Just a note

[Darrell]

I believe Aaron was just throwing out ideas since you requirements are
vague. In most cases C++ solution will be faster than pure Python. Sounds
like you want a disk based dictionary. I'd be curious how much C++ will buy
you in this case. Disk I/O will eat into your gains.

If you try this in Python avoid readline(), but there is a fast replacement
floating around.

My current paradigm is to code it in Python then push the slow stuff into
C++. The Python phase can really clarify the problem.

I'd like to hear how you solve this.

--
--Darrell

Thomas Weholt <thomas at bibsyst.no> wrote in message
news:3826A8E2.D2C557D7 at bibsyst.no...
> > Also, please have a look at bplustree.py which
> > provides many of the advantages of berkeley db but in a pure python
> > implementation.
>
> And once again the re-occurring question when talking about pure python
> implementations of things like these :
>
> 1. speed ??
> 2. size ??
>
> Like I mentioned I`ll be storing huge amounts of data. HUGE !!! Not huge
> entries/pickled objects, but many small. It`s my impression that pure
> python approaches to problems like this ( I`m thinking of pickle vs.
> cPickle) is slow when compared to similar solutions programmed in C/C++.
>
> Thomas
>
>
> --
> http://www.python.org/mailman/listinfo/python-list