Introduction to Git –- Fall 2020

Lecture 1: Why use version control?

Slides: https://hackmd.io/@hpc2n-git-2020/L1-motivation

What is version control?

In software engineering, version control (also known as revision control, source control, or source code management) is a class of systems responsible for managing changes to computer programs, documents, large web sites, or other collections of information. –- Wikipedia

Version control systems (VCS)

… systems responsible for managing changes …

Why use version control?

In an ideal world, things develop linearly:

• Every new version is an improvement upon the previous version.
  • No need to backtrack.
• Everyone known what everyone else is doing
• In the end, things are simply finished.

digraph {
  rankdir=LR
  Mon[label="Monday's\n improvements"] [fixedsize=circle]
  Tue[label="Tuesday's\n improvements"] [fixedsize=circle]
  Wed[label="Wednesday's\n improvements"] [fixedsize=circle]
  Mon -> Tue
  Tue -> Wed
}

In real world, things develop non-linearly:

• A new version can anything between
  • a complete catastrophe and
  • a major breakthrough.
• People do not know what others are doing
• Sometimes we are simply fixing earlier mistakes…

digraph {
  rankdir=LR
  Mon[label="Monday's\n improvements"] [fixedsize=circle]
  Tue[label="Tuesday's\n mistakes"] [fixedsize=circle]
  Wed[label="Wednesday's\n corrections"] [fixedsize=circle]
  Mon -> Tue
  Tue -> Wed
}

Going back to an earlier version

Sometimes, it is easier to simply backtrack to an earlier version…

digraph {
  rankdir=LR
  Mon[label="Monday's\n improvements"] [fixedsize=circle]
  Tue[label="Tuesday's\n mistakes"] [fixedsize=circle]
  Wed[label="Wednesday's\n improvements"] [fixedsize=circle]
  Mon -> Tue
  Mon -> Wed
}

Where is this earlier version?

• CTRL + Z
• my_file.txt, my_file.txt.old, …
• My project/
  • 2020-08-12/
  • 2020-08-13/
  • …
• Daily home directory backup

Challenges and obstacles

• Prone to mistakes
  • CTRL + Z has limits, overwritten/deleted files, human/hardware error
• How much to save?
  • Individual files? Everything? How much space is required?
• How to organize versions?
  • What is the difference between different versions?

Overall, difficult to manage!

What about the granularity?

digraph {
  rankdir=LR

  subgraph cluster1 {
    t1a [label="Component A\n improvement"] [fixedsize=circle]
    t1b [label="Component B\n mistake"] [fixedsize=circle]
    t1c [label="Component C\n improvement"] [fixedsize=circle]
    label="Mondays's changes"
  }

  subgraph cluster2 {
    t2a [label="Component A\n improvement"] [fixedsize=circle]
    t2b [label="Component B\n correction"] [fixedsize=circle]
    t2c [label="Component C\n mistake"] [fixedsize=circle]
    label="Tuesday's changes"
  }

  subgraph cluster3 {
    t3a [label="Component A\n mistake"] [fixedsize=circle]
    t3b [label="Component B\n improvement"] [fixedsize=circle]
    t3c [label="Component C\n correction"] [fixedsize=circle]
    label="Wednesday's changes"
  }
  
  t1a -> t2a
  t1b -> t2b
  t1c -> t2c

  t2a -> t3a
  t2b -> t3b
  t2c -> t3c
}

This compounds the problems!

How does VCS solve this?

• Stores the history using snapshots (commits)
  • Each snapshot represents the project in a given point of time
• Manages snapshots and associated metadata
  • Naming (tags), comments, dates, authors, etc
• Easy to move between different snapshots
• Can handle different degrees of granularity
• Can handle multiple development paths (branches)

Comparing and joining

• VCS makes it easy to compare different snapshots
  • Named revisions, comments, time information, author information
  • Diff tools
  • Search tools
  • Bisection search
• VCS also allows the joining (merging) of different snapshots
  • Easy to experiment with ideas

Collaboration

• One of the primary functions of VCS is to allow collaboration
• Usual setup: server (remote) + multiple clients
  • People work locally and send (push) the changes to the server
  • VCS keeps track of what has been done and by whom
• Safer since mistakes can be easily remedied
• The contributions of several people can be merged

Backup

• VCS functions as an backup
• Locally, the system maintains a copy of each file
  • Usually only the changes or the files that have changed are stored
• Globally, lost files can be recovered from the server

Integration

• VCSs such as Git have been integrated with several services
  • HackMD, Overleaf, …
• Services such as GitHub can do almost everything for you
  • Store history, distribute, testing / continuous integration, bug reports, milestones, website, …

Practical use cases

What are the practical use cases for VCS?

Source code

• Many VCSs are designed for managing source code
• Manage deployment (production, development, testing, etc)
• Manage published versions (v0.1 etc)
• Manage (experimental) features
• Bug hunting

Latex files

• Track which version of a manuscript has been
  • submitted,
  • revised and/or
  • accepted
• Collaboration between several authors

HPC: batch files and data

• Track different version of your batch scripts
  • Easy to check the used configuration afterwords
• Track input and output files
  • Limited to smallish files