A pleasant walk through computing

Home

The 50-10 Code Cube - Revising Pomodoro for Software Development

I've been resolving three aspects of programming:

• The cost of interruptions
• Methods for staying on task
• The benefits of taking breaks

Each of these has studies behind it (see References), as well as anecdotal information. The simple question I wanted to answer is:

What's an effective work-break cycle for programming?

The three most common cycles I know of are:

• Pomodoro - 25 minutes work, 5 minutes break
• 52/17 - 52 minutes work, 17 minutes break
• 90/? - 90 minutes work, then a break

Pomodoro

The Pomodoro Technique is popular, and, besides the work-break cycle, has these suggestions:

• Record interruptions so you can return to them outside the work interval

The problem is, Pomodoro, to my knowledge, hasn't been studied scientifically. When I used it for non-programming tasks, I could see the benefit. The more I tried using it for programming, the less effective I found it.

Numerous studies show there's a recovery lag when resuming an interrupted task. For software developers, it takes about fifteen minutes to recover the previous context--to "get back to where I was." The worst time to be interrupted is then there's the highest mental workload. (Like when I'm in the middle of debugging a method having spent twenty minutes understanding how it relates to five other modules.)

I think it's a mistake, though, to think only in terms of interruptions. A programmer always ramps up when starting (or restarting) a coding session. In other words, we should plan on the first fifteen minutes being devoted to recovering context.

Breaks

I've worked for hours straight, and thought I was being productive. Maybe sometimes I was. But the research shows that regular breaks improve productivity, because the brain needs that time to recover from fatigue.

The Friction

So, where does that leave us regarding Pomodoro?

• It takes a programmer at least fifteen minutes to "ramp up"
• A Pomodoro work period is twenty-five minutes
• That leaves ten minutes or less of effective work time

Well, no wonder Pomodoro wasn't working out!

I have tried the 52/17 cycle. I found the seventeen minute break too long. I read an article where office workers tried it themselves, and it was pretty challenging. I suspect that the original study wasn't looking at programmers....

There are benefits to Pomodoro. The task planning and completion help a lot with emotionally attaching to success, and with chunking down larger items.

Proposal - Code Cube, 50-10 Time Box

I've been trying this for about a month now, and it's working well for me. So, maybe it will work for you, too. Basically, use Pomodoro Technique but with a 50 minute work cycle followed by a ten minute break. This allows for the fifteen minute ramp up, then thirty-five minutes of solid, focused work. I've found that I'm usually ready for a break at the fifty minute mark.

Even if I wasn't "ready" for a break, when I took it I found I needed it.

I use an online tool named Marinara Timer to manage my time boxes. It has a Custom timer where I configured my cycles.

1. It aligns on the hour, making it easy to understand.
2. I get enough context recovery time.
3. Four cycles gets me to lunch, the remaing four finish my day.
4. A ten minute break is enough to take a walk, catch up on email, even have a short conversation. It's an effective coding break.
5. That break has often helped me solve a thorny problem.

A complete work session looks like this. I use a Markdown document to track my work sessions, distractions and notes, but paper and pen (or pencil!) would work as well or better.

# Distractions

# Round 1
- [ ] Validation rules
- [ ] Unit tests

# Round 2
- [ ] Review code, make sure setting values correctly
- [ ] Improve queries based on understanding of rules

# Round 3
- [ ] Design workhorse feature

# Round 4
- [ ] Begin coding workhorse feature


2. Start my timer.

3. Work deliberately, with intense focus. Record distractions.

It's important to not noodle around during the work time. Manage distractions! This is a hallmark of deliberate practice, and an aspect of meditation and mindfulness.

# Distractions
- [ ] Email from Luis, not urgent but should get back to him soon.
- [ ] Birthday gift research for Joanie.

# Round 1
- [ ] Validation rules
- [ ] Unit tests


# Distractions
- [ ] Email from Luis, not urgent but should get back to him soon.
- [ ] Birthday gift research for Joanie.

# Round 1
- [x] Validation rules
- [ ] Unit tests


4. When the work cycle ends (I'm currently enamoured of the Whoosh sound), take a break.

It can be hard to stop, but do it. And, I recommend walking away from the desk, and/or closing your work windows, so you don't get sucked back into the task.

5. Repeat.

Conclusion

I was spurred to write this post because of this article. The author, Tyler Hakes, gives several examples from his workday of how he reduces interruptions in his day, and it's worth reading.

Below are a bunch of references I read both before and while creating this entry. They include links to the primary research papers that other articles refer to. I've called out some quotes that I found useful.

References

Papers

When resuming an incomplete programming task, the developer must remember their previous working state and recover knowledge about the software. Details of working state might include recalling plans, intentions, and goals. Details of knowledge might include plan progress, component mechanisms, and domain representations.

Our data suggests that people compensate for interruptions by working faster, but this comes at a price: experiencing more stress, higher frustration, time pressure and effort.

DeMarco reports that the recovery time after a phone call is at least 15 minutes.2 Even though we could not measure recovery time exactly, we believe his estimate to be valid. If more than 10 interrupts occur during a day, the time between the interrupts becomes too short to accomplish product development work.

Work fragmentation as result of interruptions usually demands extra effort to recover and resume pending activities: a study of 24 information workers found that a worker needs on average 25 minutes to get back on an interrupted task [5]. Similarly, Iqbal and Horvitz [6] found that people experience disorientation and loss of context when multitasking. Czerwinsky et al. found that after experiencing work fragmentation people found it more difficult to perform interrupted tasks and took longer to complete them [7].

LINQPad - Refresh Entity after SubmitChanges to Avoid Cached Results

Today I was using LINQPad to see how SQL record values were being affected by table triggers. The results were puzzling, so I did the same test using SQL Management Studio and got different results. Clearly, there was caching going on!

Here's roughly what my LINQPad C# Statements script looked like (table names changed to protect the innocent). I was going directly against the database, not using an Entity Framework DbContext, so LINQPad is using LinqToSql for the connection.

var reportId = "1234";
//Change Status from P to C. Trigger should reset to P because Completed is false.
SubmitChanges();

var report = ReportParents.Single(a => a.ReportID == reportId);

d.Dump();


And my result:

But in the database, both statuses were (correctly) set back to P.

HEADER
STATUS Completed
------ ---------
P      0

REPORT
STATUS Completed
------ ---------
P      0


This was happening because Linq2Sql caches query results during a single Context session. Caching can be turned off by not tracking objects (ObjectTrackingEnabled = false;), but that will prevent SubmitChanges from running.

The solution is to use Context.Refresh to requery the database for the entity.

var reportId = "1234";
//Change Status from P to C. Trigger should reset to P because Completed is false.
SubmitChanges();

//Force fresh result. Caching was preventing accurate trigger results.

var report = ReportParents.Single(a => a.ReportID == reportId);

d.Dump();


Now, running the script returns the correct results.

Create Blog Post Using Markdown Monster's Command Addin

This is a quick entry, mostly for me. I've been using the very good Markdown editor, Markdown Monster, and wanted to automate creating blog posts. I have a manual process that I thought I could somewhat automate.

MM has an addin that lets me write C# scripting and control a lot of the editor.

Through both the documentation plus trial-and-error, I was able to do the following:

1. Prompt for a post title.
2. Create the post file in a known folder.
4. Open in a tab for editing.
5. Open the post folder's "images" folder.
6. Commit and publish

Here's the code for creating the post, some of which is seen in the above screenshot.

There were some things I couldn't accomplish, noted in the code comments, probably due to my ignorance of how the MM API is supposed to work.

#r Westwind.Utilities.dll
#r Microsoft.VisualBasic.dll
using Westwind.Utilities;
using System.IO;

string defaultPath = @"c:\path\to\posts";

string title = Microsoft.VisualBasic.Interaction.InputBox("File Name", "Enter file name", "", 600, 300);

if (String.IsNullOrWhiteSpace(title))
{
return null;
}

string datetime = DateTime.Now.ToString("yyyyMMdd-HHmm");
string folder = Path.Combine(defaultPath, datetime + " " + title);
string file = Path.Combine(folder, title + ".md");
string images = Path.Combine(folder, "images");
string publishedOn = DateTime.Now.ToString("yyyy-MM-dd HH:mm");
string slug = title.ToLower();
slug = slug.Replace(" ", "_");
slug = slug.Replace("?", "_");
slug = slug.Replace("&", "_");
slug = slug.Replace(".", "_");
string meta = String.Format(@"---
Title          : {0}
PublishedOn    : {1}
Slug           : {2}
Tags           :
Status         : Draft
---
",title, publishedOn, slug);

Directory.CreateDirectory(images);
File.WriteAllText(file,meta);
//Model.ExecuteProcess("explorer.exe", images);
Model.Window.OpenTab(file);
//I can't figure out how to set the cursor position
//The code seems right, but the cursor doesn't move unless I run the command
//from the addin window. Even then, the Console.WriteLine is what
//forces it to work.
var editor = Model.ActiveEditor;
var doc = Model.ActiveDocument;
var len = doc.CurrentText.Split('\n').Length;
Console.WriteLine(len);
Console.WriteLine(editor.GetLineNumber());
editor.SetCursorPosition(0,len);
editor.SetEditorFocus();

//Couldn't get this working
//var editor = Model.ActiveEditor;
////HACK: For some reason this allows the document to be edited.
//Console.WriteLine(editor.GetLineNumber());
//editor.SetSelection(meta);
//editor.SaveDocument();


To open the images folder, I created this command-let:

#r Westwind.Utilities.dll

using Westwind.Utilities;
using System.IO;
using System.Windows;

string folder = Model.ActiveDocument.Filename;
folder = Directory.GetParent(folder).FullName;

string images = Path.Combine(folder, "images");
if (Directory.Exists(images)) {
Model.ExecuteProcess("explorer.exe", images);
}
else {
MessageBox.Show("No folder " + images);
}


And, to publish, I do this. I have a batch file in my blog folder that takes care of the actual commit/push to Git.

#r Westwind.Utilities.dll

using Westwind.Utilities;
using System.IO;
using System.Windows;
using System.Diagnostics;

string folder = Model.ActiveDocument.Filename;
folder = Directory.GetParent(Path.Combine(folder,@"..\..\..\")).FullName;

string cmd = Path.Combine(folder, "publish.cmd");
cmd = @"""" + cmd + @"""";
Console.WriteLine(cmd);
//Model.ExecuteProcess("cmd.exe", "/C " + cmd);
var si = new ProcessStartInfo();
si.FileName = "cmd.exe";
si.Arguments = "/C " +  cmd;
si.WorkingDirectory = folder;

Process.Start(si);