Lesson Learned: Don’t Use Low-Level Lib To Test High-Level Code

Summary: Using a fake http library to test logic two levels above HTTP is unnecessarily complex and hard to understand. Fake instead the layer directly below the logic you want to test and verify the low-level HTTP interaction separately. In general: Create thin horizontal slices for unit testing, checking each slice separately with nicely focused and clear unit tests. Then create a coarse-grained vertical (integration-like) test to test across the slices.

The case: I want to test that the method login sends the right parameters and transforms the result as expected. Login invokes post-raw which calls an HTTP method. Originally I have tried to test it by using the library clj-http-fake but it proved to be unnecessarily complex. It would be much better to fake post-raw itself for testing login and test the original post-raw and its HTTP interaction separately, using that library.

Most interesting links of Mars ’13

Recommended Readings

A lot of stuff this month since I have finally got time to review some older articles. Quite a few articles by Fowler. Few really great (yet short) talks on agile & SW development.

Top

• Agile in a Nutshell (originally Agile Product Ownership in a Nutshell) by Henrik Kniberg – the best explanation of the agile development process ever, in just 15 minutes and with wonderful animation; every developer should see this. Some highlights: the most important task of product owner is to say NO so that backlog doesn’t grow infinitely; at start, the estimates of size and value will suck and that’s OK because the value is in the conversation, not in the numbers (that are anyway just relative); the goal is to maximize outcome (value), not output (# features). Compromises between short-term vs. long-term goals, knowledge vs. customer value building etc. Build the right thing (PO) x build it right (devs) x build it fast (SM). Technical debt x sustainable pace. As I said – you MUST see it.
• Martin Fowler: The Value of Software Design (talk, 22 min, from 0:45:00 til 1:07; Feb 2013) – a balanced argument for the value of good software design and internal code quality based on paying off by enabling us to keep our development speed. Discusses the DesignStaminaHypothesis (bad design => rapid decline of development speed), TechnicalDebt, TechnicalDebtQuadrant (Prudent x Reckless, Deliberate x Inadvertent), TradableQualityHypothesis. According to the experience of Fowler and others, the good design payoff point “it’s weeks, not months.”
• What Does It Take To Become A Grandmaster Developer? – great post about cognition and learning, valuable references, quotes from an interesting study of good vs. mediocre developers. We have mental capacity for ~7 chunks of information => great performers recognize patterns and see and understand thus higher-level chunks and have many “chunks” (patterns encountered previously) readily available. You need deliberate effort to learn more chunks – especially initially but you must always try to get out of your comfort zone to grow. Experienced collegues can help a lot in acending the learning curve.

Agile, organization, innovation, project management

• How to Prioritize a User Story Map – we all know that we should prioritize features by their value, risk, and lack of knowledge and that we should slice the features thin so that they fit into short iteration and can be deployed soon to produce feedback, right? Here we see a nice example of what happens if not done so and how to do feature slicing better.
• Bob Marshall: Rightshifting – according to the author, 80% of knowledge work organizations are very ineffective, wasting resources on non-value-adding activites; only few are effective, even fewer highly effective. Rightshifting is the attempt at shiting them to the right, towards higher effectiveness. Links to a few videos explaining it more. Related: Steve McConnell’s Business Case for Better Software Practices, referring to a study by SEI; “The actual distribution of software effectiveness is asymmetric. Most organizations perform much closer to the worst practice than to the best.” – the best performing 10 times better then the worst/average (productivity, speed, defects, value)
• On Antifragility in Systems and Organizational Architecture – introduces the concept of antifragility, based on Nassim Taleb’s book Antifragile that compares fragile, robust, and antifragile systems and organizational structures (which is also applicable to SW systems); robust = resists change (unless too large); antifragile: learn, adapt; closely related to DevOps and continous delivery
• M. Fowler: PurposeOfEstimation – many Agilist disdain estimation, this is a balanced view: “estimation is valuable when it helps you make a significant decision.” (F.ex. when deciding what we (don’t) have resources for or when in need of coordinating related activities.) It is evil when used as commitments that people are forced to stick to and blamed for not managing to do so. “Above all be wary of anyone who tells you they [estimates] are always needed, or never needed.” A. Ferguson: “[..] it is poor project management (whether by project managers or other team members) that results in a client who thinks estimates are fixed, or that raw estimates = actual effort/duration”.
• Ron Jeffries: Estimation is Evil – discusses the problems estimates can cause, issues with requirements gathering up front and their volatility, transparency and politics. Very valuable, highly recommended. See the “favorite quotes” at the bottom of this post. Also contains an interesting lesson learnt from the failed Chrysler C3 project: don’t try to build a grand new system to replace and fix the old one, fix one problem at a time – worth reading for this alone.
• Interview with Steve Blank: Why Big Companies Can’t Innovate – the 2013 list of the world’s 50 most innovative companies has only a few large, established firms (those that have built innovation into its DNA such as Apple and Google). Established companies are less innovative because they focus in their existing business model, have risk-aversion (while there are many failures on the way to a new business model); finally “the people who are best suited to search for new business models and conduct iterative experiments usually are not the same managers who succeed at running existing business units.” – and thus aren’t given the chance. “[..]  the process of starting a new business [..] is fundamentally different from running an existing one. So if you want your company to grow organically, then you need to organize your efforts around these differences.”

Architecture & Ops

• M. Fowler: Schemalessness + NoSQL and Consistency (20 + 20 min) two short, very good, balanced talks about NoSQL. He explains schemalessness and consistency and points out common misunderstanding about them so if you are into NoSQL, watch it.
• What Powers Instagram: Hundreds of Instances, Dozens of Technologies (2012) – interesting high-level overview of the Instagram infrastructure based on AWS and Python (25 XL instances running Django/Gunicorn behind ELB with 3 Nginxes, sharded PostgreSQL with streaming replication on 12 QXL mem instances with software raid and XFS to freeze when snapshoting, media in S3, Redis, Solr for geo-search, Memcached. Gearman for task queues, pyapns for notifications. Munin for monitoring.)
• The Netflix API Optimization Story – how Netflix redesigned its APIs to improve performance, reduce chattiness, and power product development and experimentation. The common REST API has become a development bottleneck and a lowest common denominator solution (w.r.t. supporting various clients). The main changes were: usage Hystrix for fault tolerance, each device team managing their own end-points in any JVM languges (primarily Groovy) and re-using common APIs (i.e. pushing some device-specific code to the server) => able to experiemnt more quickly, using the Functional Reactive Programming Model and asynchronous APIs (to abstract away thread-safety and parallel execution implementation details from the device teams so that code can execute sync. or async. without them needing to know).
• Me: Overview of current monitoring libs for Java – Netflix’ Servo, Yammer’s Metrics, JavaMelody, JavaSimon.
• Debug Servlets, or ‘HTTP Won; Use It’ – expose all debugging info of your services over HTTP – it makes debugging much simpler. We do a part of it and it really helps. Expose config (values, where they come from), logs, log configuration, JMX (setting it up otherwise not trivial), version, build number, git hash, server time (timezones tricky), metrics, stack dumps, app-specific status (Hadoop: live nodes, data size etc.). The author recommends JavaMelody to collect & visualize many common metrics. Not on security: Make sure to hide passwords and make the endpoints visible only internally. (Tip: consider Jolokia for exposing JMX over HTTP, see below.)
• JVM Crash/Core Dump Analysis – 3 common categories of  JVM crash causes (JVM/JIT/JNI) and how to recognize and troubleshoot them

Other

• How to lose wight in the browser:  The definitive front-end performance guide – a site by a number of experts from Twitter, Opera, Google, and other places with best practices for performant web sites (HTML, CSS, JS, jQuery, images). Ex.: styles up top, scripts down bottom; minify your html, css and JS; async script loading; combine css/JS files into one; cache array lengths while looping; use css sprites for icons.
• Luke Stevens: The harsh truth about HTML5′s structural semantics (part 1) – “HTML’s structural elements — article, section, nav and aside — are, at first glance, some of the easiest parts of the HTML5 specification to understand and implement. However, they’re actually some of the most poorly specified, poorly understood, and poorly implemented parts of HTML5.” Interesting: The “research” leading to their establishment was quite random, ignoring a crucial source of information (css IDs).
• Marco Emmanuel Patiño: Six non-technical books every programmer should read – 1. Team Geek: A Software Developer’s Guide to Working Well with Others (-> effective communication and collaboration), 2. The Pragmatic Programmer: From Journeyman to Master, 3. The Passionate Programmer: Creating a Remarkable Career in Software Development, 4. Clean Code: A Handbook of Agile Software Craftsmanship, 5. 97 Things Every Programmer Should Know: Collective Wisdom from the Experts (online), 6. Code Simplicity: The Fundamentals of Software.
  • Related: Top 5 Java programming books – Best of lot (actually 8) – 1) Head First Java, 2) Effective Java, 3) Thinking in Java, 4) Head First Design Pattern, 5) Concurrency Practice in Java, 6)Java performance, 7) Java Puzzlers, 8) Head First Object Oriented Analysis and Design.
• Humans as slaves of chemistry: America’s Real Criminal Element – Lead – a fascinating article about how whole nations can be seriously influenced by a single chemical substance. Aside of that it is also fascinating to observe how we tend to search for causes in our domain of expertise (police, sociologists, …) and of interest while denying other possible causes, no matter how strong are the proofs. If the facts presented are true, then the fivefold increase in serious crimes in (not only) America since 60s has been caused by the increase of lead in the environment (pushing many people over the edge of ocassional violent loss of control). How many social problems in the world have similar industrial causes? Are we careful enough with what we let into our air and bodies?

Languages

• newcoder.io: Learning more Python via projects – an excellent next step when you have learned Python syntax via LPHW or similar; in this tutorial series you will be building real-world apps while learning more of Python. You will play with, Data Visualization, APIs, Web Scraping, Networks, GUI.
• Brian McCallister: Go is PHP for the Backend – a very good explanation why you might want to use Go and that you have to first learn “the Go way” to avoid insanity, since it is very opinionated and different from what you might be used to. Some pros: “native code, UNIX friendly, higher level then C, lower level then Python or Ruby, garbage collected, strongly typed, good performance, good concurrency support, etc.”
• The Neophyte’s Guide to Scala 1 to 15 (list) – a good follow-up on the Cursera FP in Scala course, a series of blog posts exploring some topics more in depth. F.ex.: extracotrs (unapply, for pattern-matching), the broad applicability of pattern matching, pattern matching anonymous functions & partial functions #4, usiong Option idiomaticly #5, nice FP error handling with the Try type #6, Futures, etc. Higly recommended! Thx to Jakob Lind

Libs

• Jolokia is remote JMX with JSON over HTTP: a REST API bridged to JMX, with support for security, fine-grained access control, bulk operations. Especially useful if you either 1)  need to perform bulk operations (e.g. get multiple values) or 2) want to access them from something that doesn’t support JMX. JSON is in general very easy to use and navigate. You can install Jolokia as a WAR (or mebedd its Servlet), a JVM agent, or attach it on-the-fly to a running JVM.
• The Appeal and Controversy of ZeroMQ – why to use 0MQ? It is a messaging library that focuses on performance, decentralization and simplicity, solving some really hard problems (sending async. messages w/o locks, distribuing to specific subscribers) and providing a simple API. Main pros: decentralized (no central broker), many languages; cons: no security (but you can use it over SSH).

Talks

• Tim O’Reilly: Create More Value Than You Capture (30 min + questions) – build apps that matter, that change how we do things. Thinking just about money is bad. Try to make the society better, smart, create more value than you capture, solve important problems, help people. Ex. startups: Uber, Square, Code for America.
• TED: Bruce Feiler: Agile programming — for your family (20 min) – an inspirational talk, based on positive experience from multiple families, about applying the agile thinking and values to make our families happier by empowering the children (enlist them in their upbringing, deciding on goals, rewards, punishments), letting them know who they are, being adaptive, having regular “retrospectives” (that eventually become cherrished memories). Backed by research. Did you know that the #1 wish of children isn’t that parents spend more time with them but that they are less stressed?

Clojure Corner

• What’s new in Clojure 1.5.x – reducers, new threading macros (cond->, as->, some->, ..), various improvements, improved performance, erro messages, doc strings, bug fixes
• Stuart Sierra: On the Perils of Dynamic Scope – summary: don’t create macros like with-connection binding to a thread-local var; make all methods take the resource as a parameter – thus the user has the freedom to decide when to close the resource and isn’t limited to a single thread and can use lazy sequences
• Logic programming is overrated – core.logic is essentially only a complex DSL for doing an exhaustive search and there is already a nice, clean tool for that: the for comprehension. A logic puzzle can be much more clearly and also efficiently using for. But it is not completely useless – logic programming is good e.g. for running programs backwards, unification is important for writing type checkers, and the new constraint programming piece has good potential. Read also Logic Programming is Underrated, which provides a faster core.logic solution than for-comprehension and provides some pointers rgarding the practical usefulness of core.logic.
• Prismatic – Graph: Abstractions for Structured Computation – How to reduce the complexity overhead in large, real-world, FP systems by decoupling what is done from how it is executed. Graph is a Clojure library enabling a declarative way to describe how data flows between (mostly pure) functions => “It allows us to formalize the informal structure of good FP code, and enables higher-order abstractions over these structures that can help stamp out many persistent forms of complexity overhead.” By decoupling the description of how data flows and the actual execution, we can execute it in different ways (parallelized, with memoization, lazy/eager) and apply various interceptors (for logging etc.). See especially the part “Graph and complexity overhead.”
• Mike Anderson: Game development in Clojure : Alchemy 7DRL post-mortem (and the previous 7 daily updates, Alchemy @ GitHub) – an interesting report about game making in Clojure during 7 days, in as functional and immutable style as possible while keeping it sufficiently fast. How do you represent & handle statuf game objects, the world map, game state? The design of the game, what was easy and what hard with Clojure. Tl;dr: search it for “Some parting thoughts” (Clojure productive, immutability hard but pays off, prototype objects great, more typing would have helped). “Making everything immutable in Clojure is harder than it would have been in an OOP language like Java where everything can be encapsulated in mutable classes. In particular, the state update functions are tricky to make both correct and performant. The payoff is big however: in terms of the simplicity and effectiveness later on, and in the conceptual clarity being able to treat the entire game state as an immutable value”. Having REPL is a big win.
• Refactoring Java using Clojure with the Eclipse Java development tools (JDT) (operation on AST nodes, i.e. little too low level; the Eclipse Refactoring API might be better)
• Clojure at a Bank – [Our] Clojure Code Immaturity – experiences with going from Java to Clojure: 1) too few comments, too short names => hard to learn the code; 2) not knowing clojure.core well enough => reimplementing (if-let, juxt, …); 3) structure, comment, split up your namespaces well, navigating more complicated then in Java IDEs; 4) reasonably used Macros, Protocols, Defrecords payed off;
• Datomic for Five Year Olds – explaining the key characteristics of Datomic compared to relational and NoSQL DBs (schema, architecture, programmability/language); doesn’t go into details of how it works (e.g. how does Datomic determine what subset of the DB to cache locally and what if it is few GBs); Honey Badger’s 2012 talk Exploring Datomic: a database deconstructed explores the architecture and technical details much more

Tools

• Vagrant 1.1.0 is out (what’s new?), with support for VMWare Fusion and AWS VM backends in addition to VirtualBox – use the same config to create, provision, stop, destroy and connect to a virtual machine locally or in the cloud (with limited support for shared folders, I’d suppose). V. 1.1 is backwards compatible aside of plugins, upgrade to new config optional.
• Animated presentations: ArtRage (drawing program, also for iPad), Wacom Intuos 5 (drawing tablet), Screenflow (screen & audio capture) – used for the Agile in a Nutshell (Agile Product Ownership in a Nutshell) mentioned above
• ckjm — Chidamber and Kemerer Java Metrics (via Neal Ford) – a command-line tool (also Maven/Ant plugin) to compute some metrics, outputting text or XML for further processing; the metrics: WMC: Weighted methods per class (cyclomatic complexity of its methods), DIT: Depth of Inheritance Tree, NOC: Number of Children, CBO: Coupling between object classes, RFC: Response for a Class, LCOM: Lack of cohesion in methods, NPM: Number of Public Methods, Ca: afferent coupling.
• Bulletproof Demos: Record & Replay built into Chrome – ever got a failure while demonstrating a web app though it has worked moments ago? No more! You Chrome to record your requests and responses and let its cache handle them during the real demonstration. (Mac: stop Chrome, to record run open -a “Google Chrome” –args –record-mode, to replay run open -a “Google Chrome” –args –playback-mode. Linux: google-chrome –record-mode and –playback-mode. Win.: run chrome <arg>)
• UserTesting.com (via Ash Maurya, the author of Running Lean) – on-demand usability testing; they have a large base of test users, can select those matching your criteria and unleash them upon your site guided by a script your provide, watch videos of their actions while they verbalize their thinking process, recieve written answers from them, talk to them.
• MindMup.com – opensource, free mind-mapping in the cloud by Gojko Adzic & co., with main focus on productivity. Store private maps in Goolge Drive, support for mobile devices, keyboard shortcuts. No registration needed.

Favorite Quotes

Once we estimated a project to require 9 man-months but were later told that we do not understand a thing and it may not take more then 4. At the end it took over 25 and still wasn’t done.

- paraphrasing my collegue Kim Leskovski

On collecting requirements up front:

At the very beginning, we know less about our project than we’ll ever know again. This is the worst possible moment to be making firm decisions about what we “require.”

- Ron Jeffries in Estimation is Evil: Overcoming the Estimation Obsession

On the estimate of project delivery date at its initial phase:

It’s based on an unrealistic list of requirements, using weak estimates, made at the moment of maximum ignorance, by people who are always optimistic about their own abilities.
- ibid

On planning and requirements (the Chrysler’s C3 payroll project, having a payroll expert and a team familiar with the domain):

This was one of the best-planned projects I’ve ever seen, and even so, at least one third of the requirements were added, removed, or substantially changed.
- ibid

[..] a line of code is a liability, not an asset [..]
- Jez Humble in Why Software Development Methodologies Suck

Agile is not something you do, it is something you are.
- Huib Schoots in Creating my own flow
with personal kanban, Agile Record Feb 2013

Most interesting links of February ’13

Recommended Readings

• Øyvind Bakksjø: The inverse of IoC is Control – a well-founded criticism of the (over)use of inversion-of-control containers such as Spring and Guice. Bakksjø isn’t against dependency injection in general but he prefers to do it manually, instead of magically, in a main method, using Java instead of an obscure DSL/XML. The dependencies are typically known at compile time so why postpone assembling (and defect discovery) to runtime? Why hide how parts fit together into the non-transparent magic of IoC? He has many good points though I believe there are cases when some automation of the wiring process is valuable (think of scope-dependant beans, diff. deps in diff. environments etc.) Comment by B. Borud: “Explicit wiring you can read from start to finish beats vague declarative shit that you may be able to figure out if you pay close attention.” Check out also the comments at HN.
• J. B. Rainsberger: Demystifying the Dependency Inversion Principle – a very good explanation of the value of dependency injection in the terms of general good code principles (promoting abstractions and postponing determination of details to the latest moment/client)
• CodingHorror: Coding Without Comments (2008) – a very good explanation if how to treat comments (similar what Clean Code says): write code so that it is self-explanatory (using good method and variable names), if necessary, add comments that explain why (while the code shows what/how).
• Frustrations and Aspirations of a Software Craftsman – what makes us happy or unhappy on a project? (I could add some too.) Bad: bureaucracy, old/unfit technology, lack of autonomy and credibility, uninteresting domain, demotivated people, blaming and highly competitive environment, arrogant and unskilled people, … => it is mostly about people. Good: projects where there is passion, craftsmanship, friendship and trust.
• Jetty-9 goes fast with Mechanical Sympathy – interesting how the run-time behavior might differ from what we would expect and how knowing the hardware can improve performance. Here:  false sharing of a blocking queue’s head/tail pointers and locks (close => same CPU cache row => updating one invalidates the other), using trie backed directly by IO buffers for faster String lookups etc. Result (all microbenchmark disclaimers): jetty-9 out performs jetty-8 by 30% faster and 50% less YG garbage.

Czech

• Daniel Kolman: Jak správně pojmenovat test – a good description of different approaches to name and organize tests w.r.t. the experience of the developer – starting with “testWorks1″ through one method-one test with “testMyMethodName” through a more thoughtful naming scheme such as Roy Osherove’s UnitOfWork_StateUnderTest_ExpectedBehavior (e.g. ndexOf_containsSearchedString_returnsCorrectIndex) to tests as a living documentation: tests represent a list of features of the class, each name is a sentence having the class under test as its subject (e.g. [list] holdsItemsInTheOrderTheyWereAdded). Advantages of the latter: readability, documentation, it forces us to think what actually is still the responsability of the class and what isn’t. “Stačí názvům testů věnovat dostatečnou pozornost a ony vám samy napoví, kdy už toho třída dělá moc a měli byste se zamyslet, jak kód lépe rozdělit.”

Non-IT

• Why You Should Work With Someone You Hate – working with people we like and do not have conflicts with is nice but we are likely to have similar views and ideas and miss the broader picture. Working with somebody that drives you crazy while being able to respect each other is very valuable because it opens you to different views and forces you to really defend yours. Recommended!
• Kent Beck: Focusing Talks: Three Questions – some good advices for performing technical talks and avoiding putting too much into the talk. Ask 1) Who is the audience? – be as specific and narrow as possible to be able to tune and cut the content; “Better to reach one person than zero.” 2) What is my one startling sentence? – one that makes the audience take notice; “Picking one sentence, I encourage focus and empathy in myself. Now I have one person’s attention. What do I say?” 3) What would I like the audience to do? “‘To what purpose?’ helps me discard interesting-but-tangential points, to, as Peter Jackson put it, move the ring.” “The more clear and uncompromising I can be with the three questions, though, the easier all the rest of it becomes.” “Talk about less. And less. Less than that.”

Clojure Corner

• Clojure’s Reader is Unsafe (for untrusted strings) – don’t use read-string to read untrusted strings – even if you set *read-eval* to false, it can still be persuaded to execute arbitrary code (hopefully fixed in Clojure 1.5). The right, save way is to use clojure.tools.reader.edn/read-string.

Tools

• Android: Clock to use when giving presentations – presentation countdown apps for Android
• Netflix open-sourced its Chaos Monkey and the whole SymianArmy

The Sprinting Centipede Strategy: How to Improve Software Without Breaking It

Re-published from blog.iterate.no.

Our code has been broken for weeks. Compiler errors, failing tests, incorrect behavior plagued our team. Why? Because we have been struck by a Blind Frog Leap. By doing multiple concurrent changes to a key component in the hope of improving it, we have leaped far away from its ugly but stable and working state into the marshes of brokenness. Our best intentions have brought havoc upon us, something expected to be a few man-days work has paralized us for over a month until the changes were finally reverted (for the time being).

Lessons learned: Avoid Frog Leaps. Follow instead Kent Beck’s strategy of Sprinting Centipede – proceed in small, safe steps, that don’t break the code. Deploy it to production often, preferably daily, to force yourself to really small and really safe changes. Do not change multiple unrelated things at the same time. Don’t assume that you know how the code works. Don’t assume that your intended change is a simple one. Test thoroughly (and don’t trust your test suite overly). Let the computer give you feedback and hard facts about your changes – by running tests, by executing the code, by running the code in production.

Help, My Code Isn’t Testable! Do I Need to Fix the Design?

Our code is often untestable because there is no easy way to “sense¹” the results in a good way and because the code depends on external data/functionality without making it possible to replace or modify these during a test (it’s missing a seam², i.e. a place where the behavior of the code can be changed without modifying the code itself). In such cases the best thing to do is to fix the design to make the code testable instead of trying to write a brittle and slow integration test. Let’s see an example of such code and how to fix it.

Most interesting links of July ’12

A brief one due to (thanks to?) holiday and an accompanying surprising lack of enthusiasm for the technical stuff.

Recommended Readings

• Microdesign and red flags – why using an “else” or returning a boolean as a success indicator should rise a red flag and make you think what you’re trying to achieve and whether there aren’t better options (often there are – using guard conditions instead of else and throwing an exception or returning a detailed status object instead of a boolean)
• Evolutionary architecture and emergent design: Emergent design in the wild – the final article in Neal Ford’s highly interesting series on emergent design, discussing how to recognize the “last responsible moment,” the problems with big up-front designs due to “unknown unknowns” etc., highly recommended

Useful Tools

• Capistrano – Ruby ops tool usable to execute shell commands on multiple servers via SSH (originally intended for deployment of RoR apps). Cap shell enables execution of commands on all/selected servers interactively.
  Quick start: gem install capistrano; echo “role :all, srv1.example.com, srv2,example.com” > Capfile; cap invoke COMMAND=”hostname” SUDO=1; cap shell
  Alternatives: ClusterSSH, Fabric.

Book Review: Implementation Patterns

Implementation Patterns, Kent Beck, 2007, ISBN 0321413091.

Summary: Should you read the book? Yes, the chapter on principles and values is trully enlightening. The book in general contains pearls of wisdom hidden in the mud of “I know that already, man.” I would thus recommend skimming through the book and reading only the pieces matching your level and needs.

The book seems to be targeted a lot at Java beginners (especially the chapter on collections), going into otherwise unnecessary details, yet there are many valuable advises of which some can only be appreciated by somebody with multiple years of professional programming experience. It thus seems to me that the book isn’t a perfect match for anybody but everybody will find there many useful ideas. It would best be split in two.

An experienced developer will already know many of the patterns though it’s perhaps useful to see them named and described explicitly and listed next to each – it helps to be aware and clearer of what you do and why you do it.

I’d absolutely recommend everybody to read the chapter A Theory of Programming, explaining Kent’s style of programming and the underlying key values of communication, simplicity and flexibility as well as the more concrete principles (local consequence, minimize repetition, logic and data together, symmetry, declarative expression, co-locating data and logic having the same rate of change). Also in the rest of the book there are valuable ideas that it would be a pity to miss. I list below some of those that I found particularly interesting.

Most interesting links of June ’12

Recommended Readings

Development

• Neal Ford: Evolutionary architecture and emergent design: Emergent design in the wild – discusses why not to do Big Design Up Front and how to recognize the “last responsible moment”. Quote: “It’s puzzling that so many organizations continue to use BDUF in the face of so many failures and underachieving projects.“
• Gojko Adzic: How To Solve “Not Enough Time” – everybody in IT complains about too much work. The solution acording to Gojko? Kill software already made that isn’t needed (ex.: deleted 70% functionality that wasn’t used; remember that maintenance costs grow ~ exponentially with size/complexity). Kill software in making that won’t be needed (know the value –  effect maps). Kill software that was not successful (if you measure the value of SW, you know whether its existence is worth it). Well written and inspiring. I’m looking forward to killing some SW
• Coding Horror: Postponing decisions to the last responsible moment – a brief and pretty good explanation of this key lean principle together with its connection to YAGNI (and some criticism of LRM by A. Cockburn – I agree with the postscript 2)

Special

• Build The (USS) Enterprise – A fascinating site by an engineer who dares to think big yet manages to stay rooted in reality. A huge inspiration for us all! We tend to think too small. Having a great, inspiring goal is what moves us forward. And people are known to have achieve seemingly impossible things nobody else believed in (how many believed Columbus could reach India by by sailing the opposite direcetion? and even though he actually hasn’t, the impacts of his discovery were tremendous)
• Hooking Users In 3 Steps: An Intro to Habit Testing – to be successful with a new web product in this age of distraction, you need your users to build the habit of using the app regularly (think of Twitter, Facebook, newspaper); this post describes how to find your habitual users, understand them, and optimize the application to turn more occassional users into habitual ones (identify habitual users – learn how they use the app & learn what turns random users into “devotees” – optimize the “habit path”)
• Puppet: Serve configuration from a particular Git branch on demand (original: Git Workflow and Puppet Environments) – at Comoyo we use Puppet to configure all our environments and developers need to test their changes before pushing them to the live environment, preferably without interference from other devs. This post by Adrien of PuppetLabs describes how to enable each developer to have her private branch(es) and have Pupet Master serve the config from the branch on demand, using little puppet environments and hooks magic. (Notice that if using Puppet to serve files then you’ll need to have them inside a module, which is a good practice anyway.)

Other

• DZone reference card for Jetty (DZone login required) – useful, brief overview over Jetty XML and programatic configuration, especially useful is the overview of handlers, use of servlets and webapps, SSL support, websockets
• What’s new in Groovy 2.0? – static type checking, Java 7
• Speed, memory and LOC of Python 3 vs. Java 7 [fixed link] (the computer language benchmark game) – Python tends to take more time (though not terribly more) but is quite economical with memory and visibly more productive (3* less lines of code). Ruby is similar but way slower. Scala is little slower and more memory hunrgy but also more productive. Of course performance is rarely the key factor for picking a language on often it doesn’t matter that much.

Favourite Quotes

Goodhart’s Law: once a metric becomes a target, it loses its meaning as a measure.

by WordFriday

Clojure Corner

• What Sucks about Clojure…and Why You’ll Love It Anyway – 40 min talk by Clojure Programming author Chas Emerick. Some of the negative points: Namespaces – complex (use x refer etc.; no package.* import, …), no forward references; Dynamic scope (with-foo …) has subtle, complex behavior and hard to see => avoid when you can; Using STM (effectively) is hard even though it looks easy, it’s overused, think of it rather as in-memory db with superior integr. with Clojure (ex.: what is the right ref granularity? whole word map? small particle?), nondeterministic (which tx will fail/suceed? – can’t reason) => strange error under high load/mix of operations; JVM: long startup time, …; Macros look easy but are hard (don’t compose nicely with the rest); Function composition – hard to find failure root cause.
• Why is Clojure so slow? (2/2012) – an interesting comparison of Clojure and other languages (C, F#, Groovy etc.) and an analysis of its slow start-up time (metadata and docstring building) and slower performance. (Interestingly, ClojureScript runs it 9* faster.) According to the author, it’s slower partly due to the imutable data structures (nothing new here, we knew we have to offer some performance for the increased safety and robustness). Conclusion: “Clojure is a beautiful, powerful and very useful language, but [...] not great for small script-y programs.” Also, the usefulness of the benchmark is limited if you aren’t writing games. Plans for making Clojure leaner and faster are under way.
• Stuart Sierra: Functional Relational Programming with Cascalog – a brief introduction into Cascalog with a valuable background info about Hadoop, MapReduce, relational programming in Clojure. Good links, especially the paper Out of the Tar Pit looks interesting.

Most interesting links of May ’12

This was a rich month, bringing some hope for ORM, providing a peep-hole into the bright and awesome future with in-browser video and other cool web-stuff presented at WebRebels 2012 and IDEs providing immediate feedback and visualisation. There were valuable articles about simplicity and quality in software and good talks about the lean startup (i.e. enabling innovation) and other topics.

Recommended Readings

• M.Fowler: ORM Hate – Why ORM is actually a good solution – a very valuable article where Fowler opposes the popular trend of criticising Object-Relational Mappers such as Hibernate. Yes, using an ORM is difficult and a leaky abstraction – but that’s because the problem of mapping from a rich in-memory object model to a relational store is inherently difficult (and you need to do it with or w/o an ORM tool) and because those last 10-20% of DB access require human intelligence. If you can avoid the need for ORM by using the relational model also in memory or by using a NoSQL database with a data model that fits your in-memory model then it’s great to do so but often you can’t and then using ORM is the best solution. You certainly don’t want to code your own “lightweight” ORM.
• Alan Quayle on WebRTC, the HTML5 standard for in-browser video/text communication – intro & status – this is an exciting technology coming to our browsers. Some quotes: “WebRTC enables applications such as voice calls, video chat, file sharing, messaging, white-boarding, gaming, human computer interaction, etc. without any client or plug-in download to run from a browser using simple HTML and JavaScript APIs.  Real time communications becomes pervasive on the internet. … Essentially any browser becomes a SIP end point, a telephone, an ‘open’ Skype client, an end point for any real-time communication and control. … Likely by the end of this year we’ll see Chrome and Firefox running WebRTC.“
• Communicating Sequential Processes: Theory for reasoning about concurrent, interacting processes – an inspirational reading about a much better way to do concurrency than Java threads; “… [CSP] is a language for describing patterns of interaction between concurrent objects. It is supported by an elegant, mathematical theory, a set of proof tools, and an extensive literature.” The beauty is that thanks to the theory behind, you can actually reason about the interactions and verify their correctness, contrary to the feared mess of Java threads. CSP is broadly similar to the popular Actors model and is implemented in Occam while it also influenced Erlang’s concurrency model and Go. The library JSCP brings it to Java. I guess we’re better of using Actors due to their popularity and maturity though the mathematical backing of CSP with the potential of formal proofs of correctness is indeed attractive. Any of the two is better than using threads directly because:
  

  The monitor-threads model provided by Java, whilst easy to understand, proves very difficult to apply safely in any system above a modest level of complexity. One problem is that monitor methods are tightly interdependent, so that their semantics compose in non-trivial ways [...]

• Rich Hickey introduces the Reducers library: simplicity in practice – a beautiful example of simplifying something by taking appart all the unrelated but mingled concerns and focus only on those really needed. Whether you’re interested in Clojure or not, you should read the beginning of the post where Hickey explains how the current collection functions based on first (returns 1st element) and rest (returns the remaining ones) mix too many things (ordering, output representation, etc.) and how this “new super-generalized and minimal abstraction for collections” avoids that and thus provides e.g. for doing things in parallel and composing transformation without producing intermediate collections. Beautiful! (PS: I’ve blogged about more examples of pursuing simplicity & gaining power.)
• M. Fowler: Cannot Measure Productivity – a thoughful discussion of why the productivity of programmers is hard/impossible to measure (i.e. you should concentrate on measuring other, more useful metrics) “[..] false measures only make things worse.”
• Gojko Adzic: Redefining software quality – an obligatory read that introduces a holistic view of SW quality and the quality pyramid. The key idea is that there are multiple, vertically organized facets of quality and once a more basic facet is saturated, you should move and and concentrate on the next facet and level of quality. The quality pyramid: Deployable & functionally OK > Performant & secure > Usable > Useful > Successful. Once a particular level is satisfied, it is wasteful to put more effort into it and you’ll bring much more value to the customer by focusing on the next higher level. Gojko: “Yet from what I see most software teams invest, build and test only at the lowest two levels, gold-plating things without a way to explain why that is bad.”
• Is Pair Programming for Me? – the author, who claims to have taught pair programming to 200+500 people, points out that pair programming is a skill that must be (consciously) learned, or actually a number of inter-personal skills. He also describes the cycle people go through when learning it, including a temporary downswing in productivity and negative view of pairing (therefore people should do it at least for 3-4 weeks to overcome the problems and gain the benefits).

Videos

• Bret Victor: Inventing on Principle (55 min, see at least the first 5 min) – very inspiring! Victor firmly believes that “creators need an immediate connection to what they create” and demonstrates how this can be achieve when coding image rendering, a game, an algorithm, when designing a circuit. After watching it for few minutes you will think: How could we have been working with such crappy tools without realizing how limited they are? Fortunately people started to apply the idea of an immediate connection between code and the result, f.ex. in LightTable and Bikeshed‘s IDE. On the other hand, there is an evidence that this may be too hard with the current programming languages.
• Eric Ries: Evangelizing for the Lean Startup – entertaining and enriching introduction into an approach for bringing innovation to life – f.ex. in startups – withou failing unnecessary, demonstrated on the example of the author’s failed and successful startup. Many innovators fail because they don’t realize that their key challenge is that they don’t now neither the problem (who are our customers and what they need) nor the solution (the product to satisfy the need) and thus what they need to do is to experiment and learn in the shortest cycles possible. If you wander what the buzz about lean startup is or how to build innovations, this is the ultimite source you should watch. The video has 1h but the first 20-30 min will give you a sufficient overview. The key points summarized by the Iterate lean guru Anders Haugeto:
  dd

  - There is only one way to measure progress: Progress == The amount of things you have learned from your real customers
  - Hence, you need to work a continuous loop to build, measure and learn as fast as possible. Typical iterations, like sprints, are too long, hence inefficient
  - Until you have an established product, even recognized engineering practices like TDD, sprints, refactoring, and all the XP-stuff are less important than this feedback cycle
  - Even the perfect agile method is nothing, if you’re using it to build the wrong product: How can you know you are heading in the right direction?

• WebRebels 2012 conference talks – I’d especially recommend the awakening talk by Zed Shaw pointing out that we’re building amazing things – but on top of crapy technologies without realizing anymore that the technologies are crappy and could/should be much better. Erlend Oftedal’s talk about webapp security was an (scary) eye-opener for me. If you’re considering offline webapps with HTML5′s webapp cache and/or local storage then you must listen to Jake Archibald’s painful story of various pitfalls hidden there. Christian Johansen’s Pure, functional JavaScript is a pleasure to listen to. Check out the program.

Links to Keep

• E. King: Maximizing the Value of Your Stand-up – interesting techniques to try out at your stand-ups – Speed Scrum, Pass-the-Conch Scrum (passing a token randomly to define the order), Time-Box Scrum, Challenge Scrum (the team may ask 1 question each presenter), Impediments-Only Scrum, Award Scrum (reward for best articulation of his/her information), Business Value-Focused Scrum, No-Board Scrum, Whiteboard Scrum, Buddy Scrum (report for sb. else)

Useful Tools

• puppet-lint – check code style of your Puppet files
• Guard – cross-platform tool that can watch for file changes and execute actions (“guards”) when a file is changed, useful e.g. to execute tests/style checks only on the files being modified. Includes support for many testing/checking tools and multiple notification means such as Growl.
• ThreadLogic – Thread dump analyzer that understands common patterns found in application servers and enabling the definition of custom patterns. Supports Sun, IBM, and JRockit.
• Dumbster – mock SMTP server for unit testing (start in @Before, get sent messages in the test, stop afterwards)

Quotes

Kai Thomas Gilb, in a talk proposal for JavaZone 2012:

Accurate estimation is impossible for complex technical projects, but keeping to agreed budgets, and deadlines is achievable by using feedback and change.

Clojure Corner

• StackOverflow: Clojure Performance Benchmarks – links to various discussions and benchmarks (beware that older results and facts are likely to be outdated). And of course you must keep in mind that 1) benchmark only measure what they measure, e.g. the outcomes cannot be generalized and that 2) benchmark results aren’t relevant for your problem unless you’re doing exactly the kind of operations being benchmarked (e.g. who cares that X is 100 ms slower if your code spends 1s waiting for a XML file download?) (Craig Andera had a pretty good experience report from webapp performance testing including what (not) to do)
• A good wrapup of the EuroClojure conference by Deon Garrett. Such a pity I missed it!
• Goldberg (at GitHub) – Johann Sebastian Bach’s Goldberg Variations in Overtone by @ctford; using Overtone and Clojure to build up mathematical and functional definitions of canons. Deon Garrett: “Go right now and download the code from Chris’ talk. If you don’t know Clojure, use this as an excuse to learn it – it’s that good.”

Rich Hickey interviewed by M. Fogus:

Reducing incidental complexity is a primary focus of Clojure, and you could dig into how it does that in every area.

…

In particular, the use of objects to represent simple informational data is almost criminal in its generation of per-piece-of-information micro-languages, i.e. the class methods, versus far more powerful, declarative, and generic methods like relational algebra. Inventing a class with its own interface to hold a piece of information is like inventing a new language to write every short story. This is anti-reuse, and, I think, results in an explosion of code in typical OO applications.

Bad Code: Too Many Object Conversions Between Application Layers And How to Avoid Them

Have you ever worked with an application where you had to copy data from one object to another and another and so on before you actually could do something with it? Have you ever written code to convert data from XML to a DTO to a Business Object to a JDBC Statement? Again and again for each of the different data types being processed? Then you have encountered an all too common antipattern of many “enterprise” (read “overdesigned”) applications, which we could call The Endless Mapping Death March. Let’s look at an application suffering from this antipattern and how to rewrite it in a much nicer, leaner and easier to maintain form.

The application, The World of Thrilling Fashion (or WTF for short) collects and stores information about newly designed dresses and makes it available via a REST API. Every poor dress has to go through the following conversions before reaching a devoted fashion fan:

1. Parsing from XML into a XML-specific XDress object
2. Processing and conversion to an application-specific Dress object
3. Conversion to a MongoDB’s DBObject so that it can be stored in the DB (as JSON)
4. Conversion from the DBObject back to the Dress object
5. Conversion from Dress to a JSON string

Uff, that’s lot of work! Each of the conversions is coded manually and if we want to extend WTF to provide information also about trendy shoes, we will need to code all of them again. (Plus couple of methods in our MongoDAO, such as getAllShoes and storeShoes.) But we can do much better than that!