Community
Participate
Working Groups
HEAD, was OK in I20101005-0800 package source; public class Snippet { public static void main(String[] args) { System.out.println("Hello \u00e4 \u2063, \\u00e4 \\u2063"); } } - The Externalize Strings wizard wrongly says "Some values are invalid". - The NLS hover in the .properties file wrongly says "Malformed \uxxxx encoding". The encoding/decoding code needs some more cleanup: - NativeToAscii is a bad name. The native2ascii tool is already misnamed, since it converts to "Latin1 + escaped Unicode", and not to ASCII. Our class does not even deal with a "native" encoding, so both parts of the name are wrong. It should rater be called something like PropertiesFileEscapes. - The 2 conversion methods in NativeToAscii should both do the special handling for \t etc. . Callers of getEscapedAsciiString(char) shouldn't have to do that.
(In reply to comment #0) > HEAD, was OK in I20101005-0800 > > package source; > public class Snippet { > public static void main(String[] args) { > System.out.println("Hello \u00e4 \u2063, \\u00e4 \\u2063"); > } > } > > - The Externalize Strings wizard wrongly says "Some values are invalid". > - The NLS hover in the .properties file wrongly says "Malformed \uxxxx > encoding". This is a bug in NativeToAscii.isValidNativeString(String), I will fix it. > The encoding/decoding code needs some more cleanup: > > - NativeToAscii is a bad name. The native2ascii tool is already misnamed, since > it converts to "Latin1 + escaped Unicode", and not to ASCII. Our class does not > even deal with a "native" encoding, so both parts of the name are wrong. It > should rater be called something like PropertiesFileEscapes. ...and then what do you want to call the two methods? escape(..) and unescape(...)? Dani, did not like the names getNativeString() and getEscapedNativeString(). He had suggested - native2Ascii() and ascii2Native(). I agree NativeToAscii is technically incorrect, but it is a well known name and immediately suggests what the code will do. > - The 2 conversion methods in NativeToAscii should both do the special handling > for \t etc. . Callers of getEscapedAsciiString(char) shouldn't have to do that. I can introduce a boolean parameter to do this.
> ...and then what do you want to call the two methods? escape(..) and > unescape(...)? Yes. The Javadoc of the class should tell that this class converts between Java chars and the escaped form that must be used in .properties files. Again, "native" is not involved here at all. In the native2ascii tool, "native" refers to the system's default encoding (which can be configured with the -encoding switch). In our code, "native" is not involved at all. What we do is simple escaping and unescaping. So using the term native2ascii does not suggest what we do - it's just wrong. > > - The 2 conversion methods in NativeToAscii should both do the special handling > > for \t etc. . Callers of getEscapedAsciiString(char) shouldn't have to do that. > I can introduce a boolean parameter to do this. No, don't add a boolean. The methods should be symmetrical, and they currently are not. All callers of getEscapedAsciiString(char) already escape \r, \n, \t, and \f, and that should simply be moved into the method. The other escaping can stay where it is if it is legitimate in that context.
>Dani, did not like the names getNativeString() and getEscapedNativeString(). He >had suggested - native2Ascii() and ascii2Native(). I also suggested escape***(...). So that would be fine by me as well.
Created attachment 181396 [details] fix (In reply to comment #1) > This is a bug in NativeToAscii.isValidNativeString(String), I will fix it. phew.. the \es again. This method is just wrong, got rid of it. Things should work perfectly now.. - Renamed to PropertiesFileEscape, escape(..) and unescape(...) - Made escape(..) and unescape(..) symmetrical - the patch also contains the 2 line fix for bug 328223
(In reply to comment #4) > Created an attachment (id=181396) [details] [diff] > fix Fixed in HEAD.
Now we have only one remaining bad transformation: "\b" is externalized as \b, but it should be \u0008. And BTW: PropertiesFileAutoEditStratergy contains a typo. Example with nasty cases: package pack; public class Snippet { public static void main(String[] args) { p("Hello \\b\\t\\f\\r\\n\\\\:\\!\\# !# \b\t\f\r\n\\"); //$NON-NLS-1$ p("Hello \\b\\t\\f\\r\\n\\\\:\\!\\# !# \b\t\f\r\n\\"); p("\t\tHello \u00e4 \u2603, \\u00e4 \\u2603, \\u"); //$NON-NLS-1$ p("\t\tHello \u00e4 \u2603, \\u00e4 \\u2603, \\u"); } private static void p(String s) { System.out.println(s); } }
Created attachment 181487 [details] additional fix Fixed in HEAD.
.
Verified for 3.7 M3 with I20101025-1800.
