Implementation:Iterative Dvc Dependency Repo

Knowledge Sources	Iterative_Dvc
Domains	Dependency_Management, Cross_Repository
Last Updated	2026-02-10 10:00 GMT

Overview

RepoDependency is a class defined in dvc/dependency/repo.py (168 lines) that extends the base Dependency class. It represents a dependency on a file or directory in an external DVC repository. It handles filesystem resolution through DVCFileSystem, revision locking, and cross-repo file downloads with hash caching.

from dvc.dependency.repo import RepoDependency

Source File

Property	Value
File	`dvc/dependency/repo.py`
Lines	168
Class	`RepoDependency`
Extends	`Dependency` (from `dvc.dependency.base`)

Class: RepoDependency

RepoDependency models a stage dependency that points to a path in another Git/DVC repository. The remote repo is specified via a URL, optional revision, and optional config/remote overrides.

Class Attributes

Attribute	Value	Description
`PARAM_REPO`	`"repo"`	Top-level key in serialized representation
`PARAM_URL`	`"url"`	Remote repository URL
`PARAM_REV`	`"rev"`	Requested Git revision (branch, tag, or commit)
`PARAM_REV_LOCK`	`"rev_lock"`	Locked (pinned) Git revision
`PARAM_CONFIG`	`"config"`	Optional DVC config override (path or dict)
`PARAM_REMOTE`	`"remote"`	Optional remote storage override

The REPO_SCHEMA class variable defines a voluptuous validation schema requiring the url field and accepting optional rev, rev_lock, config, and remote fields.

Constructor

def __init__(self, def_repo: dict[str, Any], stage: "Stage", *args, **kwargs)

Stores the repository definition dict, calls the parent constructor, and immediately constructs a DVCFileSystem via _make_fs(). The filesystem path is normalized to a POSIX-style path.

Properties

is_in_repo

Always returns False since the dependency refers to an external repository.

Methods

workspace_status

def workspace_status(self) -> dict

Compares the locked revision (from _make_fs(locked=True)) with the latest revision (from _make_fs(locked=False)). Returns {str(self): "update available"} if they differ, or an empty dict if the dependency is up to date.

save

def save(self) -> None

Records the current Git revision into self.def_repo[PARAM_REV_LOCK] if it has not been set yet. This pins the dependency to a specific commit.

dumpd

def dumpd(self, **kwargs) -> dict[str, Union[str, dict[str, str]]]

Serializes the dependency to a dictionary containing path and repo keys. The repo sub-dict is built by _dump_def_repo(), which includes only non-empty fields (url, rev, rev_lock, config, remote).

download

def download(self, to: "Output", jobs: Optional[int] = None)

Downloads the dependency content to the target output by:

Calling super().download() to perform the file transfer
If the target filesystem is local, iterating over downloaded files to extract DVC hash info
Saving the hash information to the output cache state, avoiding re-hashing of already known files

update

def update(self, rev: Optional[str] = None) -> None

Updates the dependency to track a new revision. If a rev is specified, it updates the requested revision. Then it re-creates the filesystem with locked=False and records the resolved revision as the new rev_lock.

changed_checksum

def changed_checksum(self) -> bool

Always returns False. From the current repo's perspective, a RepoDependency is described by its URL and rev_lock, making it effectively immutable once locked.

_make_fs (private)

def _make_fs(self, rev: Optional[str] = None, locked: bool = True) -> "DVCFileSystem"

Constructs a DVCFileSystem instance for the remote repository. Key logic includes:

Handling remote as either a string name or a dict of config values
Loading config from a file path or using a dict directly
Injecting the local repo's cache configuration into the remote filesystem config to avoid re-streaming data
Passing subrepos=True to support monorepo structures

_get_rev (private)

def _get_rev(self, locked: bool = True) -> Optional[str]

Returns the locked revision when locked=True, falling back to the requested revision. When locked=False, returns only the requested revision.

Key Dependencies

Module	Usage
`dvc.dependency.base.Dependency`	Base dependency class
`dvc.fs.DVCFileSystem`	Filesystem abstraction for remote DVC repos
`dvc.config.Config`	Loading config files for remote repos
`dvc.fs.LocalFileSystem`	Checking if download target is local
`voluptuous`	Schema validation for repo definition
`dvc.utils.as_posix`	Path normalization

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment