On 11/2/06, Ivanov, Alexey A wrote:
Hi all, I've started fixing HARMONY-1809. To remove throws clause from the declaration of replace method, as it was proposed by Oleg in HARMONY-1975, I placed removeItems() and insertItems() calls into try-catch block. This would work OK for any valid arguments. I was going to handle invalid arguments by making adjustments so that the following removeItems() and insertItems() will not throw the exception. After I wrote several tests, I faced strange behaviour of RI with regards to invalid arguments to replace. (The Javadoc say nothing about which valid ranges for replace() parameters as well as any exceptions.) RI accepts invalid arguments but the result differs from what I'd expect. For example, if the content has "text" in it, I'd expect that content.replace(-2, 4, null, 0) would give "xt" as the result. I mean the invalid start position is adjusted to 0, and the length of remove is adjusted to be 2 accordingly. But this is not the case. As the result of this call, all characters are removed leaving "" in the content. Moreover the content object becomes unusable after that: content.insertString(0, "1") throws ArrayIndexOutOfBoundsException. Similarly if number of characters to be removed is greater than the length of the content (content.replace(2, 4, null, 0) with "text" in it), the object will throw ArrayIndexOutOfBoundsException when doing insertString. Considering the fact that GapContent is pretty low-level class in text representation model and that it is protected, I think that Harmony implementation can silently ignore BadLocationException possible thrown from insertItems() and removeItems(). Taking into account erroneous behaviour of RI's replace, we can do that until an application is broken. As another option, we can throw an Error from catch block to make application which depends on implementation of replace() fast-fail. Any objections, comments, opinions?
Hi Alexey, I'm not quite convinced by your evaluation (I'm not an expert in Swing API so I may be wrong). My experiments with GapContent showed that Harmony initially created different object then RI, for example, if you create an object with GapContent() constructor, Harmony will return start==0 and end==9 while RI will return start==1 and end==10. The next point that confuses me that the spec. says about position as "logical position in the storage"...and... "This is not the location in the underlying storage array". So negative value for position may be considered as valid. Thanks, Stepan. Thanks,
Alexey. P.S. The related JIRA issues: https://issues.apache.org/jira/browse/HARMONY-1809 https://issues.apache.org/jira/browse/HARMONY-1975 GapContent Javadoc: http://java.sun.com/j2se/1.5.0/docs/api/javax/swing/text/GapContent.html Description of GapContent.replace: http://java.sun.com/j2se/1.5.0/docs/api/javax/swing/text/GapContent.html #replace(int,%20int,%20java.lang.Object,%20int) -- Alexey A. Ivanov Intel Middleware Product Division
-- Stepan Mishura Intel Middleware Products Division ------------------------------------------------------ Terms of use : http://incubator.apache.org/harmony/mailing.html To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
