's git-annex key (only set if the download was successful and the file is tracked by git-annex) and any error messages from the addurl process (only set if the download failed).

@dataclass
class Report:
    downloaded: int
    failed: int

Report is used as the return value of download(); it contains the number of files successfully downloaded and the number of failed downloads.