Ak ste sa tento rok zúčastnili Webexpa 2011, mohli ste na chodbách stretnúť až siedmych ľudí s visačkou iNOVA sys. Najviac nás, okrem technologických noviniek, zaujímali startupy a všetko čo sa okolo nich točilo. Nuda? Na tri dni pre nás neexistovala.

Stručný súhrn vecí, ktoré ma potešili:

• Tripomatic – môj favorit Startup show, čiže v súťaži o 1 milión Kč za najlepší prihlásený startup, zaslúžene zvíťazil
• Tomáš Beránek a jeho prednáška “zo života” o tom, ako (ne)balil svoju zákonitú – samozrejme prítomnú na prednáške (http://webexpo.cz/praha2011/prednaska/o-uprimnosti-a-dalsich-fatalnich-chybach-v-algoritmech-predvztahove-komunikace/)
• Množstvo známych tvárí, ktoré sme na chodbách a dvore postretávali, či už ide o chalanov z Lingibli, Relbitu, z HotGloo, rovnako ako šumné hostesky (vlani sa mi ale tie kroje páčili o trošku viac …
• Zlatý kliniec programu John Vanhara a jeho American Dream v praxi

Čo ma naopak sklamalo:

• Za jedlo si organizátori WebExpa ešte stále hviezdu Michelina nezaslúžia – ale treba uznať, že všadeprítomné jednohubky boli výborné
• K tričku som sa tento rok nedostal, keďže ich dav rozkmásal a pobral skôr, ako som si vôbec všimol, že sa niečo deje. Prečo nám ich nedali do igelitiek pri check-ine? Ahááá, to by sa tam nezmestilo toľko letákov.

A aký je môj celkový dojem?

Výborný. A to si myslím, že hovorím za všetkých z iNOVAsysu. Vďaka WebExpu sme sa viezli domov zasa o kus múdrejší a inšpirovanejší a plní elánu zdolávať nové výzvy internetového sveta.

For a medium-size IT company as we are, there seems to be no way, how to help fighting terrorism.  But we are different. We see things others overlook. We are highly motivated. We are frigging smart. We have a damn mission.

Hereby, in the following twenty-or-so lines, we will prove the world, that Barack Obama IS NOT Osama bin Laden by any means! We will reveal the secret connection between English literature and English breakfast. We will present some shocking facts about Justin Bieber (we hilighted this solely for higher Google ranks, not as if we enjoyed his music). We will prove much more about the secret connections behind the hood! (If you still don’t see the connection to my previous post, stay cool. This is a follow-up post indeed.).

OK, so now, as I have your interest, let’s focus on the list of celebrities we will test against name silimarity (with an algorithm explained in the previous article). Look at the list of the names and give an educated guess, which pair will look the most similar for the algorithm we designed previously. Let’s go:

• Barak Obana vs Barack Obama
• Barack Obama vs Osama bin Laden
  
• Donald Trump vs Donald Duck
• Andi Roddick vs The flying spagetti monster
  
• Bill Cosby vs Ted Mosby (the big battle for the most awesome TV character!)
• James Joyce vs Jam n Juice

Got the guesses?! Fantastic! Now we will give you a chance to check your intuition against a supercool algorithm of ours called levenstheinX. This algorithm is intended to take a pair of names and return a number from zero to one. The bigger this number is, the more similar the names are.  The listing of the algorithm can be found at the very end of this blog post.

Without further ado, let’s reveal the top 3 similar pairs of names! Drumrolls please….

3. James Joyce vs Jam n Juice with a score of 0.636 that means a 63% similarity!

Congratulations. Let’s see who placed second…

2. It is Donald Trump vs Donald Duck scoring 0.667!  Man, get him to know Scrooge the Duck!

And the winner of the contest for the most similar names iiiiiis…..

1. Barak Obana vs. Barack Obama with only 17% uncertainty of the equality (score was 0.83).

While the winners enjoy their fame, we have one more cool announcement to make. We are ABSOLUTELY sure, that Barack Obama is nothing like the nasty terrorist Osama Bin Laden with their test scoring a straight zero (and this is also it for our contribution to the war against the terrorism). Awesome.

PS: What is the similarity score you would guess for the following: Justin Bieber vs. Just in beaver? (Find the answer at the very bottom of the page).

Hey techies! Bet you are damn curious about the supercool algorithm that produced these outputs. Here it goes, feel free to study it, feel free to use it, it is all yours:

function levenstheinX($s1,$s2){ 

  if(strlen($s1) > strlen($s2)) levenstheinX($s2,$s1);

  $score = levenshtein($s1, $s2);

  if(strlen($s1) > 0 ){

    $diff = (double)$score / strlen($s1);

  }else{ $diff = 0.0;}

  return (1.0 - $diff);

 }

THE ANSWER to the Bieber related question is 69%  

…the fact that there are basically only three types of errors in typed names:

• Mishearing a name –  can be caused by unusual pronounciation, different accent or some distraction when a name is spoken to a person writing it down. For example Eric Clapton can be easily heard as Eric Clayton. There are also names that have multiple textual representations, for example Ján Kováč , a typical slovak name is pronounced the same way as Ján Kovács, a hungarian variation of the surname.
• Mistyping a name – this kind of errors can be caused by distraction, fatigue, or dyslexia when inputting a name. It results in mistyped, exchanged or missing letters. As an example imagine someone really tired writing David Coperfeld instead of David Copperfield. By merely seeing these two names, you are pretty sure they represent the same name. But how do you convince your program to feel the same way about it?
• Different representations – when you intergrate data from different sources, you may find the same names represented differently. Some of them may use accent marks and diacriticals, others may not. One source holds string as UTF-8 strings, other uses local encodings.

OK. But I have two names and I need to figure out an algorithm to see if they represent the same real name or not. What should I do?

Solve all the three problems above, I say. Once you don’t know, from which type of errors your data suffer the most, you have to design all three algorithms. Before you get scared, just look how simple it is:

• Misheard names – this is basically just an acoustic problem. If two different names sound very similarly, they are pretty likely to be interchanged when interpreted by a listener. And vice versa – names like Rafael Nadal and Andy Roddick have a very little chance of being misheard in favor of one another. The good news is, that there are a few known algorithms to determine whether two words sound the same or not. See soundex, or Metaphone algorithm entries on Wikipedia. What happens here is, that every name is normalized into a code. Names that sound similar share the same code and that is the solution to our problem (but be sure to check if the algorithm is suitable for the language of your names).
• Mistyped names – this kind of  errors is even easier to figure out. Let’s just set up a metric we basically already defined above – let’s say that two strings differ by N letters, if exactly N letters have to be removed, added, or interchanged in the first string in order to end up with the second one. For example John Doe differs from Jon Do in 2 letters (remove h; remove e). More precisely, this metric is called edit distance. The greater the edit distance between two strings is, the more typing errors had to be made during the input. And that’s it. Let’s say people have 10% chance of mistyping a name. So what is the change of two typing errors in one name? Yep, it is less,  1% or so. What we can deduct from this fact is, that the lower the edit distance of two names is, the bigger the chance that they represent the same name.
• Different representations – this is pretty simple to solve. Once you convert both names to the same encoding, normalize them to unaccented and lowercase strings. Be aware that this may create some type I errors making names Emil Sanchéz the same as Emil Sanchez. But we believe that it eliminates substantially bigger amount of type II errors, so it should be a benefit for you in general.

So much for the theory. I presented three approaches to compare names against mistyped, misheard, and “misrepresented” characters.

To keep this post concise, I decided to split it in two parts. Next time, I will show a practical implementation of all the theory above. Bet you’re curious about the Obama question .

Startupcamp.sk events

These events are dedicated to provide a place for networking, startup pitches, and investor meetups and to educate the community by inviting speakers with presentations that it might have interest in. Local Startupcamp meetups are currently organized:

• monthly in Bratislava (by Radovan Grezo and Matej Ftacnik) – 100 – 200 attendees
• monthly in Kosice (Marek Kuzma) – about 40 attendees
• bimonthly in Piestany (Zuzana Urbanova) – roughly 20 attendees
• proposal to host in Zilina (Frantisek Milec)

For more information, videos, posts and info regarding startupcamp, go to startupcamp.sk

STARTeurope.at events

STARTeurope is an entrepreneurial organization based in Vienna. They organise startup events called STARTup Live. The list of STARTup Live events held:

• STARTup Live Vienna (27.5-29.5.2011) – there is still time to register. STARTeurope has asked me to be one of the mentors for the participating startups! It’s a huge privilege and responsibility to be a mentor at an event like this and I am really grateful for this opportunity. More info here.
• STARTup Week Vienna (03.10-07.10.2011) – one of the biggest Startup oriented events in Central Europe. A week full of valuable information, great speakers and helpful mentors. www.startupweek2011.at

Barcamp.sk events

Barcamp is an international network of user-generated open conferences on which the attendees become speakers. It is a meeting of people that was created from the need for sharing and for learning in an open environment. The main goal of these conferences are not only presentations and demos but mainly the discussions and the cooperation on the speaker with the audience. The next barcamp will be:

• Barcamp Zilina (28.5.2011) – it’s going to be filled with exceptional speakers and great people, so be sure to come and see. I believe there is still some free space and there are buses beeing organized from Bratislava and Kosice for free for the participants. More info at www.barcamp.sk

Other events

There are also other conferences and events organized (e.g. TEDx, OpenCoffee, Forwarders.sk, EnglishCoffee, Rubyslava, CloudCoffee and some “professional” conferences from exponet.cz or itapa.sk) but today I wanted to introduce to you my personal favorite ones that are useful and beneficial for startups. If I have forgotten about some events, let me know in the comments.

Once you have determined what pays the bills in the company, think about your core business. Is your core business bringing the most into the company? If yes, then everything is alright. If not, you should definitely re-think what really is your core business.

Once you know what is your core business, try to look at all those other activities in the company. Have you caught yourself doing a job you don’t like or someone else might do it better than you? Probably yes. Why should you do It? It’s a good thing to stay faithful to your core business and be the best at that. You don’t want to end up being that company, that can do literally anything, but nothing of good quality.

The good thing is, that there is no reason why you should do stuff you are not good in, or just simply don’t enjoy doing. If your client wants something you cannot deliver by yourself be sure to have a good network of reliable business partners you can outsource to. That will enable you to concentrate on your core business, deliver additional outstanding services to your client and also have a blooming relationship with your business partners. It’s essentially a win-win-win situation.

To give you a real life example, we at inovasys have tried to do graphic design (as I like to call it developer graphics) by ourselves in the early days. The only outcome of that was that we were frustrated and the client was not satisfied by our work. Since that bad experience we have partnered with several graphic designers and graphic studios. Now our client can choose from several different styles and approaches, we can concentrate on what we are good at and give the client a better product.

You should rather be really good in one thing than mediocre or even crappy at several ones.

• Can it withstand market pressures from competitors ?
• Can it deliver more economic value to customers than alternatives ?

First, take your idea’s ambient temperature. Take a look at the fire chart I’ve put together for you according to Doblin’s Ten Types of Innovation, where each flame represents one type. The idea behind it is simple : the more ways in which your idea is innovative, the better it can withstand the offerings of your competitors. According to a research by Harvard Business Review, ideas that innovate in six or more areas have a good starting position for success. If your idea excels in three or less types, it might not be the next best thing.

Next, measure your idea’s heat index. Is your idea hot ? Good. Now will it deliver value to your customers ? To answer this question, I use the slightly simplified version of the EVE (Economic Value Estimation)  index. The algorithm is simple : EVE = IEV / RV, where IEV stands for Incremental Economic Value (the sum of decreased costs or increased revenues associated with each innovation type) and RV is Reference Value (what customers pay for the next-best alternative of your product).

The EVE index shows, by innovation type, how much increased value your idea offers over what’s already available on the market. Be careful when interpreting the factor, as it depends on the industry – in B2B, an index of 1,5 can be considered hot, but in a CPG context, the number will most likely need to be around 4. That being said, consider anything below 1,1 as cold and a nonstarter.

I had a database table containing a TEXT column holding an arbitarly long string in each record. There was also a surrogate column called crc, holding a CRC32 hash of the text from the same record. This is a standard approach to allow queries to use indices when querying  TEXT columns for an exact match (there is no way to put an index on a TEXT column, so you put it on the hashed column and query against it). So I had the following super-slowly behaving select:

SELECT id FROM test_table WHERE crc=CRC32('abc');

Here starts my lesson. I ran the above select over and over again, but it still kept ignoring the index on the crc column. After unsuccessfully trying to force the MySQL optimizer to use my index, I raged. All I kept getting were full table scans on the simplest select imaginable. I nervously wandered around the room, almost sure I discovered a bug in the MySQL optimizer.

What I learnt is, that in situations like this, taking a 30 minute break can make much more sense than furiosly surfing the internet and submitting posts to forums and bugtrackers. Clearing your mind with some completely unrelated activity makes you re-focus on the problem once you return to it, discarding all (mostly wrong) assumptions you made in emotion, giving you a much better chance of finding a solution.

So how about my problem with the malfunctioning index? Once I returned with a clear mind, I almost instantly realized that my crc column was a varchar, but CRC32 function in the where clause returned an INT. This type mismatch prevented MySQL from using the index. Fixing the query was a matter of seconds now:

SELECT id FROM test_table WHERE crc=CAST(CRC32('abc') AS CHAR);

PS: I just think this just proves one of the principles of agile development – overworking does not mean working more effectively, nor  getting more done. Knowing when to stop is a valuable trait, too .

A few months ago, we decided to access the database of one of our products with MySQL stored procedures. Suprisingly, the transition from the usual PHP querying based on mysql_query was not as smooth as we thought. We started receiving annoying ‘Commands out of sync’ errors during subsequent queries for no obvious reason. It took us some time to figure out, and to fine-tune our retrieval procedure so we decided to share.

Firstly. Forget about PHP’s mysql extension if you want to run stored code. It will not work. You will keep getting Out of sync exceptions upon the second and all subsequent queries and there seems to be no way around it yet. You will have to use mysqli to do the dirty job (fortunately, it is a standard PHP extension as well).

The problem with mysql_query is that it does not support multiple resultsets. And that is exactly what your stored procedures do – they tend to return more than one resultset. Whenever you call an SP, the exit status is secretly carried in an (empty) resultset along. If you add it up with your own output from a procedure, some of the resultsets from the query will be ignored by your PHP retrieval code. When you try to run another query, the pending resultset(s) will still be in the buffer. This leaves the engine uncertain whether you finished processing the previous result and it…(rather than doing something unexpected) results in an exception.

To avoid this hassle we need to retrieve all resultsets properly before sending another requests. To do this, we will use mysqli_next_result and mysqli_store_result functions to loop over the remaining resultsets and clear the buffer. Working snippet of code follows:

$mysqli = new mysqli("server", "user", "pass", "db");

$query = $mysqli->multi_query("CALL proc_name($proc_param1,$proc_param2)");

if ($query) {

   $result = $mysqli->use_result();

   if ($result) {

      while ($row = $result->fetch_assoc()) {

         do_whatever_with_the_result($row);

      }

      $result->free();

   }

   //clear the resultset buffer

   while ($mysqli->more_results()){

      $mysqli->next_result();

      $result = $mysqli->use_result();

      if ($result instanceof mysqli_result) {

         $result->free();

      }

   }

}

Simple enough. If you are familiar with the mysql extension, mysqli commands should be readable to you without looking to the manual (but be sure to give a glance at  the official manual) .

Hope this article got you started, if you wanted to figure out why the hell those SPs don’t work . Feel free to post comments.

Tip: You may want to use the multiple resultsets “bug” (we know, it is NOT a bug, stay cool) as a feature and accept more results from stored procedures (by not discarding the resultsets right away).