CSC/ECE 517 Fall 2011/ch1 1f rs
Comparison of Version Control Systems : The Programmer View
Introduction
Version control also called Sub-Version Control or Revision Control is indispensable in large projects. They make sure that projects are not spinning out of control by letting different programmers work on the project from a different perspective without getting in each other’s way and without doing any damage that cannot be undone.
Some of the most popular version control systems are RCS, SVN, CVS, Mercurial and Git. This article intends to compare these version control systems in a programmer’s viewpoint.
Choice of Version Control Systems
There are a number of solutions which are available for programmers. We have put together a definitive feature comparison for reference on deciding about the best choice for a project.
The main difference between Version Control Systems is whether they are Client-Server based or peer-to-peer based. Either they have a centralized repository where code is checked out, edited and checked in with changes, or a setup where the code is frequently updated from different peer sources, a more decentralized network to keep your code updated with changes.
The article further compares all the Version Control Systems like RCS, CVS and Distributed Version Control Systems. Also, the article compares multiple Version Control applications like Mercurial, Git, Clearcase and CVS. Comparison is done in the view of programmer. Various commands used in different applications are discussed.
Comparison of Version Control Systems
Local Version Control
This is a primitive approach of version control. The system operates on single files only.<ref>http://en.wikipedia.org/wiki/Revision_Control_System</ref> It has no way of working with an entire project. Changes to a file are stored as 'deltas' with the aid of a diff utility. Since the approach maintain many copies of files, it is generally cumbersome with reduced efficiency.
Centralized Version Control
CVS implements a Client-Server model to achieve version control. A central repository is maintained which is used by all the developers of a project. Programmers can only have a working copy of the project. The model allows a developer to checkout only the required files or branches. In this model all operations on the code (commit, exchange of code etc.) involve communication with the central server via a network. This means that these operations are not available when the user is offline. Also issues in communication with the server may disrupt the development activity. Any changes to the project on the server involves an authentication process. Hence only users with certain privileges can commit code on the central server . A developer can checkpoint or create branches on his work only by committing his changes to the central repository. Since all the developers commit changes on the same repository, it is generally challenging to resolve conflicts when merging or reconciling changes from other branches.
Distributed Version Control
This system implements a peer to peer model to achieve version control. No reference copy of the codebase exist by default. However in practice, an official reference copy is indeed maintained. The model requires the developers to possess a full blown backup of the repository , including the entire history. As a result, the developers can collaborate directly without the need of a central authority. Hence it allows developers to be productive even when offline. (e.g. when travelling). Also the overhead of communicating with a central repository is overcome in this approach. Branching and merging fit much better in this approach as each user has his own branch. The administrator's major work is to merge appropriate changes from branches of different users. This is relatively simpler compared to the centralized system. As a result, the distributed system scales much better with the number of people.
Summary of Centralized versus Distributed Approach
| Centralized Version Control | Distributed Version Control |
|---|---|
| Client - server model | peer- to -peer model |
| Single central repository. | Multiple repositories |
| Changes can be committed only Online | Changes can be committed offline |
| Allows developer to checkout required files/branches. | Complete repository has to be downloaded by the programmer |
| Merging and Branching operations are complicated as
multiple programmers work on the same repository at the same time. |
Branching is simple as each programmer gets a branch |
| Repository maintenance becomes more challenging with number of people | Scales better with number of people |
Comparison of Version Control Applications
Overview of Version Control Applications
RCS: The Revision Control System (RCS)
RCS manages multiple revisions of files. RCS operates on a set of files. RCS automates the process of storing, retrieval, logging, identification, and merging of multiple revisions. It is useful for text that is modified frequently which includes source code, programs, documentation, graphics, papers, and letters. This is the most basic and most primitive Version Control System known to programmers.
More information about RCS can be found at RCS
ClearCase Version Management System
ClearCase is a comprehensive software configuration management system. It is used to manage multiple versions of evolving software systems, tracks which versions were used in software builds, used to perform builds of individual programs or complete releases according to user-defined version and system specifications.
These capabilities enable ClearCase to address the issues and critical requirements of organizations that develop software:
Effective development — ClearCase enables users to work efficiently, allowing them to fine-tune the balance between sharing each other's work and isolating themselves from destabilizing changes done. ClearCase automatically manages the sharing of both source files and the files produced by software builds.
Effective management — ClearCase tracks the software build process, so that users can determine what was built, and how it was built. Further, ClearCase can instantly recreate the source base from which a software system was built, allowing it to be rebuilt, debugged, and updated — all without interfering with other programming work.
Enforcement of development policies — ClearCase enables project administrators to define development policies and procedures, and to automate their enforcement.
More information about ClearCase can be found at ClearCase
CVS: The Concurrent Versions System (CVS)
CVS is a Client-Server software revision control system in the field of software development. Version control software keeps track of all the changes in a set of files, and makes sure that several developers (potentially widely separated in space and/or time) can collaborate in real time. The CVS server runs on Unix-like systems with client software that runs on different operating systems on multiple client machines. It is regarded as the most mature version control system because it has been developed for such a long time and has stood the test of time. A fork project of CVS, CVSNT was created to run CVS on Windows servers, and it is currently being actively developed to increase functionality and make it more usable on Windows platform.
More information about CVS can be found at CVS
SVN: Sub Version (SVN)
SVN was created as an alternative to CVS that would be useful to fix some known issues in CVS and also maintaining high compatibility with it. SVN is free and open source with the difference of being distributed under the Apache license as opposed to GNU which distributes licenses for CVS. To prevent the database from being corrupted, SVN adopts a technique called 'Atomic Operations'. Either all of the modifications made to the source are committed or none are committed. Partial changes will not enter the original source. Many developers have switched to SVN as it is a newer technology that starts with the best features of CVS and improves upon them. While branch operations in CVS are expensive and do not really lend themselves to long-term forks in the project, SVN is designed to allow for it. Hence it lends itself better to large forked projects with many directions. Some known disadvantages of SVN includes slower operation speeds and the lack of distributed revision control. Distributed revision control uses a peer-to-peer model as compared to using a centralized server to store code modifications. Peer-to-peer model work better for open source projects which are spread over multiple far away physical locations. The downside to a dedicated server approach is the issue of not having redundancy in the case when the server is down leading to no access to clients.
More information about SVN can be found at SVN
Git:
Git takes a totally diverse approach that differs greatly in comparison to CVS and SVN. The original concepts for Git were to make a faster, distributed version control system that would openly invalidate conventions and practices used in CVS. It is primarily developed for Linux and has the highest speeds on there. It will also run on other Unix-like systems, and Windows agents are known as msysgit. As there is no centralized server, Git is not very suitable for single developer projects or small teams as the code may not necessarily be available when using a normal computer. Workarounds exist for this problem. Developers look out for Git’s improved speed as a decent trade off for the hassle.
More information about Git can be found at Git
Mercurial:
Mercurial is a distributed revision control tool which began close to the same time as Git. Mercurial was originally made to compete with Git for Linux kernel development. Mercurial is primarily implemented in Python as opposed to C, but there are some instances where C is used. This is a major difference compared to other version control systems in Developer's point of view. Users feel that Mercurial is similar to SVN with respect to certain features, as well as being a distributed system. This acts as an advantage for the tool as the learning curve for those already familiar with SVN will be less steep. The documentation for Mercurial is very well compiled and facilitates understanding the differences faster. One of the major drawback to Mercurial is that it does not allow for two parents to be merged. Unlike Git, it uses an extension system rather than using scripts. That may be ideal for some section of programmers, but many find the speed of operation of Git to be a feature they do not want to trade off for anything.
More information about Mercurial can be found at Mercurial
Basic Features
| Software | Repository model | Concurrency model | Programming Languages | Scope of Change | Atomic commits | File Renames | Interactive Commits | Partial Checkout/clone | Platforms supported |
|---|---|---|---|---|---|---|---|---|---|
| RCS | Set of files | Merge | C | File | No | No | No | Yes | Unix-like |
| CVS | Client-Server | Merge | C | File | No | No | No | Yes | Windows, Mac OS X |
| SVN | Client-Server | Merge or Lock | C | Tree | Yes | Yes | No | Yes | Windows, Mac OS X |
| Clear Case | Client-Server | Merge or Lock | C, Java, Perl | File | Partial | Yes | No | Yes | Linux, Windows, AIX, Solaris, HP UX, i5/OS, OS/390, z/OS |
| Git | Distributed | Merge | C,Perl | Tree | Yes | Partial | Yes | No | POSIX, Windows, Mac OS X |
| Mercurial | Distributed | Merge | Python, C | Tree | Yes | Yes | Yes | No | Windows, Mac OS X |
| Feature | Feature Description |
|---|---|
| Repository Model | Describes the relationship between various copies of the source code repository. |
| Concurrency Model | Describes how changes to the working copy are managed to prevent simultaneous edits from causing nonsensical data in the repository. |
| Programming Languages | The coding language in which the application is being developed |
| Scope of Change | Describes whether changes are recorded for individual files or for entire directory trees. |
| Atomic commits | Refers to a guarantee that all changes made are merged, or that no change at all will be made. |
| File Renames | Describes whether a system allows files to be renamed while retaining their version history. |
| Interactive Commits | Interactive commits allow the user to cherrypick the patch-hunks that become part of a commit, instead of having only a file-level granularity. |
| Partial Checkout/clone | Ability to check out or clone only a specified subdirectory from a repository. |
Basic Commands
| Software | Clone | Pull | Push | Checkout | Add | Remove | Merge | Commit | Revert | Rebase |
|---|---|---|---|---|---|---|---|---|---|---|
| RCS | Not Supported | Not Supported | Not Supported | co | Not Supported | rcsclean | rcsmerge | ci | Not Supported | Not Supported |
| CVS | Not Supported | Not Supported | Not Supported | checkout | add | rm | update -j | commit | rm, update | Not Supported |
| SVN | svnadmin hotcopy | svnadmin load | svnadmin dump | checkout | add | rm | merge | commit | revert | Not Supported |
| Clearcase | Not Supported | Not Supported | Not Supported | checkout | mkelem | rmelem | merge | checkin | unco/rmver | findmerge |
| Git | Clone | Fetch | Push | Clone | add | rm | merge | commit | checkout | rebase |
| Mercurial | Clone | Pull | Push | Clone | add | rm | merge | commit | revert | rebase |
| Command | Command Description |
|---|---|
| Clone | Create an identical instance of a repository |
| Pull | Download revisions from a remote repository to a local repository |
| Push | Upload revisions from a local repository to a remote repository |
| Checkout | Create a local working copy from a (remote) repository |
| Add an element | Mark specified files to be added to repository at next commit |
| Remove an element | Mark specified files to be removed at next commit |
| Merge | Apply the differences between two sources to a working copy path |
| Commit | Record changes in the repository |
| Checkout | Restore working copy file from repository |
| Rebase | Forward-port local commits to the updated upstream head |
Sample Workflow involving Clearcase
This section provides an example of a sample workflow used by a programmer to maintain the code in Clearcase.
Joining a UCM Project: After project managers have established the UCM project environment, developers can then set up their work areas. This is accomplished by joining the UCM project. More information about this can be found here
Working with views: A view is the mechanism Rational ClearCase uses to provide access to specific versions of elements or specific parts of code under source control. Rules are the basis to determine which version of each element is visible and accessible through the view for a developer. More information about views can be found here
Sample command to create a new view: ct mkview -tag <view_name>
Sample command to create a new file:ct mkelem <file_name>
Checking in files and checking out files: Checking files in and out are two basic operations you will perform on a regular basis by a developer. Whenever the developer needs to edit a file, he has to checkout the file. And after completion of coding, he needs to check in the file. More information about the process can be found here
Sample command to check out a file: ct co -nc <file_name>
Sample command to check in a file: ct ci -nc <file_name> (when you are working in the same branch where the file is checked out).
Delivering your work: After testing your work in your private workspace, and you are sure that your work is error-free, you are ready to submit your work to the integration stream so that other project members can use the latest version of your work. More information about the procedure of delivering code can be found here
Merging your work: After completion, the development branch must be merged to integration branch where software developed by different teams are merged and tested for seamless operation. More information about this procedure can be obtained here
References
<references/>