Monday, January 9, 2017

Is there a perfect API Design?

Part I:The Problem

I write a verification framework called ApprovalTests. It uses itself to test itself and is generally bug free. So when exploratory tester extraordinaire Maaret Pyhäjärvi wanted to use it as a test target we were both rather excited. I was excited to show off how good automated testing and TDD can be. She was excited to show off how much it still failed to cover.

She won.

To be clear, the stuff I tested was pretty solid. However, I was woefully incomplete on the system as a whole. A mere hour of testing discovered gaping holes in different environments, documentation and onboarding of new users, and usability issues with my API.

These are hard problems, many of them still haven't been solved and I wanted to talk about one in particular that I am still struggling with today:


I use something called reporters and annotations in ApprovalTests. It means you can write code like


The issue was in discoverability. If you start typing this you get very little help from your editor:

You get some other reporters [KDiff, Tortoise] but not many and these happen to be useless and DiffReporter will use them if they exist on your system anyways.

If you typed "Reporter" instead this goes away and the over 50 options will present themselves to you, but it's not intuitive to this and a few painful usability tests showed this to me over and over as I watch is silence and frustration.

Renaming can fix this


I ran an online poll, most people preferred this. 70% vs the 30% that preferred the previous version.
I think I'm in the 30%, I prefer my classes to have Noun names, but I might be partial because this is what I'm used to.

So I'm faced with changing a lot of the API.

The issue is:

How do I know which one is the right answer?

How do I know there *is* a right answer?

Part II: 2 Pepsis and the world of choice

Malcolm Gladwell did a great Ted Talk on choice. In it he talks about a taste test to find the perfect sweetness for Pepsi. There were 2 peaks, so they averaged them out, but this isn't the right answer, the right answer is there are 2 preferences for sweetness. Because of this as a culture we have changed from a single 'perfect' spaghetti sauce to an aisle of choices

Maybe we should have the same occur with API's?
Maybe the issue isn't to have either UseReporter or OnFailure. Maybe I need to have both.

This requires a bit of finesse. I don't yet have answers on how to version and package these solutions.
Should they be in separate nuget packages? Should they interplay with each other?
How to I balance having a clear way with choice paralysis.

I don't know the answers to these questions, but I am beginning to see that maybe there isn't the one perfect API...

Saturday, July 30, 2016

5 ways to do Decision trees in C#

At Agile2016 we spent the mornings doing Mob Programming working on the Tennis Refactoring Kata from Emily Bache.

Thursday we ran accross this issue:

#1. If Structure

This is fairly easy to code as is, the trouble is it looks horrible:

So we started to play around with alternative structures...

#2. Pass in result

One solution is to call all the methods but abort early if you've already found a solution.
This looks a lot nicer, the problem is it leaks into the underlying methods
So we keep looking..

#3. Linq

If you like linq you can solve this with a bit of functional code

This reads nicely, but is a bit confusing as to why it works and means all your methods must not use any parameters.

but It can also be used with the yield keyword, which is neat and allows parameters but splits logic.
so we kept looking...

#4. Do If Null

We were able to get closer to the 'Pass in result' without leaking with the addition of an extension method.
at this point someone pointed out the ?? operator...

#5. ?? Operator

The ( a ?? b ) operator says if  a is null do b. The resulting code is rather nice
It's worth noting that if your language has 'truthiness' (like javascript) you can also do this with the or (||) operator

Which is Best?

Of course that's up to you. For us most liked the ?? operator best, while a few thought the Linq solution was the nicest.

Which one do you like? 

Wednesday, June 15, 2016

Interview with Bryan Beecham about Pairing w/me the first time

I put a interview Maaret recorded with Bryan Beecham online yesterday.
It's about the first time Bryan paired with me using Strong-Style Pairing.

Here's the video:

and the Transcript:

 So the first time i met llewellyn I was in a session at an agile conference. They were talking about a developer and tester pairing and the benefits of that and it was a great session and the really cool thing was that it was very hands on so I ended up pairing at a table and I had this guy. This guy there with like kinda a little cocky he's got blue hair and I'm pairing with him. And I'm like great, I'll learn. I know how to code a bit... but we are using eclipse (i think) and I don't know eclipse that well... and he asked me that, maybe he noticed my hesitation and he said how well do you know this IDE?
 and I said "ah, it's not my thing"
 and he said "that's ok, because I'm going to tell you what to type."
   and I thought maybe he thinks I'm new to pair programming or something and I don't know about this but no, I was in for a whole new world of understanding because the person, who turned out to be llewellyn, introduced me to some, I guess now it's been modified to strong pairing which has some other meanings to it, but he told me what to do and told me what to type and when you think  
 oh yeah, type something very high high level, but no he was telling me on a very low level of type this word. press enter press semicolon you need a curly brace there and I was not allowed to use the mouse it's like no, no no no it's
 alt+r for that
and word by word, but I'm like 'keep it cool' 'keep it cool' because inside i'm about to boil over and just be furious, right, but then I noticed something was happening.

  I was noticing that I started to know what extract method was I started to pickup on somethings I started to understand the syntax better and his language started changing as this happened and he started talking at a more abstract level and his instructions got moved to a higher level and we were actually picking up speed and I was like 'holy smokes there's something to this' so although he's like completely obnoxious and annoying at first that there's really something here.

  So I that was my introduction to Llewellyn and I would like to say that as this point I feel like he is becoming a  very good good friend of mine  

Friday, June 3, 2016

Using intellisense for documentation

A few days ago I created an open source project called EclipseTips with help from Aki Salmi

It's a small jar that has classes without implementation that provide useful cheat sheet hints while you are working.

Here's what it looks like

The really interesting thing about this is it is a hack about using editor inline help solely for the purposes of help. This is something I have not seen done before.

Keep in mind that these java classes serve no purpose other than to provide documentation.

Because this is a new area, I wanted to share some lessons learned as well as places I think this could be expanded to.

Lessons Learned from version 1

Hack then Generate

Javadocs is a reasonable place to add documentation when coding, but it is not a good place write real documentation. Once you get the general idea and a few working prototypes, you are going to want to create a way to generate the documentation classes.  

For example, here's how we created the Completions Tips data. We then generated both the mac and windows tip objects from here.

You can take a full look at the generation process here.

Simple Html

While the documentation rendering doesn't support much, you can use simple html like tables and images and UTF8 characters to make your text more readable. Just keep it simple.

Use tl;dr curation

A lot of the value is in the discoverability and quick readability. We approached this in a couple ways.
  • We don't show all the tips, only the ones we judge 'most' useful
  • We grouped things so you could focus on pieces, but we also had a full listing under .Mac()
  • We named the groups starting with an underscore ( _Lines ) so they would be together in the editor

Possible Next Steps

Other Editors

  • Visual Studio / Resharper (for vs keybindings & R# keybindings)
  • IntelliJ

Example / Starter / Cookbooks for Api

When you are getting started you often find your way to websites to lookup quick getting started examples. This as stuff that doesn't exactly belong in the actual code itself. Especially as the examples might be how you discover the packages, class and methods you want to call.

I could easily see documentation classes for things like linq, unit testing, markdown, regex, etc.

Of course, you might want to put that directly on the methods (as documentation was originally intended) but this also opens up the ability to write a documentation project for a project you don't actually own yourself.

Let's see what the future will hold...

I'm excited to see what will come out of this discovery. If you have ideas please reach out to me on twitter ( @LlewellynFalco ) Share what you've created, or ping me and we can pair and create it together.

Tuesday, March 22, 2016

The coordination tax of working alone

If you have team members working together on a project but separately on their own piece of the project then you need to coordinate that work.

There are lots of interestingly different ideas on how much 'cost' there is to this coordination. Ranging from it's a very low overhead, to it requires a full time manager who is the highest paid person on the team. Often these drastically different opinions are held on the same teams by the same people.

What I want to explore today however is what does the actual coordination looks like and what it would mean to get it right.

To start off with, there are a few types of coordination:

Multi-functional coordination

If someone is building the database and someone is building the UI those things need to talk to each other; usually through a middleware build by a third person. If they don't line up the costs are large indeed, ranging from "we need to redo this" to the far worse "Let's ship the wrong thing". 

The coordination includes:
  • knowing what each piece looks like
  • which functionality goes where
  • how to connect them
If you know the answers to all of this, then I believe you can make a fairly good assembly line fashion process. 

However, I would like to state categorically 
"If you have not yet built it, you don't know everything involved"

or as the great Woody Zuill said

"It's in the doing of the work that we discover the work that needs to be done"

And I don't mean done 'something like it'. If you've ever done a remodel on your house you will notice the estimates don't match the end result. It's not that this is the first remodel the contractor has ever done. Each situation has it own unique unknown unknowns.

Unfortunately, in software we are always doing something new because duplication is a copy and paste away.

Coordination of these types of tasks is expensive. You need an level of detail and level of oversight that is time consuming to obtain. If you don't get it right there is a lot of waste in the process that is silent.

Multi-person coordination

Even if you have people with the exact same skill set you will often have them working separately in an effort to maximize throughput. Here the coordination is in reducing duplication and keeping the code united. You don't want different people creating the same things or worst the same things in a different way where it's hard to realize they are even the same things. Ideally you want everyone producing unique work in the best possible way.

Perfect coordination 

I think perfect coordination is easiest to picture in a single entity. This entity is aware of the details and the larger picture. It has the skill sets for all aspects of the code. It incorporates new knowledge and processes  the moment it learns them. It has a memory of the previous work and knows when it should reuse pieces. It has high self awareness and doesn't have to spend a lot of time in meeting talking about work.

For many small projects this entity has been a single person.

I hypothesise that it is also the reason a Mob can be effective.


Sunday, March 6, 2016

Two people working on the same thing

One of the most common objections to Pair or Mob Programming is

"How can more than one people working on the same thing be productive?"

Let's put pairing that aside for a second though and look at common scenario:

Duplication in the average code base

You have a fairly big system, maybe 300,000 lines of code. How much of that is some form of duplication? Keep in mind that duplication is more than just copy and paste. It's anytime you need to change something in more than 1 place. This inconsistency is particularly insidious because it means new creative solutions for each one, plus a harder time identifying the problems later.

Note: I have seen cases of over 1500 instances of duplication in a single code base of this size. (If this seems familiar, hire me :-)

This duplication is one of the main causes of technical debt.

Obviously, you should remove this. You can see some techniques to do just that in our Practical Refactoring video but let's instead focus on the reason this duplication get created in the first place.

Keep in mind that often this duplication isn't written by the same single person. Often in is many people, over many months. Often they aren't even aware there is simular code somewhere else in the system. Who has time to just read over 300 thousand lines of code? [this is roughly 15 copies of the last Harry Potter Book] 

I doubt anyone would actively say "Please have a couple people write the same thing over 4-5 places in different ways". But, of course, no one is saying that. These areas of duplication are smaller than the tasks that are ever discussed. It's just something that happens but is unaccounted for. So let's account for it:

Time to do the Math

We need an amount of duplication which is a bit of a problem as there isn't a nice industry number for this. So I'll use a range from 5%-50% for the common scenario of 300,000 Lines of Code (LOC). 


300,000 * 0.05 = 15,000 LOC. 
15,000 LOC / (100 LOC/Day) = 150 days to create.


300,000 * 0.50 = 150,00 LOC.
150,000 LOC / (100 LOC/Day) = 1500 days / (300 day/year) = 5 years to create.

Meaning the average project has inadvertently budgeted between 1/2 man year to 5 man years to have 2 people work on the same thing for the purpose of making the code worse.

For which I have to ask:

"How can more than one people working on the same thing to be productive?"

Thursday, March 3, 2016

Defining 'Technical'

TL;DR: Technical can mean advanced, lingo and/or labs.

I'll like to a moment to get technical about the concept of technical talks...

It is often discussed before, during and after a conference about the 'right' amount of 'technical' talks. This is a very hard subject and in some of my recent discussion I started to get a glimpse into why:

"Technical is a poorly defined umbrella term."

As I started to break down what it means to be a technical talk, I started to see 3 axis emerge on aspects of a talk that can be considered technical.

The aspects of Technical

1) Lingo vs. Clear Words

"When my WCF backed by a RDMS wasn’t rendering in WPF I opened MSDN and said WTF"
If you are a .net developer this statement makes perfect sense, but everyone can understand
"When my program wasn’t connecting to the database I had to read the documentation online"

Pros: Lingo conveys large info in small words.
Cons: It isolates many people and makes them feel stupid while making the speaker look smart.

2) Advanced vs. Beginner

How many prerequisite do you need to attend this session?

If you hear the statement
"Then just apply your basic differential equations..."
you better have taken a few years of math. Whereas

"With just your basic arithmetic you can..."
is probably accessible to most people.

To be clear, I'm not trying to put a value judgement on the topic. Advanced topics don't inherently hold more value than beginner topics, but they do have more barriers to entry.

Pros: There are many great and valuable things in the advanced space.
Cons: Understanding the prerequisite most people have is very difficult and error prone. These sessions also have a limited audience by their very nature.

3) Labs vs. Concepts

Is this where the rubber meets the road or a 10,000 foot overview?

Will the presenter just mention that
"It’s saved in a settings configuration" or are we going to the command line and opening emacs?

It's worth mentioning that there is a fairly solid middle ground between concepts and labs that I would refer to as examples/demos. Like a lab where you are watching rather than doing.

Pros: Labs get to the real knowledge that concepts simply gloss over. For example: I understand the concept of flying a plane, but you don't want me to be your pilot.
Cons: Labs cover a much smaller space in a much larger time scale than concepts do. They are also very specific to a particular context that might not transfer to your own.

In the end we have a complicated 3D space looking something like this:
This is a very complicated space and most talks lie somewhere in between. However, I have found the most value from examining the 8 outermost points.

The 8 Extremes of Technical

1) Clear Beginner Concepts

This the least technical presentation.
That's not to demean the presentation. These talks for valuable and deliver a lot of (albeit shallow) understanding to a large audience.
This is the most common type of talk at most conferences.

2) Clear Advanced Concepts (GOOD)

This is technical in that you need a fair amount of knowledge to get into the conversation.
I have seen a fair amount of these talks at strangeloop. They are the kind of conversation where you understand every word but your head is still hurting.
These can also be the topics that open your eyes and stick with you for ages.

3) Advanced Concepts with lot’s of lingo

These are error prone. Like the above they make your head hurt but there is a good chance you didn't understand every word and you are too embarrassed to ask for clarification.
While Academic conferences seem to have a bunch of these the truth is they are everywhere.

4) Beginner Concepts with lingo (BAD)

I can see no use for this. Lingo implies a prerequisite for understanding but beginner states none is needed.
Obviously there is some use of this from a dark place where the speaker wants to feel superior or confuse the listener, but I believe more of this happens by accident because of
'the curse of knowledge'. Namely, once you internalize the lingo you forget that it is even is lingo; It's meaning just seems so clear.

5) Beginner Labs with lot’s of lingo (BAD)

See above, except because you are in the details the confusion is even more damaging.

6) Advanced Labs with lot’s of lingo

Interestingly, this seems to be a fairly good place for lingo. While I would still suggest brief definitions when introducing the terms (to avoid having people too embarrassed to ask) the efficiency that is provided by lingo is very beneficial here, you have people that have been in the space a while to fulfill their prereqs and you are using specific technology that needs names and labels.

7) Clear Advanced Labs 

Even better.

8) Clear Beginner Labs (GOOD) 

Beginner and Introductory talks are fantastic. They are heavily attended and help bring up new people (Doubles every 5 years means 50% of industry is less than 5 years of experience). Making these labs clear welcomes newbies and gives them a good experiences and solid learning. Labs help to make the learning concrete. I personally really enjoy these type of presentations and give many of them myself (so I'm a bit biased).
They do a have a couple pitfalls though.

  • They require more effort on the student. This is particularly bad if they are held right after lunch or competing with an easier session.
  • Sometimes you can't see the forest for the trees. Students will feel that they are missing out on the big important stuff because they are struggling with the details.
  • Experts like to discount them. "that's too basic" or "We've all seen that before" is the most common thing I hear from experienced people.

So those are my thoughts into everything that usually gets wrapped up in the term 'technical'.
Hopefully this blog post rates as a Clear, Advanced and somewhere in between Concepts/Lab.
If not, let me know in the comments.