Re: [TCLCORE] A hash table benchmark - or why did we really discuss last year the hash table algorithm if tcl is that good.

[Alexandre F. <ale...@gm...>]

On Thu, Jan 10, 2013 at 11:35 AM, Martin Lemburg <mar...@gm...> wrote:
> Hi,
>
> via the Codeproject Daily newsletter I got in touch with the following hash
> table benchmark articel:
>
>
> http://www.codeproject.com/News.aspx?ntag=19837497437284805&_z=0391382
>
>
> The author really would like to know details from under the hood to
> understand why tcl 8.5 is so "unbeatable" in its hash table performance.
> Any one there able to explain him?

Two obvious things:

 (1) IIUC, the test keys are random digit strings. This is a "sweet
spot" by design for Tcl's hash, because of the ubiquity of integer
indices in typical programs (IOW, the hash function is nearly
injective for digit strings). This may not be the case of the other
languages' hashes.

 (2) To analyze such things, it is good to separate algorithmic
performance from implementation. For the former, compute an histogram
of bucket depths. The latter should only be done iso-hashfunction.

-Alex