Skip to content

Files

Latest commit

 

History

History
159 lines (120 loc) · 5.59 KB

version_control_in_datascience.md

File metadata and controls

159 lines (120 loc) · 5.59 KB

Version Control in Data Science

Use case 1: When working with multiple features (Version Control Git Branches V1 V2)

  • STEP 1: You have a local version of this repository on your laptop, and to get the latest stable version, you pull from the develop branch.

    Switch to the develop branch

    git checkout develop

    Pull latest changes in the develop branch

    git pull

  • STEP 2: When you start working on this demographic feature (new feature), you create a new branch for this called demographic, and start working on your code in this branch.

    Create and switch to new branch called demographic from develop branch

    git checkout -b demographic

    Work on this new feature and commit as you go

    git commit -m 'added gender recommendations'
git commit -m 'added location specific recommendations'

  • STEP 3: However, in the middle of your work, you need to work on another feature. So you commit your changes on this demographic branch, and switch back to the develop branch.

    Commit changes before switching

    git commit -m 'refactored demographic gender and location recommendations '

    Switch to the develop branch

    git checkout develop

  • STEP 4: From this stable develop branch, you create another branch for a new feature called friend_groups.

    Create and switch to new branch called friend_groups from develop branch

    git checkout -b friend_groups

  • STEP 5: After you finish your work on the friend_groups branch, you commit your changes, switch back to the development branch, merge it back to the develop branch, and push this to the remote repository’s develop branch.

    Commit changes before switching

    git commit -m 'finalized friend_groups recommendations '

    Switch to the develop branch

    git checkout develop

    Merge friend_groups branch to develop

    git merge --no-ff friends_groups

    Push to remote repository

    git push origin develop

  • STEP 6: Now, you can switch back to the demographic branch to continue your progress on that feature.

    Switch to the demographic branch

    git checkout demographic

Use case 2: Usage of github for tracking Hyperparameter tuning (Version Control Git Commit Messages V1 V2)

  • Step 1: You check your commit history, seeing messages of the changes you made and how well it performed.

    View log history

    git log

  • Step 2: The model at this commit seemed to score the highest, so you decide to take a look.

    Checkout a commit

    git checkout bc90f2cbc9dc4e802b46e7a153aa106dc9a88560

    After inspecting your code, you realize what modifications made this perform well, and use those for your model.

  • Step 3: Now, you’re pretty confident merging this back into the development branch, and pushing the updated recommendation engine.

    Switch to develop branch

    git checkout develop

    Merge friend_groups branch to develop

    git merge --no-ff friend_groups

    Push changes to remote repository

    git push origin develop

Use case 3: Working with team (Version Control Merging Branches On A Team V1 V2)

  • Step 1: Andrew commits his changes to the documentation branch, switches to the development branch, and pulls down the latest changes from the cloud on this development branch, including the change I merged previously for the friends group feature.

    Commit changes on documentation branch

    git commit -m "standardized all docstrings in process.py"

    Switch to develop branch

    git checkout develop

    Pull latest changes on develop down

    git pull

  • Step 2: Then, Andrew merges his documentation branch on the develop branch on his local repository, and then pushes his changes up to update the develop branch on the remote repository.

    Merge documentation branch to develop

    git merge --no-ff documentation

    Push changes up to remote repository

    git push origin develop

  • Step 3: After the team reviewed both of your work, they merge the updates from the development branch to the master branch. Now they push the changes to the master branch on the remote repository. These changes are now in production.

    Merge develop to master

    git merge --no-ff develop

    Push changes up to remote repository

    git push origin master

Note on Merge Conflicts

  • For the most part, git makes merging changes between branches really simple. However, there are some cases where git will be confused on how to combine two changes, and asks you for help. This is called a merge conflict.

  • Mostly commonly, this happens when two branches modify the same file.

  • For example, in this situation, let’s say I deleted a line that Andrew modified on his branch. Git wouldn’t know whether to delete the line or modify it. Here, you need to tell git which change to take, and some tools even allow you to edit the change manually. If it isn’t straightforward, you may have to consult with the developer of the other branch to handle a merge conflict.

  • You can learn more about merge conflicts and methods to handle them here https://help.github.com/articles/about-merge-conflicts/.

Additional resource: http://nvie.com/posts/a-successful-git-branching-model/