Homebrew VC1

Hip-Hop Approach

Homebrew VC2

Filing Clerk Approach

Homebrew VC3

Cloud Storage Approach

MIME-Version: 1.0
Received: by 10.220.194.194 with HTTP; Thu, 19 Jun 2014 10:10:44 -0700 (PDT)
Date: Thu, 19 Jun 2014 13:10:44 -0400
Delivered-To: fitch@datamininglab.com
Message-ID: 
Subject: Backup of mumble_final_report_and_data_v24.zip
From: Simeon Fitch 
To: Simeon Fitch 
Content-Type: multipart/mixed; boundary=001a11c2dd740cbee804fc33756a

--001a11c2dd740cbee304fc337568
Content-Type: text/plain; charset=UTF-8

(backup to self)

--001a11c2dd740cbee304fc337568--
--001a11c2dd740cbee804fc33756a
Content-Type: application/x-zip; name="mumble_final_report_and_data_v24.zip"
Content-Disposition: attachment; filename="mumble_final_report_and_data_v24.zip""
Content-Transfer-Encoding: base64
X-Attachment-Id: f_hwmbquyc0

iVBORw0KGgoAAAANSUhEUgAAAnYAAAGCCAIAAADi3Rk8AAAKr2lDQ1BJQ0MgUHJvZmlsZQAASA2t
...
@datamininglab.com>@datamininglab.com>+qyohum9qg8bex3z3cxcjtaxynahmfmu8iw_dxd1n1svz3g@mail.gmail.com>

Abdication-Oriented VC1

Abdication-Oriented VC2

My-Tools-Take-Care-Of-It-For-Me Approach

• Microsoft Office Revision Tracking (definitely not the TARDIS.)
• Google Docs

Abdicating Control Hurts!

Time Lord VC1

Architecture Types:

• Centralized: CVS, SourceSafe, Subversion, Team Foundation Server
• Distributed: Git, Mercurial, Bazaar

Time Lord VC2

• A Time Lord with both hearts pumping will use a Distributed Version Control System (DVCS).
• The software team at ERI uses Git and Mercurial, both DVCSs.

Time Lord VC3

System selection:

• If the client dictates a VCS, there is no decision...
  • Learn and embrace what they have as quickly as you can, with enthusiasm!
  • Any version control is better than no version control, as long as you understand its limitations and it doesn't create a false sense of security.
• If there's a system tightly integrated into the modeling platform...
  • Push it to its limits, and augment with Git as needed (i.e. for documentation or data)
  • Tip: There's a "portable" (non-installing) version of Git available for your thumb drive.

Time Lord VC4

• If the VCS system selection is completely up to you, my recommendation is to use Git with SourceTree.

• Mercurial with SourceTree would be my second choice, but be aware of which features are implemented as plugins and how to install them.

Time Lord VC5

Analytics tools context:

• RStudio supports Git
• Knime supports most everything, via Eclipse platform
• SAS Data Integration Studio supports CVS and Subversion
• SPSS Collaboration and Deployment Services: didn't have time to figure it out.

Terms

• Init
• Clone
• Add
• Checkout
• Commit
• Push
• Pull
• Branch
• Merge
• Diff

Checkins

Checkouts and Editing

Diffs

Branching

Merging

Conflicts

Appendix

Warning: random stuff ahead.

Versions of What?

• Ultimately, all the electronic by products of your technical work.

• It's like flossing: you don't have to floss the teeth you don't want to keep.

• The time series data over all your technical artifacts.

• The question is more appropriately stated: what do you not version control.

What to Version

Let's ask Hadley!

• Examples:
  • An implementation of the Grammar of Graphics in R
  • Data from the 2006 ASA data expo
  • An R package contain all baby names data from the SSA
  • Wickham family recipes
  • Advanced R programming: a book
  • ideas

Notable Quotes

But don't take my word for it...

No Such Thing as a One-off Script

Devalytics: Making your analytical work easy to replicate, build upon, and scale while saving significant amounts of time in the process

In the same way the DevOps mantra is "Infrastructure as code," today’s data scientists need to think of all their scripts as actual software that will require ongoing maintenance, enhancement, and support. To paint with a broad brush, there is no such thing as a one-off script. As soon as anyone else has access to it, or if it even sticks around on your local filesystem, it will almost inevitably be reused and applied to different situations in the future.

Evaluate Effectiveness

How effective is your time travel machinery?

A lot of the analysts and data scientists we know don’t do version control... If you ask them what do you do if you want to work on different versions of an experiment with different parameters, they say, ‘Well, I make a copy of my files. I have scripts and script [copies]. How do you ensure you can get back to your old versions of your work? You use version control.

Engineering Practices

Faster iteration is a competitive advantage.

It terrifies me that... many data scientists don't use source control. If you ask them they might say something like "Sure I use version control! I email myself every new version of the R script. Plus it's on dropbox." ...Successful data scientists learn the value of good pipeline engineering [because] a well-engineered pipeline gets data scientists iterating much faster, which can be a big competitive edge...

Enabling Reproducibility

To stand behind our results, we have to be able to reproduce them!

Rule 1: For Every Result, Keep Track of How It Was Produced

Rule 4: Version Control All Custom Scripts

Rule 5: Record All Intermediate Results, When Possible in Standardized Formats

Rule 7: Always Store Raw Data behind Plots

Collaboration

Solving big, interesting problems requires working effectively with others.

While these tools and technologies are fundamentally changing how we collaborate on science, there is still considerable room for improvement in how we are using them.

New Ways of Working

Everything we do needs to support who we are. —D. Bailey

Version control systems (VCS)... are now finding new applications in science. ...Git, provides a lightweight yet robust framework that is ideal for managing the full suite of research outputs such as datasets, statistical code, figures, lab notes, and manuscripts. ...this tool can be leveraged to make science more reproducible and transparent, foster new collaborations, and support novel uses.

"Big Data Needs to Grow Up"

We have to adopt software and practices to manage the accelerating information flow.

We are in an Information Revolution... but it is entering a new stage... [with] previously unimaginable quantities of data to measure, analyze and act on. These new data sources promise to transform our lives... but we need to get much better at handling all that data we’re producing and collecting.