From: Josef 'Jupp' SCHUGT <jupp@...>
Date: 2003-11-22T05:45:07+09:00
Subject: Re: A question about Charsets

Hi!

* Eric-Roger Bruecklmeier; 2003-11-21, 13:01 UTC:
> Josef 'Jupp' SCHUGT schrieb:
>>> in a new application i have to read dBase III Files which were
>>> generated in a DOS enviroment. How can i convert the Data into
>>> Windows codepages from ruby?
>>
>> Map each Byte to the corresponding one using a hash. You need the
>> codepages.
> 
> That's the way i do it now, but it's slow :-(

When I find my code in tons of trouble, friends and collegues come to
me, speaking words of wisdom: write in C. (Sung to: 'Let it be' by
the Beatles).

Speedup calls for a C extension. I'll skip the 'intro to C
extensions' stuff (Thomas and Hunt have that) and directly go to the
implementation of the mapping algorithm.

Suppose s points to array of char to be converted. Suppose you simply
need to map code 0 to 1 and vice versa. In that case use this:

for (p = s; *p; p++) {
  switch (*p) {
    case  0: *p = 1; break;
    case  1: *p = 0; break;
  }
}

You don't need to map any char in the ASCII printable range which
saves a lot of coding. The resulting code is extremely fast.

> Exactly that's the problem, the software has to be portable :-(

The above code is extremely portable. An additional advantage: You
can give the codes in decimal or hexadecimal values.

For 16 Bit codes things are more complicated. You then need

for (p = s; *p; p+=2) {
  switch (*p << 8 + *(p+1)) { /* or the other way round, depends */
    case  0: *p = 1; break;
    case  1: *p = 0; break;
  }
}

and lots of additional cases.

Viel Erfolg,

Josef 'Jupp' Schugt
-- 
                        .-------.
message > 100 kB?   /   |       |
sender = spammer?  /    | R.I.P.|
text   = spam?    /  ___|       |___