cruppstahl’s blog

I’m terribly late with the next release, which I already boiled down to the win32-port, without the port to Darwin/bigendian-architectures. I have two excuses for being late: my daily work currently demands a lot of time, and the second reason is one of the most stupid bugs I’ve ever had (and in this case I can’t blame anyone but myself). Here’s what happened:

Whenever I declare a structure, which is stored in the file, I have to pack it. It looks like this (simplified):

#include "packstart.h"
struct packed { ... } ;
#include "packstop.h"

All the packing-logic (in this case, printing #pragma pack) is in the header-files. The first few lines of packstart.h and packstop.h are macros to prevent the compiler from parsing the file twice:

#ifndef HAM_PACKSTART_H__
#   define HAM_PACKSTART_H__
#endif

When including packstart.h twice, it was only parsed for the first time. And therefore the very same structure sometimes was compiled with, sometimes without packing. This resulted in segmentation faults, but also in strange situations where the debugger showed values which were different from those used by the compiler.

When I finally saw that, after one week of debugging, I had to bite in my keyboard, or I’d have started screaming…

I really hope that this is the last bug on win32. I will finish the code and then start testing, which will take some more days. I also have decided to use Visual C++ 5.0 as my development environment, and not Visual Studio 6, as I originally planned. I still prefer vi, but Visual C++ 5.0 (Visual Studio 2005) is available as a free edition and it looks nicer and seems to be more convenient then Visual Studio 2003.

chris Coding, hamsterdb

After about 5 weeks of development (and 3 weeks of vacation), release 0.4.0 is ready!

This release features some minor bugfixes, and the implementation of database cursors. You can use cursors to bi-directionally iterate over your database, and to insert, erase or replace entries.

Cursors are very fast, especially the iterating and “move to next/previous element” are very cheap operations.

To see cursors in action, i have added two small example programs.

• db2.c Copies one database into another; you can also use this to copy data from/to in-memory-databases
• db3.c Demonstrates how to sort data with hamsterdb; reads text from stdin, breaks it up into words and inserts all words in the database. Afterwards, the words are printed in sorted order

The already existing sample is here.

See the CHANGELOG for a detailled list of changes.

The newest hamsterdb version can be downloaded here:
http://www.crupp.de/dl/hamsterdb-0.4.0.tar.gz.

chris Coding, hamsterdb, Libraries