Data Versioning with DVC

# Track a single file
dvc add data/train.csv

# This creates:
# data/train.csv.dvc  - Pointer file (commit this to Git)
# .gitignore update   - data/train.csv added to .gitignore

# Track a directory
dvc add data/images/

# Commit the pointer files to Git
git add data/train.csv.dvc data/.gitignore
git commit -m "Track training data with DVC"

# Push data to remote storage
dvc push

outs:
- md5: a1b2c3d4e5f6a1b2c3d4e5f6
  size: 52428800
  hash: md5
  path: train.csv

# This lightweight file is what Git tracks.
# The actual data (52 MB) is stored in DVC cache
# and pushed to remote storage.

# Push all tracked data to remote storage
dvc push

# Pull all tracked data from remote storage
dvc pull

# Push/pull specific files
dvc push data/train.csv.dvc
dvc pull data/train.csv.dvc

# Fetch without checkout (download to cache only)
dvc fetch

# Checkout from local cache (no download)
dvc checkout

# Tag the current version
git tag data-v1

# Update the data
# ... modify data/train.csv ...

# Track the new version
dvc add data/train.csv
git add data/train.csv.dvc
git commit -m "Update training data v2"
git tag data-v2
dvc push

# Switch back to v1
git checkout data-v1 -- data/train.csv.dvc
dvc checkout data/train.csv.dvc

# Switch to v2
git checkout data-v2 -- data/train.csv.dvc
dvc checkout data/train.csv.dvc

import dvc.api

# Get URL of a tracked file (for cloud-native access)
url = dvc.api.get_url(
    path='data/train.csv',
    repo='https://github.com/user/project'
)

# Read file contents directly
with dvc.api.open(
    'data/train.csv',
    repo='https://github.com/user/project',
    rev='data-v1'   # specific Git tag/branch/commit
) as f:
    import pandas as pd
    df = pd.read_csv(f)

# Get parameters
params = dvc.api.params_show()
print(params['train']['learning_rate'])