Luc Maisonobe a écrit :
> Nikola Petrović a écrit :
>> Hello all,
>> I have a problem with base64 codec and encoding ascii characters. Here`s the
>> code:
>> int ok=0, bad=0;
>> for(int i=128; i<256; i++){
>> char c = (char)i;
>> str = "" + c;
>> String enc = new
>> String(Base64.encodeBase64(str.getBytes()));
>> String dec = new
>> String(Base64.decodeBase64(enc.getBytes()));
>> if(str.equals(dec)) ok++;
>> else bad++;
>> }
>> System.out.println("ok: " + ok);
>> System.out.println("bad: " + bad);
>>
>> I get this:
>> ok: 80
>> bad: 48
>>
>> So, simple encoding and decoding of ascii chars > 128 doesn`t work for me.
>> Is there some explanation?
>
> Both the String.getBytes() and the String constructors from bytes rely
> on some charset to convet characters to bytes. I guess your JVM
> configuration has some inconsistencies in the default charset. On my
> Linux box with a default charset set to UTF-8, your code works well.
>
> For robustness, I suggest you explicitely set the charset as follows:
>
> String enc = new String(Base64.encodeBase64(str.getBytes("UTF-8")),
> "US-ASCII");
> String dec = new String(Base64.decodeBase64(enc.getBytes("US-ASCII")),
> "UTF-8");
You should also have a look at the str = "" + c statement. I guess it
also depends on the default charset. UTF-8 uses at least two bytes
sequences for special characters, whereas a single byte from 128 to 256
is valid only in some encodings (like ISO-8859-x), with encoding
specific meaning. For example in ISO-8859-1 the character 190 (0xBE) is
the 3/4 character (¾ in HTML), but in ISO-8859-15 it is the
capital Y with diaresis (Ÿ in HTML).
Beware that building a Sting with invalid bytes sequences lead to
undefined results.
Luc
>
> Luc
>
>
>> I tried this on os x, java5 and java6, doesn`t work, and on java6 on solaris
>> pc it does. That confuses me totally :)
>>
>> The other part of the problem is probably linked to this one. I`m trying to
>> encode a string in my java code, and decode it in some php code that I use.
>> Is that possible? I guess since base64 is a standard it should work, and yet
>> it doesn`t on os x, nor on solaris :( Maybe it is linked somehow to the
>> ascii char problem?
>>
>> Thanks,
>> Nikola
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: [EMAIL PROTECTED]
>> For additional commands, e-mail: [EMAIL PROTECTED]
>>
>>
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
