Knowing the kdb community, someone will post a benchmark full of nonsensical values in a plain text file available on an ftp site that shows kdb winning by 100000000 of some unknown unit...

In my experience the kdb community does exactly the opposite, showing a real world use case (high frequency trading for example) using real world data and shows that their platform is significantly faster than the competitors. Is it worth the price? Not so sure, but it certain use cases it certainly is.

Yeah kdb is built so that the entire core of the db fits in L1 CPU cache. It really is an impressive feat of engineering and optimization.

I did hate it when we used it however. It required more or less a team of people doing querying on it fulltime. Have you seen the queries in Q? It looks like someone set the baud speed wrong on a serial connection:

From wikipedia's page on Q (the query language of kdb):

    The factorial function can be implemented directly in Q as
    
    {prd 1+til x}
    or recursively as
    
    {$[x=0;1;x*.z.s[x-1]]}

Makes sense to me, assuming 'til' is 'iota' and 'prd' is 'fold-with-multiplication'.

I don't know Q, but I'd guess you read it right-to-left like J/K/APL. `x` is the right argument (in J it's `y`), so if we were to call `factorial 5`:

1. Create an array of [0,x) | 0, 1, 2, 3, 4

2. Add 1 to each element of the array | 1, 2, 3, 4, 5

3. Fold the array with multiplication | 120

Array languages are super elegant and fun once you use them a bit.

It's not like that first factorial function is actually any worse than the usual Haskell one:

    fac = \n -> product [1..n]

The recursive one is also much the same as a Haskell one, but Q is hardly built around idiomatic recursion.

This[1] would be the one, right?

http://kparc.com/q4/readme.txt