Category Archives: Programming

Thinking small and big (in programming)

When programming, thinking small often allows for a quick start but after a while your project slows down as it grows with pain. Based on every such painful experience there are countless of good practices in programming for thinking big. However, thinking big is difficult and comes with overhead, and if you think too big there is a risk you will only think big, and not think about your problem and your actual code very much.

I will give a number of examples of smaller and bigger thinking (don’t get stuck reading the list).

Memory usage
Small thinking: everything fits in RAM
– smaller: everything fits in CPU cache, or CPU registries
Big thinking: using external storage, streaming, compression
– bigger: scaling out

Small thinking: single process and thread
Big thinking: multi threading, sending work to other processes
– bigger: scaling out

Thinking small: portable by standard compliance
– smaller: single platform
Thinking big: target specific tweaks, build and configuration options
– bigger: target specific dependencies (Mysql for Linux, MS SQL for windows)

Source management
Small thinking: versioned tarballs
– smaller: just a single local file
Big thinking: a git/svn repository
– bigger:several repositories, bug tracker, access rights

Small thinking: single standard compile command
– smaller: no building required in the first place
Big thinking: make
– bigger: autoconf, tools and configuration required (Babel)
– even bigger: build a build-and-config-system (like menuconfig for Linux kernel)

Small thinking: assert()
Big thinking: unit tests
– bigger: test driven development, test coverage analysis
– even bigger: continuous integration

Small thinking: command line options
– smaller: hard coded
Big thinking: configuration file
– bigger: configuration (G)UI
– even bigger: download configuration, find out configuration itself, selection of different configurations (like XML-file, JSON-file or database)

Error handling
Small thinking: crash with error message
Big thinking: log file(s), verbose levels
– bigger: error recovery, using system logs (like Windows event log)
– even bigger: monitoring, choice of different external log systems

Thinking small: single CLI or GUI
Thinking big: build a backend library or server that allows for different UIs

Dependencies (code)
Thinking small: only standard library
Thinking big: require libraries (external and own code broken out to libraries)
– bigger: optional dependencies, supporting different libraries that do the same thing
– even bigger: dependencies can be loaded dynamically during run time

Dependencies (databases, services)
Thinking small: no dependencies
Thinking big: external storage
– bigger: allow multiple clients against common storage
– even bigger: distributed, scaled out, storage

Small thinking: functions and data
– smaller: rely on global data
Big thinking: encapsulation (OO-style), immutable data (FP-style)

Small thinking: functions are specific for data
Big thinking: generic functions by templates, interfaces, generators, iterators

Small thinking: code is fast enough
Big thinking: architecture allows scaling out for more performance as required

Small thinking: manual copy-replace is good enough
Big thinking: testing, continuous integration, rollback, zero-downtime

Thinking small: automation not needed since all tasks are simple enough
Thinking big: automation makes complex tasks fast and easy

One size does not fit all
The important thing to understand is that there is no silver bullet. Each program, problem or project has its own requirements and its own sweet spot of small-vs-big. But this my change over time; even if you need to be a bit big later on, it may not help in the beginning.

Perhaps obvious; it is not meaningful to say that a full CI-environment is better than assert() (and you may argue that they are entirely different things). Having global data is not (for all problems) worse than having completely immutable state. And so on.

Your need for big varies within a project: you may need a very big and configurable build process to build something that has very small requirements when it comes to scaling out.

There are no safe default choices
You need to make qualified small-vs-big choices. If you are an experienced programmer you often don’t need to think much about it. If you work within an environment where you already master tools that are available (perhaps even mandatory) you can use them with little overhead. However, if you take your environment (perhaps just an IDE, or more) for granted and you rely on it, it may not be so easy for someone else to pick up where you left.

Just as you can fall behind others who use better tools you can grow fat and fall behind those who use fewer tools (and stay smaller).

If in doubt: start small
Going small in every aspect is often not good enough (except for very isolated problems). But it can be a good start and if (or when) it fails you will learn from your mistake. You will understand how to architect your software better and you will understand why some (big) tools and practices really exist. That is good wisdom!

Going big in every aspect is most definitely going to make you fail. You may need to do this for building systems like SAP or Windows (but such large project often do fail). If you fail with something far too big it is hard to learn from it. Chances are you never really got down to the requirements and chances are much energy was spent integrating tools and frameworks into a development and operation environment that worked at all.

Small sometimes goes a long way
There are often theoretical discussions about small-vs-big. Big often looks attractive and powerful. However, some problems are just hard regardless how you solve them, and a small solution is often more right on target.

There was a macro-kernel vs micro-kernel discussion. A micro kernel is a big solution: more encapsulation, more isolation, less global data, more dynamic loading and so on. Linux is obviously more successful than HURD (the GNU micro kernel), mostly because it actually works.

Agile and Refactoring
Agile and refactoring are about encouraging you to start small, make things that are good enough for now, and fix them later on (if ever needed). Often the problem down the road is not what you expected when you started.

Architecture, Microservices, UNIX
The UNIX principle is that everything is a program that does one thing well.
Microservices is much the same thing except it spans over several networked services.

This works. Because most of the time, for most purposes, the developers (of UNIX and Microservices) can think small. Most programs in a UNIX system, like most services in a Microservice architecture, are for most practical purposes small programs or small services.

UNIX: some programs need to be highly secure, some need an interactive UI, some need to log, some have high performance requirements, some have dynamic dependencies and some are better not written in C. This is why you should build a microservice architecture (not a monolith) and this is how you should build it (unless you are as good as Torvalds and you can land a Monolith in C – but that works thanks to very good architecture and practices – and Linux is still just the Kernel in a much bigger system).

Limited time
All software projects have limited time available. Time is spent on:

  1. Understanding requirements
  2. Producing code that correctly and efficiently matches the requirements
  3. Test and deployment
  4. Solution architecture
  5. Tools and frameworks: understanding and integration

#1 deliver value even on its own (sometimes a technical solution is not even required).
#4 and #5 only deliver value if the they save total time (by lowering #2 and #3).
#2 sometimes is just not possible without #5, then please go ahead with #5.

But if #2 takes one week if you use Notepad to code a single index.html file containing HTML+CSS+JavaScript (and this solves the requirements), then there must be a good case for spending time on #4 and #5 (going big) instead of just solving the problem (staying small).

#4 and #5 produce what I call invented problems; problems that you did not have in the first place, that are not related to your requirements but comes with your tools. The most obvious example is licencing issues. If you go multithreading and/or use an external database you suddenly have deadlocks, race conditions, transactions and semaphores to worry about: is that price worth it for what you get from the database or multithreading? Deployment (and server configuration) is absolutely necessary, often rather complicated, and delivers no value to the customer what so ever.

Always ask yourself: how hard would it be to solve this problem using the smallest reasonable set of tools?

Maintain vs Replace
Many big practices are about producing maintainable code. Often this never pays off:

  • There is no need for the code anymore
  • The code does what it needs to do, and no change is required
  • Even though the code itself is maintainable, no one understands the problem and the solution well enough to actually improve (or even change) it

When (if) the moment of change actually comes, often a fresh start is anyway the best solution. If programs are made small and they do one thing well (so it is quite easy to test that another program can replace it) replacing them is not a big deal.

This means that ugliness (global variables, lack of encapsulation, hard coded limitations, lack of proper test coverage, inability to scale, and so on) often is not a problem. On the other hand, a (big) program that is not fit for purpose (not correct and efficient) never produce much value in the fist place.

Performance (and Scaling)
Golden rule of optimization:

  1. Don’t
  2. Experts only: see (1)

This is not entirely true but most of your code is not performance critical. In computing, there are two ways you can get faster:

  1. Go small: find ways to make your code require less resources
  2. Go big: assign more resources to run your code

The truth is that modern hardware is extremely powerful. Even a Raspberry Pi V1 (with 700MHz CPU and 512MB RAM) can serve enormous amounts of network requests or crunch amazingly many numbers. If a Raspberry Pi is not enough for you, you either have

  1. very many users
  2. a very complicated/large/heavy problem
  3. coded a solution that mostly wastes resources

If you know that #1 (only) is your case, go ahead and scale out big. Be sure to know your bottlenecks and seriously consider your storage model.

If #2 is your case you need to sit down and think.

If #3 is your case, you should have stayed small from the beginning. It is probably cheaper to rewrite significant parts of your solution in C (or another language that uses minimal resources) and keeping all data in RAM, than it is to scale your code out.

Availability (and Redundancy)
You may need high availability: downtime, unexpected or not, is expensive.

The big solution is to go for redundancy: if one goes down the other takes over. This can be the right thing to do – when everything else has already been tried. Sometimes the cure is worse than the disease.

The small solution is to keep your program simple and when something unexpected happens, let it crash. This way you will quite soon (pre production) have nailed out the critical errors. And if you can not make it stable no redundancy or fault-tolerant environment will really save you.

When going big understand the cost. The road to hell is sided by good intentions. Beware of grande architectures.

Lodash Performance Sucks!

To continue my Functional Programming Sucks series of posts I will have a closer look at reduce().

I complained with Lodash (and Underscore) for different reasons. One complaint was performance, but I just read the code and presumed it was going to be slow without measuring. Then I complained with the performance of Functional Programming in general.

I thought it would be interesting to “improve” the Functional code with Lodash functions, and to my surprise (I admit I was both wrong and surprised) I found Lodash made it faster! After reading a little more about it I discovered this is a well known fact.

So, here are four different implementations of a function that checks if the elements (numbers) in an array are ordered (cnt is incremented if the array is sorted, such was the original problem).

// Standard reduce()
    this.test = function(p) {
        if ( false !== p.reduce(function(acc,val) {
            if ( false === acc || val < acc ) return false;
            return val;
        }, -1)) cnt++;

// Lodash reduce(), and some other Lodash waste
    this.test = function(p) {
        if ( false !== LO.reduce(p,function(acc,val) {
            if ( false === acc || val < acc ) return false;
    //      if ( !LO.isNumber(acc) || val < acc ) return false;
            return val;
        }, -1)) cnt++;

// My own 4 minute to implement simpleReduce(), see below
    this.test = function(p) {
        if ( false !== simpleReduce(p,function(acc,val) {
            if ( false === acc || val < acc ) return false;
            return val;
        }, -1)) cnt++;

// A simple imperative version
    this.test = function(p) {
        var i;
        for ( i=1 ; i < p.length ; i++ ) {
            if ( p[i] < p[i-1] ) return;

// my own implementation reduce()
    function simpleReduce(array, func, initval) {
         var i;
         var v = initval;
         for ( i=0 ; i<array.length ; i++ ) {
             v = func(v, array[i]);
         return v;

The interesting thing here is that the standard library reduce() is the slowest.
However, my simpleReduce is faster than Lodash reduce().

(seconds) reduce()
Std Lib Lodash Simple Imperative
Raspberry Pi v1 (ARMv6 @ 700) 21 13 9.3 4.8
MacBook Air (Core i5 @ 1400) 0.46 0.23 0.19 0.16

The conclusion is that from a performance perspective Functional Programming sucks. Lodash sucks too, but a little bit less so than the standard library (however, if you decorate all your code with isEmpty, isString, isNumber and that crap it will get worse).

That said, the generic nature of Lodash comes at a cost. The most simpleReduce() imaginable outperforms Lodash. As I see it, this leaves Lodash in a pretty bad (or small) place:

  • Compared to the standard library it is an extra dependency with limited performance benefits
  • The generic nature of Lodash comes at both a performance cost and it allows for sloppy coding
  • A hand written reduce() outperforms Lodash and is a good excercise for anyone to write. I expect this is quite true also for other functions like take or takeRight.
  • For best performance, avoid Functional Programming (and in this case the imperative version is arguably more readable than the FP reduce() versions)

Whats up with the Standard Library???
JavaScript is a scripted language (interpreted with a JIT compiler) that has a standard library written in C++. How can anything written in JavaScript execute faster than anything in the standard library that does the same thing?

First, kudos to the JIT designers! Amazing job! Perhaps the standard library people can learn from you?

I can imagine the standard library functions are doing some tests or validations that are somehow required by the standard, and that a faster and less strict version of reduce() would possibly break existing code (although this sounds far fetched).

I can (almost not) imagine that there is a cost of going from JS to Native and back to JS: that function calls to native code comes with overhead. Like going from user space to kernel space. It sounds strange.

I have read that there are optimizations techniques applied to Lodash (like lazy evaluation), but I certainly didn’t do anything like that in my simpleReduce().

For Node.js optimizing the standard library truly would make sense. In the standard library native code of a single-threaded server application every cycle counts.

UPDATE: I tried replacing parts of the above code: 1) the lambda function that is passed to reduce(), 2) the imperative version, with native code. That is, I wrote C++ code for V8 and used it instead of JavaScript code. In both cases this was slower! Obviously there is some overhead in going between native and JavaScript JIT, and for rather small functions this overhead makes C++ “slower” than JavaScript. My idea was to write a C++ reduce() function but I think the two functions I wrote are enough to show what is happening here. Conclusion: don’t write small native C++ functions for performance, and for maximum performance it can be worth to rewrite the standard library in JavaScript (although this is insane to do)!

All FP-sucks related articles
Functional Programming Sucks)
Underscore.js sucks! Lodash sucks!
Functional Programming Sucks! (it is slow)
Lodash Performance Sucks! (this one)

Functional Programming Sucks! (it is slow)

Update 2017-07-17: Below i present numbers showing that functional code is slower than imperative code. It seems this has changed with newer versions of Node.js: functional code has not turned faster but imperative code has become slower. You can read a little more about it in the comments. I will look more into this. Keep in mind that the below findings may be more accurate for Node.js v4-6 than for v8.

Functional programming is very popular with contemporary JavaScript programmers. As I have written before, Functional programming sucks and functional libraries for JavaScript also suck.

In this post I will explain more why Functional Programming sucks. I will start with the conclusion. Read on as long as you want more details.

Functional Programming practices are bad for performance
It is very popular to feed lamda-functions to map(), reduce(), filter() and others. If you do this carelessly the performance loss is significant.

It is also popular to work with immutable data. That is, you avoid functions that change (mutate) current state (side effects) and instead you produce a new state (a pure function). This puts a lot of pressure on the garbage collector and it can destroy performance.

The Benchmark Problem
Sometimes I entertain myself solving problems on I particularly like the mathematical challenges in the Project Euler section (the Project Euler is also an independent organisation – HackerRank uses the challenges in Project Euler to create programming challenges).

This article refers to Project Euler 32. I will not go into details, but the solution is basically:

  1. Generate all permutations of the numbers 1,2,3,4,5,6,7,8,9 (there are 9! of them)
  2. For each permutation, check if it is “good” (very few are)
  3. Print the sum of the good instances

The first two steps give good benchmark problems. I have made different implementations of (1) and (2) and then compared the results.

Benchmark Results
I have three different permutation generators (all recursive functions):

  1. Pure function, immutable data (it may not be strictly pure)
  2. Function that mutates its own internal state, but not its input
  3. Function that mutates shared data (no allocation/garbace collection)

I also have three different test functions:

  1. Tests the orginal Project Euler problem
  2. Simplified test using reduce() and lamda function
  3. Simplified test implemented a standard loop

I benchmarked on two different systems using Node.js version 6. I have written elsewhere that Node.js performance on Raspberry Pi sucks.

(seconds) Project Euler Test Simplified Test
Test Function: Functional Imperative
Permutation Gen: Pure Semi Shared Shared Shared Pure
Raspberry Pi v1 (ARMv6 @ 700) 69 23 7.4 21 3.7 62
MacBook Air (Core i5 @ 1400) 0.77 0.29 0.13 0.40 0.11 0.74

Comparing columns 1-2-3 shows the performance of different generators (for Project Euler test)
Comparing columns 4-5 shows the performance of two different test functions (using fast generator)
Comparing columns 5-6 shows the performance of two different generators (for fast simple test)

This shows that the benefit of using shared/mutable data (not running the garbage collector) instead of immutable data is 5x performance on the Intel CPU and even more on the ARM. Also, the cost of using reduce() with a lamda function is more than 3x overall performance on the Intel CPU, and even more on the ARM.

For both the test function and permutation generation, making any of them functional-slow significantly slows down the entire program.

The conclusion of this is that unless you are quite sure your code will never be performance critical you should avoid functional programming practices. It is a lot easier to write imperative code than to later scale out your architecture when your code does not perform.

However, the pure immutable implementation of the permutation generator is arguably much simpler than the iterative (faster) counterpart. When you look at the code you may decide that the performance penalty is acceptable to you. When it comes to the reduce() with a lamda function, I think the imperative version is easier to read (and much faster).

Please notice that if your code consists of nice testable, replaceble parts without side effects you can optimize later on. The functional principles are more valuable at a higher level. If you define your functions in a way that they behave like nice FP functions it does not matter if they are implemented using imperative principles (for performance).

Generating Permutations
I used the following simple method for generating permutations. I start with two arrays and I send them to my permute-function:

  head = [];
  tail = [1,2,3,4];


My permute-function checks if tail is empty, and then: test/evalute head.
Otherwise it generates 4 (one for each element in tail) new sets of head and tail:

  permute( [1] , [2,3,4] )
  permute( [2] , [1,3,4] )
  permute( [3] , [1,2,4] )
  permute( [4] , [1,2,3] )

The difference in implementation is:

  • Pure version generates all the above 8 arrays as new arrays using standard array functions
  • Semi pure version generates its own 2 arrays (head and tail) and then uses a standard loop to change the values of the arrays between the (recursive) calls to permute.
  • Shared version simply creates a single head-array and 9 tail-arrays (one for each recursion step) up front. It then reuses these arrays throughout the 9! iterations. (It is not global variables, they are hidden and private to the permutation generator)

The simplified test
The simplified test checks if the array is sorted: [1,2,3,4]. Of all permutations, there is always exactly one that is sorted. It is a simple test to implement (especially with a loop).

// These functions are part of a "test-class" starting like:
function testorder1() {
    var cnt = 0;

// Functional test
    this.test = function(p) {
        if ( false !== p.reduce(function(acc,val) {
            if ( false === acc || val < acc ) return false;
            return val;
        }, -1)) cnt++;

// Iterative test (much faster)
    this.test = function(p) {
        var i;
        for ( i=1 ; i<p.length ; i++ ) {
            if ( p[i] < p[i-1] ) return;

I tried to optimise the functional reduce() version by breaking out a named function. That did not help. I also tried to let the function always return the same type (now it returns false OR a number) but that also made no difference at all.

All the code
For those who want to run this themselves or compare the permutation functions here is the entire program.

As mentioned above, the slowest (immutable data) permutation function is a lot smaller and easier to understand then the fastest (shared data) implementation.

'use strict';


function arrayToNum(p, s, e) {
    var r = 0;
    var m = 1;
    var i;
    for ( i=e-1 ; s<=i ; i-- ) {
        r += m * p[i];
        m *= 10;
    return r;

function arrayWithZeros(n) {
    var i;
    var a = new Array(n);
    for ( i=0 ; i<a.length ; i++ ) a[i] = 0;
    return a;


function permutations0(n, callback) {


function permutations1(n, callback) {
    var i;
    var numbers = [];
    for ( i=1 ; i<=n ; i++ ) numbers.push(i);

function permute1(head, tail, callback) {
    if ( 0 === tail.length ) {

    tail.forEach(function(t, i, a) {
        permute1( [t].concat(head),



function permutations2(n, callback) {
    var i;
    var numbers = [];
    for ( i=1 ; i<=n ; i++ ) numbers.push(i);

function permute2(head, tail, callback) {
    if ( 0 === tail.length ) {
    var h2 = [tail[0]].concat(head);
    var t2 = tail.slice(1);
    var i  = 0;
    var tmp;
    while (true) {
        permute2(h2, t2, callback);
        if ( i === t2.length ) return;
        tmp   = h2[0];
        h2[0] = t2[i];
        t2[i] = tmp;


function permutations3(n, callback) {
    var i;
    var head  = arrayWithZeros(n);
    var tails = new Array(n+1);

    for ( i=1 ; i<=n ; i++ ) {
        tails[i] = arrayWithZeros(i);

    for ( i=1 ; i<=n ; i++ ) {
        tails[n][i-1] = i;

    function permute3(x) {
        var j;
        var tail_this;
        var tail_next;
        var tmp;
        if ( 0 === x ) {
        tail_this = tails[x];
        tail_next = tails[x-1];

        for ( j=1 ; j<x ; j++ ) {
            tail_next[j-1] = tail_this[j];

        while ( true ) {
            head[x-1] = tail_this[j];
            if ( j === x ) return;

            tmp            = head[x-1];
            head[x-1]      = tail_next[j-1];
            tail_next[j-1] = tmp;



function testprint() {
    this.test = function(p) {

    this.done = function() {
        return 'Done';


function testorder1() {
    var cnt = 0;

    this.test = function(p) {
        if ( false !== p.reduce(function(acc,val) {
            if ( false === acc || val < acc ) return false;
            return val;
        }, -1)) cnt++;

    this.done = function() {
        return cnt;


function testorder2() {
    var cnt = 0;

    this.test = function(p) {
        var i;
        for ( i=1 ; i<p.length ; i++ ) {
            if ( p[i] < p[i-1] ) return;

    this.done = function() {
        return cnt;


function testeuler() {
    var sums = {};

    this.test = function(p) {
        var w1, w2, w;
        var m1, m2, mx;
        w =  Math.floor(p.length/2);
        w1 = 1;
        w2 = p.length - w - w1;
        while ( w1 <= w2 ) {
            m1 = arrayToNum(p,     0, w1      );
            m2 = arrayToNum(p,    w1, w1+w2   );
            mx = arrayToNum(p, w1+w2, p.length);
            if ( m1 < m2 && m1 * m2 === mx ) {
                sums['' + mx] = true;

    this.done = function() {
        var i;
        var r = 0;
        for ( i in sums ) {
            r += +i;
        return r;


function processData(input, parg, targ) {
    var r;

    var test = null;
    var perm = null;

    switch ( parg ) {
    case '0':
        perm = permutations0;
    case '1':
        perm = permutations1;
    case '2':
        perm = permutations2;
    case '3':
        perm = permutations3;

    switch ( targ ) {
    case 'E':
        test = new testeuler;
    case 'O1':
        test = new testorder1;
    case 'O2':
        test = new testorder2;
    case 'P':
        test = new testprint();

    r = perm(+input, test.test);

function main() {
    var input = '';
    var parg = '1';
    var targ = 'E';
    var i;

    for ( i=2 ; i<process.argv.length ; i++ ) {
        switch ( process.argv[i] ) {
        case '0':
        case '1':
        case '2':
        case '3':
            parg = process.argv[i];
        case 'E':
        case 'O1':
        case 'O2':
        case 'P':
            targ = process.argv[i];

    process.stdin.on('data', function (s) {
        input += s;

    process.stdin.on('end', function () {
       processData(input, parg, targ);


This is how I run the code (use a lower value than 9 to have fewer than 9! permutations)

### Project Euler Test: 3 different permutation generators ###
$ echo 9 | time node projecteuler32.js 3 E
8.95user ...
b$ echo 9 | time node projecteuler32.js 2 E
25.03user ...
$ echo 9 | time node projecteuler32.js 1 E
70.34user ...

### Simple check-order test, two different versions. Fastest permutations.
b$ echo 9 | time node projecteuler32.js 3 O1
23.71user ...
$ echo 9 | time node projecteuler32.js 3 O2
4.72user ...

(the timings here may not exactly match the above figures)

All FP-sucks related articles
Functional Programming Sucks)
Underscore.js sucks! Lodash sucks!
Functional Programming Sucks! (it is slow) (this one)
Lodash Performance Sucks!

Underscore.js sucks! Lodash sucks!

In a world of functional programming hype there are two very popular JavaScript frameworks: underscore.js and Lodash. Dont use them! It is a horrible idea. They suck just like functional programming sucks!

They make claims like:

  • Lodash: A modern JavaScript utility library delivering […] performance
  • Underscore: JavaScript library that provides a whole mess of useful functional programming helpers.

The road to hell is sided by good intentions. This is how it goes.

1. Sloppy data types
There are many good things about JavaScript. The sloppy dynamic typing is perhaps not one of them. The following are for example true:

  • ’27’ == 27
  • undefined == null
  • 0 == ”
  • ‘object’ === typeof null

Now that I consider myself an experienced programmer I find it quite convenient to not need to be explicit about data types. But I dont mix and change types! If a variable is a number from the beginning it keeps being a number. I carefully pick types: Objects, Arrays, Number, String and stick to that type (for a variable or property). Occationally – mostly for return variables – I break the rule.

Lodash and Underscore is about allowing yourself to be sloppy:

  • Dont know if its an object or an array: use map, foreach, filter, reduce and many more
  • Dont know if it is empty (or what being empty even means): use isEmpty
  • Dont know if it is String Object or a String primitive or something else: use isString

If you dont know what it is is you already have a much bigger problem than how to do something with it.
If you mix String Objects and String primitives AND other things, and you want to know if it is any kind of string you are doing something wrong.

So Step 1 with Lodash and Underscore is that you

  1. Add a depenceny
  2. Allow sloppy and inconsistent typing
  3. No one can now presume anything about your types anymore
  4. Your code is now impossible to maintain or extend without lodash or underscore

2. Performance!
My experience after many years in software development is that when an application is not well received by the users it is very often because of (bad) performance. And bad performance causes weird, hard to reproduce, bugs and instability as well.

An important type of optimization that the JIT can do relies on the runtime generating classes with strict types for your objects (it guesses and learns the types as the program runs). If you allow a property to assume values of different types you are likely to destroy this optimization.

Lets look at the little function isEmpty.

/** Underscore **/
_.isEmpty = function(obj) {
    if (obj == null) return true;
    if (isArrayLike(obj) && (_.isArray(obj) || _.isString(obj) || _.isArguments(obj))) return obj.length === 0;
    return _.keys(obj).length === 0;

/** Lodash **/
function isEmpty(value) {
    if (isArrayLike(value) &&
        (isArray(value) || isString(value) ||
         isFunction(value.splice) || isArguments(value))) {
        return !value.length;
    return !nativeKeys(value).length;

If you KNOW the datatype you can just do:

/** String **/
if ( 0 === s.length )

/** String that may be null **/
if ( null === s || 0 === s.length )

/** Array **/
if ( 0 === a.length )

/** Object **/
function objectIsEmpty(o) {
    for ( x in o ) return false;
    return true;

(you may want to check o.hasOwnProperty(x) depending on what you actually want – but if you dont know what you want using Lodash or Underscore will produce equally unexpected results as my code)

The worst thing with the Underscore and Loadash implementations are the last lines:

    return _.keys(obj).length === 0;
    return !nativeKeys(value).length;

Unless the JIT compiler and runtime is very smart those two will produce an entire array of strings on the heap – just to check if the array length is 0! Even though this in most practical cases will have an acceptable overhead it can get very expensive.

This IS the beauty of FP generic programming. Focus on WHAT, not HOW, and any little innocent check (like isEmpty) can turn horribly expensive.

However, to be honest, I took some plain functional code and replaced plain JavaScript with Lodash functions (forEach, reduce, isNumber, isEmpty and a few more) and the code actually got faster! Still much slower than imperative code, but slightly faster than not using Lodash.

3. Complexity and Optimization
Now that you have added an extra dependency, made your data objects sloppy, made your application harder for the JIT to optimize, perhaps your application is not as quick as you need it to be. If you are writing a front end you are probably quite fine. But if you are coding a Node.js backend, performance matters a lot more and waste is more unacceptable. If you are really unlucky these sloppy types give you hard to find bugs and your backend is not completely stable and reliable.

What do you do? Common practices in the business could be things like:

  • Minification/uglification
  • Scale out service architecture
  • Failover technology
  • Spend time optimizing code and modules

This is how little sloppyness, laziness and convenience when making initial decisions about your architecture later can cause you huge problems and costs.

Of course, just including lodash and using isEmpty() in a few places is not going to do much (if any) harm.
But finding lodash or underscore generally preferable to not using them is one kind of low quality thinking that causes software to be bad.

Be explicit, careful, consistent and smart about the data types you use.
Be restrictive with libraries and frameworks that introduce overhead and hide relevant details.

Use the standard library. However, you can find that for example the Array functions of Lodash outperform the standard library equivalents. That may not be true in the future (and I really wonder how it can happen at all).

All FP-sucks related articles
Functional Programming Sucks)
Underscore.js sucks! Lodash sucks! (this one)
Functional Programming Sucks! (it is slow)
Lodash Performance Sucks!

Functional Programming Sucks*

(*like everything else that is mindlessly applied the wrong way to solve the wrong problems)

Functional Programming Rocks!
You can do amazing things with Haskell. The historical importance of Lisp is huge. Math, computer science, algorithms and engineering can come together beautifully with functional programming. Writing small reusable, testable functions with no side effects is a great thing to do in any language – but more than anywhere else those virtues are emphasized in functional programming. I love first class functions!


The unfortunate JavaScript Hype
As JavaScript made map, filter and reduce part of the standard those have become the foremost frontier of functional programming. Being quite serious, many people argue that if you replace your for-loops with map, filter, reduce and forEach you have achieved something significant when it comes to readability. What do you think?

// non-functional smelly loop
redFruits = [];
for ( f in fruits ) {
  if ( 'red' === fruits[f].color ) redFruits.push(fruits[f]);

// fantastic functional alternative
greenFruits = fruits.filter(function(f) {
  return 'green' === f.color;

As a bonus with functional, you can use lamda-functions… that is functions, with no name, that can not be reused. You can even use arrow-functions if you think this is too easy a concept and you think the code gets more readable by omitting the confusing word function.

While there are a lot of balanced articles about using functional ideas in JavaScript, for a lot of people eliminating for-loops at any cost seems to be functional enough.

The argument here is that your code should be declarative, it should express intention rather than instructions. In the case above with the fruit-filter it does make some sense. But if that argument should make any sense, it requires that the programmer does not abuse map, filter, forEach or reduce to eliminate a for-loop just for the sake of it. I have seen things like:

// functional
applePrice = Object.keys(fruits).filter(function(f) {
  return 'apple' ===;

// when this would work
for ( f in fruits ) {
  if ( 'apple' === fruits[f].name ) {
    applePrice = fruits[f].price;

I have several objections with the functional version. The Object.keys thing is hardly pleasant to the eye. It is false marketing to use filter: the intention is find or search, not filter. So you still need to read the details (just as with the for-loop), you are just first fooled into thinking its a filter (and chaining is very popular, so then you have no function names and no variable names). But perhaps the worst problem is the lack of error handling. Functional code is not meant to have side effects, and error handling is exactly that. If ‘apple’ is not found you get an ugly Error. You can of course try/catch, or make temporary array variable apples and check that its length is one, but people who prefer functional style usually dont do it.

I understand your objection: people can write crappy code with any language an paradigm and just becuase I have seen bad applications of filter does not mean there is anything wrong with filter or FP. Of course. The bad thing is recommending people to use it like a silver bullet. The bad thing is that good FP is actually difficult and junior programmers will get it wrong trying to be fashionable.

Here is another example to show I am not inventing this rant.

Functional is preferable-Hype
Another favorite example of this functional hype is from rosettacode. Look at this amazing collection of implementations of a simple algorithm: the Luhn Algorithm. I suggest:

  1. Read the introduction so you get some idea
  2. Look at the C-implementation: imperative and simple. Testable and no side effects.
  3. Look at the functional implementations: C++11, Common Lisp, Haskell, JavaScript (ES5.1 version), PicoLisp, Python, Rust, Scheme

Look at Scala: there are two versions, a Functional Style (recommended) and an Imperative style. The people att IOCCC would be jealous with this shit!

I can only come to one conclusion: the emperor is naked.

I mean, if you code in PicoLisp, for fun or for a good reason, and you have to implement the Luhn algorith, please do so! But to say “recommended” about the functional Scala code or to think that the C++11 code is in anyway more reasonable than the original C-code… it makes no sense. No real systems programmers will choose Rust over C based on this. And Python – a language known for its clarity… such a sad example.

Trendy fashionable “functional programmers” suck
So, the problem here is not primarily with Function Programming itself or the code that guru coders write with Functional Programming. The problem is that people with far too little theoretical background, training and experience follow the hype and the result is ugly.

Functional Programming Sucks (at)
There are some things that Functional Programming is not very good at: State and Side Effects. This is by design. From a FP perspective (and a general perspective as well) State and Side Effects are nasty and problematic: they should be avoided and when they cant be avoided they need special attention (like Monads).

The problem here is that JavaScript is mostly a programming language for the web. A simple (modern, single page) web application:

  1. Loads data from a server
  2. Presents data to the user
  3. Lets the user update the data
  4. Sends the data back to the server

This is (almost) all about state and side effects! You may have some functions for validation, filtering, sorting and calculations on the client, and those benefit from Functional Programming ideas. But your core enterprise is about state, side effects and error handling! Functional Programming is so unsuitable for this that a simple web page just can’t be written functionally, unless you start breaking rules. And this is of course what programmers end up doing because in the end of the day they have a real application to build for real users.

The difficult thing is to break the right rules in the right way at the right place for the right reason. This is architecture – mixing concepts, and designing how your application lives, its breaths and heartbeats. Architecture is difficult, and it is not made easier relying on unsuitable silver bullets.

Functional Reactive Programming
There is one thing that is even more hyped and even more theoretic than Functional Programming, and that is Functional Reactive Programming.

But it makes sense (as I understand it). It is not about taking Functional Programming one step further but about putting it in context. You have data that changes (signals, events, behaviours, whatever). That data can be pipelined with high quality through FP logic to produce the content of your GUI or the requests to the server.

Keep the functional parts functional, and let other parts of your application just deal with I/O, GUI and state. Dividing your code into separate modules with clear responsibilites and clear interactions have always been a good idea.

My experience is that when a real world application is not well received by the users a lot of the time its because performance sucks. When performance is bad usability is also bad, and stability gets bad (especially on the server side).

Functional programming is typically bad for performance.

  • anonymous functions can often not be JIT compiled and will never be optimized
  • recursion is nice, but rarely faster than a loop
  • the chaining concept creates arrays and arrays and arrays just for throwing away, which is fun for the garbage collector, but it is not fast
  • the immutable object concept is about generating new RO copies of object instead of changing objects, which is expensive and wasteful

Perhaps worse, since functional programming and proper use of and map(), filter(), reduce() and friends is hard (easy things get difficult) not so experienced programmers end up writing implementations with unnecessary computational complexity ( O(n) turns into O(n^2) ). It is not funny – you cant afford that for most anything that goes to production.

Agile, refactoring and generic code
It is HARD to design code properly from the beginning! Good objects, classes, functions, modules, packages, namings, dependency trees and architecture dont come for free! Agile and Refactoring are about accepting that the design will not be optimal the first time, thus we should not bother too much about it, but rather fix the code when we have learnt more about the problem and our code starts getting (too) ugly.

A strong argument for FP is that it is highly generic – which is true. But until a programmer has spent much time with her domain and problem she will not know what things can and should be made generic. Making things too generic is called overengineering, and it is perhaps the worst sickness in our industry.

I usually:

  • start with one source file rather than many
  • allow myself some copy-paste until I see what code really gets repeated
  • make my code as specific as possible, unless I see an obvious generalisation or the actual need for generalisation emerges
  • dont worry too much about global variables in the beginning (after a while there will be a natural place for them or for what the represent)
  • allow quite long functions until I see what parts of them actually do something meaningful on its own
  • code quite defensively with lots of error handling (it usually pays of quite quickly)

This works for getting quick practical results. Later, during refactoring, when the code base has grown, and when I have learnt more about the domain, I can break out pieces of critical code that creates nice generic functions. But thinking FP first – no way!

All FP-sucks related articles
Functional Programming Sucks (this one)
Underscore.js sucks! Lodash sucks!
Functional Programming Sucks! (it is slow)
Lodash Performance Sucks!

Sort strings without case sensitivity

In JavaScript, I wanted to sort arrays of strings without caring about case. It was more complicated than I first thought.

The background is that I present lists like this in a GUI:

  • AMD
  • Apple
  • Gigabyte
  • IBM
  • Intel
  • Microsoft
  • MSI
  • Nokia
  • Samsung
  • Sony

I want AMD and MSI (spelled in all caps) to be sorted without respect to case. Standard sort() would put MSI before Microsoft.

Obviously I am not the first one wanting to do this and I found an article on stackoverflow. It suggests the following solution:

Use toLowerCase()
You can make your own string compare function that uses toLowerCase and send it as an argument to sort():

function cmpCaseless(a,b) {
    a = a.toLowerCase();
    b = b.toLowerCase();
    if ( a < b ) return -1;
    if ( a > b ) return  1;
    return 0;


This has a number of problems. The article above mentions that it is not stable. That is probably true in some cases but I was of course worried about performance: making two String objects for each compare should make the garbage collector quite busy, not to mention the waste of copying and lowercasing potentially quite long stings when usually the first character is enought. When I started experimenting I found another more critical flaw though: in Swedish we have three extra characters in the alphabet; Å,Ä,Ö, in that order. The above cmpCaseless orders Ä,Å,Ö, which sounds like a little problem, but it is simply unacceptable.

Use localeCompare
There is a more competent (or so I thought, read on) way to compare strings in JavaScript: the localeCompare function. This one simply treats A,Å,Ä and O,Ö as the same character, which is far more unacceptable than the toLowerCase problem.

However, it also has a “locales” option (a second optional argument). If I set it to ‘sv’ I get the sort order that I want, but performance is horrible. And I still have to use toLowerCase as well as localeCompare:

function localeCompare(a,b) {
    return a.toLowerCase().localeCompare(b.toLowerCase());

function localeCompare_sv(a,b) {
    return a.toLowerCase().localeCompare(b.toLowerCase(), 'sv');

localeCompare() has an extra options argument with a “sensitivity” parameter, but it is no good for our purpuses.

Rolling my own
Of course, I ended up building my own function to do caseless string compare. The strategy is to compare one character at a time, not making any new String objects, and fallback to localeCompare if both characters are above the 127 ASCII characters:

function custom(a,b) {
    var i, al, bl, l;
    var ac, bc;
    al = a.length;
    bl = b.length;
    l = al < bl ? al : bl;
    for ( i=0 ; i<l ; i++ ) {
        ac = a.codePointAt(i);  // or charCodeAt() for better compability
        bc = b.codePointAt(i);
        if ( 64 < ac && ac < 91 ) ac += 32;
        if ( 64 < bc && bc < 91 ) bc += 32;
        if ( ac !== bc ) { 
            if ( 127 < ac && 127 < bc ) {
                ac = a.substr(i,1).toLowerCase();
                bc = b.substr(i,1).toLowerCase();
                if ( ac !== bc ) return ac.localeCompare(bc);
            } else {
                return ac-bc;
    return al-bl;

One fascinating thing is that here I can use localeCompare() without 'sv'.

Test for yourself
I built a simple webpage where you can test everything yourself.

Defining a string sort order is not trivial, when you dont just have ASCII characters. If you look at the ascii table you see that non alphabetic characters are spread out:

  • SPACE, #, 1-9, and many more come before both A-Z and a-z
  • Underscore: _, and a few other characters come after A-Z but before a-z
  • Pipe: | and a few other characters come after A-Z and a-z

When it comes to characters behind ASCII 127, it just gets more complicated: how do you sort european language latin letters, greek letters and arrows and other symbols?

For this reason, I think it makes sense to define your own sorting function and clearly define the behaviour for the characters you know that you care about. If it really matters in your application.

My function above is significantly faster than the options.

These results can probably be inconsistent over different web browsers.

Angular.js Hello World program

Usually when writing a small utility I use the command line. But sometimes a simple web page with JavaScript gives a more functional UI.

I like AngularJS, but I don’t start from scratch very often, and when I do, I need to start searching for code to copy. So, here is my Angular Hello World template.

<!DOCTYPE html>
<html ng-app="theApplication" ng-controller="theController">
    <meta charset="utf-8">
    <title>Hello World</title>

  <script src="angular.js"></script>
    angular.module('theApplication', []).
    controller('theController', ['$scope',function($scope) {
      $ = 'Zo0ok';

    Hello {{ name }}

Obviously, if you don’t host angular.js yourself you need to use something like

<script src="">


Playing with smart.js and V7

I have been playing quite much with Node.js lately, and I have put some effort into trying to build it for typical OpenWRT hardware. It turns out that Node.js/V8 is not, and will never be, suitable for hardware without an FPU and at least 128MB RAM.

My curiousity led me to smart.js which is based on the V7 javascript engine. Among the positive details were posix compability, ECMA script 5.1 support, HTTP support (not very different from Node.js, just much less of it), and supposed to be the fastest non-compiled JavaScript engine.

smart.js seems to be the IoT-platform, while v7 is just the JavaScript engine. smart.js had some peculiar properties:

  • The source zip I downloaded from github failed to compiled becuase of a missing .git-directory. Using git clone solved this.
  • The binary always reads and executes smart.js from the same folder. If you give more command line arguments, it executes those after smart.js.
  • I found no way to pass command line arguments.

It comes with a little set of IoT-example files, and I found this a little confusing.

I found out it was possible to use v7 standalone. It is a very nice v7.c file that is simply compiled:

$ gcc -O3 -DV7_EXE -o v7 v7.c -lm

This is very promising – a lot easier than building node.

I found that neither console.log or process.argv exist and found ways to work around this. My plan was to build a little benchmark suite and test different things. I did not get so far: the below program just creates 100 random objects (TestBit) 10 times and stores in arrays.

if ( 'undefined' === typeof process ) {
  process = {
    exit : exit,
    argv : [ 'v7', 'v7bench.c' ]  // faking command line arguments

if ( 'undefined' === typeof console ) {
  console = {
    log : print

var TestBit = function() {
  this.n  = Math.random();
  this.s  = '' + this.n;
  this.s1 = this.s.substr(2,1);
  this.s2 = this.s.substr(3,2);
  this.s3 = this.s.substr(5,3);
  this.b  = this.s > 0.5;

var generateTestBits = function(n) {
  var i;
  var r = new Array(n);
  for ( i=0 ; i<n ; i++ ) {
    r[i] = new TestBit();
  return r;

var Timer = function() {
  this.start =;
  this.last  = this.start;

  this.split = function() {
    var r;
    var n =;
    r = n - this.last;
    this.last = n;
    return r;

var timer = new Timer();

var i = 0;
var tests = [];
var tmp;


while(i<10) {
  console.log('' + ( i*100) + ':' + timer.split());

This simple program unfortunately proved that performance is much worse than I could expect. Below are timings for generating the 1000 objects, in steps of 100 at a time. Benchmarks on a RPi2 900MHz ARMv7 CPU.

There is a flag (-vo) described as “object arena size” that I decided to play with (values 1, 10, 100 gave very similar results as 1000)

       default    -vo 1000    -vo 5000    -vo 10000
 100      0.2s       0.12s        6.8s          28s                         
 200      0.4s       0.16s        8.6s
 300      1.3s       0.23s       12  s
 400      7.4s       0.24s        2.6s
 500     13  s       0.37s       15  s
 600     35  s       0.93s       32  s
 700     47  s       2.9 s       48  s
 800     79  s       4.6 s       45  s
 900    160  s       5.9 s       36  s
1000    160  s       8.0 s       37  s

Well, generating 100 random objects on a RPi2 using a simple interpreter could take 0.2 seconds (the top left value): quite reasonable I guess. As the number of generated objects grows the performance is completey ruined. With default parameter, a full 1.6 seconds is spent to generate one single “TestBit”. Memory usage of the v7 process is insignificant.

I dont know what the “object arena size” setting does, but it obviously changes something. The 400-value for the vo=5000 series is actually reproducable.

Even for an IoT-platform I find this performance (and the lack of predictability) unacceptable. V7 officially aims to be the fastest interpreted JavaScript engine out there: it must be cabable of handling what is actually small arrays (10×100 elements) of small objects.

Effects of cache on performance

It is not clear to me, why is Node.js so amazyingly slow on a Raspberry Pi (article 1, article 2)?

Is it because of the small cache (16kb+128kb)? Is Node.js emitting poor code on ARM? Well, I decided to investigate the cache issue. The 128kb cache of the Raspberry Pi is supposed to be primarily used by the GPU; is it actually effective at all?

A suitable test algorithm
To understand what I test, and because of the fun of it, I wanted to implement a suitable test program. I can imagine a good test program for cache testing would:

  • be reasonably slow/fast, so measuring execution time is practical and meaningful
  • have working data sets in sizes 10kb-10Mb
  • the same problem should be solvable with different work set sizes, in a way that the theoretical execution time should be the same, but the difference is because of cache only
  • be reasonably simple to implement and understand, while not so trivial that the optimizer just gets rid of the problem entirely

Finally, I think it is fun if the program does something slightly meaningful.

I found that Bubblesort (and later Selectionsort) were good problems, if combined with a quasi twist. Original bubble sort:

Array to sort: G A F C B D H E   ( N=8 )
Sorted array:  A B C D E F G H
Theoretical cost: O(N2) = 64/2 = 32
Actual cost: 7+6+5+4+3+2+1     = 28 (compares and conditional swaps)

I invented the following cache-optimized Bubble-Twist-Sort:

Array to sort:                G A F C B D H E
Sort halves using Bubblesort: A C F G B D E H
Now, the twist:                                 ( G>B : swap )
                              A C F B G D E H   ( D>F : swap )
                              A C D B G F E H   ( C<E : done )
Sort halves using Bubblesort: A B C D E F G H
Theoretical cost = 16/2 + 16/2 (first two bubbelsort)
                 + 4/2         (expected number of twist-swaps)
                 + 16/2 + 16/2 (second two bubbelsort)
                 = 34
Actual cost: 4*(3+2+1) + 2 = 26

Anyway, for larger arrays the actual costs get very close. The idea here is that I can run a bubbelsort on 1000 elements (effectively using 1000 memory units of memory intensively for ~500000 operations). But instead of doing that, I can replace it with 4 runs on 500 elements (4* ~12500 operations + ~250 operations). So I am solving the same problem, using the same algorithm, but optimizing for smaller cache sizes.

Enough of Bubblesort… you are probably either lost in details or disgusted with this horribly stupid idea of optimizing and not optimizing Bubblesort at the same time.

I made a Selectionsort option. And for a given data size I allowed it either to sort bytes or 32-bit words (which is 16 times faster, for same data size).

The test machines
I gathered 10 different test machines, with different cache sizes and instructions sets:

	QNAP	wdr3600	ac20i	Rpi	Rpi 2	wdr4900	G4	Celeron	Xeon	Athlon	i5
								~2007   ~2010   ~2013
L1	32	32	32	16	?	32	64	32	32	128	32
L2				128	?	256	256	512	6M	1024	256
L3							1024				6M
Mhz	500	560	580	700	900	800	866	900	2800	3000	3100
CPU	ARMv5	Mips74K	Mips24K	ARMv6	ARMv7	PPC	PPC	x86	x64	x64	x64
OS	Debian	OpenWrt	OpenWrt	OpenWrt	OpenWrt	OpenWrt	Debian	Ubuntu	MacOSX	Ubuntu	Windows

Note that for the multi-core machines (Xeon, Athlon, i5) the L2/L3 caches may be shared or not between cores and the numbers above are a little ambigous. The sizes should be for Data cache when separate from Instruction cache.

The benchmarks
I ran Bubblesort for sizes 1000000 bytes down to 1000000/512. For Selectionsort I just ran three rounds. For Bubblesort I also ran for 2000000 and 4000000 but those times are divided by 4 and 16 to be comparable. All times are in seconds.


	QNAP	wdr3600	ac20i	rpi	rpi2	wdr4900	G4	Celeron	Xeon	Athlon	i5
4000000	1248	1332	997	1120	396	833		507	120	104	93
2000000	1248	1332	994	1118	386	791	553	506	114	102	93
1000000	1274	1330	1009	1110	367	757	492	504	113	96	93
500000	1258	1194	959	1049	352	628	389	353	72	74	63
250000	1219	1116	931	911	351	445	309	276	53	61	48
125000	1174	1043	902	701	349	397	287	237	44	56	41
62500	941	853	791	573	349	373	278	218	38	52	37
31250	700	462	520	474	342	317	260	208	36	48	36
15625	697	456	507	368	340	315	258	204	35	49	35
7812	696	454	495	364	340	315	256	202	34	49	35
3906	696	455	496	364	340	315	257	203	34	47	35
1953	698	456	496	365	342	320	257	204	35	45	35


	QNAP	wdr3600	ac20i	rpi	rpi2	wdr4900	G4	Celeron	Xeon	Athlon	i5
1000000	1317	996	877	1056	446	468	296	255	30	45	19
31250	875	354	539	559	420	206	147	245	28	40	21
1953	874	362	520	457	422	209	149	250	30	41	23

Theoretically, all timings for a single machine should be equal. The differences can be explained much by cache sizes, but obviously there are more things happening here.

Mostly the data makes sense. The caches creates plateaus and the L1 size can almost be prediced by the data. I would have expected even bigger differences between best/worse-cases; now it is in the range 180%-340%. The most surprising thing (?) is the Selectionsort results. They are sometimes a lot faster (G4, i5) and sometimes significantly slower! This is strange: I have no idea.

I believe the i5 superior performance of Selectionsort 1000000 is due to cache and branch prediction.

I note that the QNAP and Archer C20i both have DDRII memory, while the RPi has SDRAM. This seems to make a difference when work sizes get bigger.

I have also made other Benchmarks where the WDR4900 were faster than the G4 – not this time.

The Raspberry Pi
What did I learn about the Raspberry Pi? Well, memory is slow and branch prediction seems bad. It is typically 10-15 times slower than the modern (Xeon, Athlon, i5) CPUs. But for large selectionsort problems the difference is up to 40x. This starts getting close to the Node.js crap speed. It is not hard to imagine that Node.js benefits heavily from great branch prediction and large cache sizes – both things that the RPi lacks.

What about the 128k cache? Does it work? Well, compared to the L1-only machines, performance of RPi degrades sligthly slower, perhaps. Not impressed.

Bubblesort vs Selectionsort
It really puzzles me that Bubblesort ever beats Selectionsort:

void bubbelsort_uint32_t(uint32_t* array, size_t len) {
  size_t i, j, jm1;
  uint32_t tmp;
  for ( i=len ; i>1 ; i-- ) {
    for ( j=1 ; j<i ; j++ ) {
      jm1 = j-1;
      if ( array[jm1] > array[j] ) {
        tmp = array[jm1];
        array[jm1] = array[j];
        array[j] = tmp;

void selectionsort_uint32_t(uint32_t* array, size_t len) {
  size_t i, j, best;
  uint32_t tmp;
  for ( i=1 ; i<len ; i++ ) {
    best = i-1;
    for ( j=i ; j<len ; j++ ) {
      if ( array[best] > array[j] ) {
        best = j;
    tmp = array[i-1];
    array[i-1] = array[best];
    array[best] = tmp;

Essentially, the difference is how the swap takes place outside the inner loop (once) instead of all the time. The Selectionsort should also be able of benefit from easier branch prediction and much fewer writes to memory. Perhaps compiling to assembly code would reveal something odd going on.

Power of 2 aligned data sets
I avoided using a datasize with the size an exact power of two: 1024×1024 vs 1000×1000. I did this becuase caches are supposed to work better this way. Perhaps I will make some 1024×1024 runs some day.

JavaScript: switch options

Is the nicest solution also the fastest?

Here is a little thing I ran into that I found interesting enough to test it. In JavaScript, you get a parameter (from a user, perhaps a web service), and depending on the parameter value you will call a particular function.

The first solution that comes to my mind is a switch:

function test_switch(code) {
  switch ( code ) {
  case 'Alfa':
  case 'Mike':

That is good if you know all the labels when you write the code. A more compact solution that allows you to dynamically add functions is to let the functions just be properties of an object:

x1 = {

function test_prop(code) {
  var f = x1[code];
  if ( f ) f();
  else call_default();

And as a variant – not really making sense in this simple example but anyway – you could loop over the properties (functions) until you find the right one:

function test_prop_loop(code) {
  var p;
  for ( p in x1 ) {
    if ( p === code ) {

And, since we are into loops, this construction does not make so much sense in this simple example, but anyway:

x2 = [
  { code:'Alfa'     ,func:call_alfa    },
  { code:'Bravo'    ,func:call_bravo   },
  { code:'Charlie'  ,func:call_charlie },
  { code:'Mike'     ,func:call_mike    }

function test_array_loop(code) {
  var i, o;
  for ( i=0 ; i<x2.length ; i++ ) {
    o = x2[i];
    if ( o.code === code ) {

Alfa, Bravo…, Mike and default
I created exactly 13 options, and labeled them Alfa, Bravo, … Mike. And all the test functions accept invalid code and falls back to a default function.

The loops should clearly be worse for more options. However it is not obvious what the cost is for more options in the switch case.

I will make three test runs: 5 options (Alfa to Echo), 13 options (Alfa to Mike) and 14 options (Alfa to November) where the last one ends up in default. For each run, each of the 5/13/14 options will be equally frequent.

Benchmark Results
I am benchmarking using Node.js 0.12.2 on a Raspberry Pi 1. The startup time for Nodejs is 2.35 seconds, and I have reduced that from all benchmark times. I also ran the benchmarks on a MacBook Air with nodejs 0.10.35. All benchmarks were repeated three times and the median has been used. Iteration count: 1000000.

(ms)       ======== RPi ========     ==== MacBook Air ====
              5      13      14         5      13      14
switch     1650    1890    1930        21      28      30
prop       2240    2330    2890        22      23      37
proploop   2740    3300    3490        31      37      38
loop       2740    4740    4750        23      34      36

Well, most notable (and again), the RPi ARMv6 is not fast running Node.js!

Using the simple property construction seems to make sense from a performance perspective, although the good old switch also fast. The loops have no advantages. Also, the penalty for the default case is quite heavy for the simple property case; if you know the “code” is valid the property scales very nicely.

It is however a little interesting that on the ARM the loop over properties is better than the loop over integers. On the x64 it is the other way around.

Variants of Simple Property Case
The following are essentially equally fast:

function test_prop(code) {
  var f = x1[code];   
  if ( f ) f();       
  else call_x();                        

function test_prop(code) {
  var f = x1[code];   
  if ( 'function' === typeof f ) f();
  else call_x();                        

function test_prop(code) {

So, it does not cost much to have a safety test and a default case (just in case), but it is expensive to use it. This one, however:

function test_prop(code) {
  try {
  } catch(e) {

comes at a cost of 5ms on the MacBook, when the catch is never used. If the catch is used (1 out of 14) the run takes a full second instead of 37ms!