it's a simple problem, which can be stated as follows:
- the char data type in Java does not denote a character, rather, it
denotes a UTF-16 encoding element
- some Unicode characters, i.e., those in the BMP, are represented by one
char element (char[1]), while other Unicode characters require two char
elements (char[2]);
- in order to make use of non-BMP characters, of which there are now many
standardized instances, one must either pass a char array, e.g., char[2],
or, alternatively pass an int, which is capable of representing all Unicode
code points in the range of 0 ... 0x10;
at some point, FOP needs to support the effective use of characters outside
the BMP coding space, and, consequently, those FOP interfaces that use the
char type need to be upgraded to int;
I am referring to FOP defined interfaces mind you, not Java defined
interfaces; in general, the Java interfaces provide mechanisms to address
this problem; for instance, see the discussion in the preamble of the
definition of
java.lang.Characterhttp://java.sun.com/javase/6/docs/api/java/lang/Character.html,
the pertinent point of which I repeat below:
- The methods that only accept a char value cannot support supplementary
characters. They treat char values from the surrogate ranges as undefined
characters. For example, Character.isLetter('\uD840') returns false, even
though this specific value if followed by any low-surrogate value in a
string would represent a letter.
- The methods that accept an int value support all Unicode characters,
including supplementary characters. For example,
Character.isLetter(0x2F81A) returns truebecause the code point value
represents a letter (a CJK ideograph).
what I believe the original commenter is pointing out (and that I am
agreeing with) is that FOP needs to take care to not use the *char* type for
interface parameters that are intended to denote a Unicode character; or, if
they do, then an overloaded version of the same interface that uses the *int
* type should also be provided;
for example, the following interfaces need to be upgraded to int or to have
an overloaded int variant:
org.apache.fop.fonts.Font.getKernValue(char ch1, char ch2);
org.apache.fop.fonts.Font.getWidth(char charnum);
org.apache.fop.fonts.Font.mapChar(char c);
org.apache.fop.fonts.Font.hasChar(char c);
org.apache.fop.fo.CharIterator.replaceChar(char c);
org.apache.fop.fo.flow.Character.getCharacter();
org.apache.fop.util.CharUtilities.*;
...
i have already upgraded al of the CharUtilities.* methods to use int instead
of char in my present work on complex script support, but there are a
variety of other internal interfaces as noted above that need to be upgraded
as well. if you like, I can fold this into my present work, or assign it a
new bug number (which may be the best for tracking purposes);
regards,
glenn
On Fri, May 21, 2010 at 7:31 AM, Peter B. West li...@pbw.id.au wrote:
I'm puzzled by this discussion. AFAIK, Java has rejected moving to 32 bits
in Java 5. Instead, they are supporting supplementary characters. There's a
discussion here:
http://java.sun.com/developer/technicalArticles/Intl/Supplementary/
Peter West
Lord, to whom shall we go?
On 21/05/2010, at 11:11 PM, Glenn Adams wrote:
I concur with this change, and have already made some changes in this
direction in my work on adding complex script support.
Please note that it is not quite so simple as merely changing from char
to int in some locations. It is also necessary to convert from UTF-16 to
UTF-32, i.e., to the full Unicode code point value, which can range from
0x00 through 0x10 (see Unicode 5.2, Section 3.3, Item D9). It is
probably not a good idea to make this conversion too early, but rather, to
defer it until certain well defined interface points, which need to be
documented as taking the full Unicode code point, and not merely a UTF-16
code element.
On Fri, May 21, 2010 at 3:46 AM, Vincent Hennebert vhenneb...@gmail.com
wrote:
Hi,
Author: jeremias
Date: Thu May 20 09:52:27 2010
New Revision: 946585
URL: http://svn.apache.org/viewvc?rev=946585view=rev
Log:
Changed many variables and parameters from int to char because AFP
font support mostly uses Unicode code points unlike Type 1 and TrueType
support which use internal character code points (the result of
Font.mapChar()). This should improve code readability.
Not sure this is a desirable change. char can only address characters
from the Basic Multilingual Plane. Java 1.5 have started to use int to
overcome that issue actually. So unless there is a fundamental
limitation in AFP such that characters beyond the BMP will never be
usable, I think we want to stick to int.
snip/
Vincent
