jsyntaxpane - Project Hosting on Google Code

Yesterday 42 hours ago issue 75 (NPE when there is no parent assigned yet) Status changed by ayman.alsairafi - Did you try that patch? I recall getting some other errors later, as stated in my earlier comment. Status: New Did you try that patch? I recall getting some other errors later, as stated in my earlier comment. Status: New Last 7 days Dec 25, 2009 issue 75 (NPE when there is no parent assigned yet) commented on by rjakob13 - Please add this patch or make sure you don't assume there is a parent. This bug causes issues in code you cannot alter, e.g. in a swing application that uses JEditorPane for syntax highlighting combined with e.g. JXErrorPane that uses JEditorPane as well. We cannot patch all other Java libraries to make sure they assign a parent before calling setContentType. Please add this patch or make sure you don't assume there is a parent. This bug causes issues in code you cannot alter, e.g. in a swing application that uses JEditorPane for syntax highlighting combined with e.g. JXErrorPane that uses JEditorPane as well. We cannot patch all other Java libraries to make sure they assign a parent before calling setContentType. Last 30 days Dec 22, 2009 issue 129 (How do I set start line number?) changed by ayman.alsairafi - Probably the best way is to extend the LineNumberRuler class in jsyntaxpane.components package. You then need to somehow pass the start line to that class, and update the component whenever it is changed (Property change maybe). The line numbers are painted in the class's paintCOmponent. It should be easy to add the start line property. Status: WontFix Owner: ayman.alsairafi Probably the best way is to extend the LineNumberRuler class in jsyntaxpane.components package. You then need to somehow pass the start line to that class, and update the component whenever it is changed (Property change maybe). The line numbers are painted in the class's paintCOmponent. It should be easy to add the start line property. Status: WontFix Owner: ayman.alsairafi Dec 12, 2009 issue 129 (How do I set start line number?) reported by room3studio - My making Flip+Zooming Text Editor which uses multiple editor pane but use same text source something like this: http://www.sigchi.org/chi97/proceedings/poster/leh-fg5.gif I want to use jsyntaxpane show zoomming Code but i can't find the way to set different start line number. Please help me. Thank you. My making Flip+Zooming Text Editor which uses multiple editor pane but use same text source something like this: http://www.sigchi.org/chi97/proceedings/poster/leh-fg5.gif I want to use jsyntaxpane show zoomming Code but i can't find the way to set different start line number. Please help me. Thank you. Dec 06, 2009 Building Wiki page commented on by ayman.alsairafi - I'm not familiar with other IDE's, but Netbeans has excellent Maven support, almost out of the box. With Maven, the JFlex builds will be done for you automatically. If you are on Linux, then it should also be fairly easy. If I just package all sources in a zip file, I cannot guarantee the latest dependencies. I also think that if you need the source, and managed to use svn to get it, using maven to build should be easy for you :-) If you have specific issues, please open them in the issues list. We can track them better there. I'm not familiar with other IDE's, but Netbeans has excellent Maven support, almost out of the box. With Maven, the JFlex builds will be done for you automatically. If you are on Linux, then it should also be fairly easy. If I just package all sources in a zip file, I cannot guarantee the latest dependencies. I also think that if you need the source, and managed to use svn to get it, using maven to build should be easy for you :-) If you have specific issues, please open them in the issues list. We can track them better there. Dec 06, 2009 UsingNew Wiki page commented on by ayman.alsairafi - I just saw this. sorry for delay. Please open an issue next time, as it is easy to track. You can get the SyntaxKit instance in several ways. 1. After you create the EditorPane, and set you contentType you can call the getEditorKit to get the SyntaxKit object 2. You can get the class name of any kit, not just SyntaxKit by calling JEditorPane.getEditorKitClassNameForContentType. Method 1 is preferred. I just saw this. sorry for delay. Please open an issue next time, as it is easy to track. You can get the SyntaxKit instance in several ways. 1. After you create the EditorPane, and set you contentType you can call the getEditorKit to get the SyntaxKit object 2. You can get the class name of any kit, not just SyntaxKit by calling JEditorPane.getEditorKitClassNameForContentType. Method 1 is preferred. Dec 06, 2009 issue 128 (Find Dialog Hidden When In Modal Dialog) reported by mmacfadden - What steps will reproduce the problem? 1. Create a modal dialog class with the editor pane in it. 2. Show the dialog. 3. User the find action. What is the expected output? What do you see instead? I should see the find dialog. The find dialog shows, but is shown behind the modal dialog. What version of the product are you using? On what operating system? 0.9.5 on Windows XP Please provide any additional information below. The find dialog flashes momentarily and then gets moved to the background. I am guessing this is because is is not a modal dialog and the modal dialog that contains the editor pane is. Not sure what the solution is other than giving an option to make the find dialog modal itself. I realize you might not always want the find dialog to be modal, but in this scenario its the only way it would work I think. What steps will reproduce the problem? 1. Create a modal dialog class with the editor pane in it. 2. Show the dialog. 3. User the find action. What is the expected output? What do you see instead? I should see the find dialog. The find dialog shows, but is shown behind the modal dialog. What version of the product are you using? On what operating system? 0.9.5 on Windows XP Please provide any additional information below. The find dialog flashes momentarily and then gets moved to the background. I am guessing this is because is is not a modal dialog and the modal dialog that contains the editor pane is. Not sure what the solution is other than giving an option to make the find dialog modal itself. I realize you might not always want the find dialog to be modal, but in this scenario its the only way it would work I think. Dec 06, 2009 UsingNew Wiki page commented on by mmacfadden - Well the class does have the method. Its not static though, so the question is how do you get the instance of the kit so you can set them. Well the class does have the method. Its not static though, so the question is how do you get the instance of the kit so you can set them. Dec 06, 2009 issue 127 (Incorrect syntax highlighting for variables that are also ty...) changed by ayman.alsairafi - jsyntaxpane is only a lexer. It recognizes tokens only, without any knowledge of their context. It does not know anything regarding the language grammar. There is no way, with the current design to work around this. This will has to wait a while. Could be a long one :-( Status: ForFuture Owner: ayman.alsairafi Labels: Type-Enhancement jsyntaxpane is only a lexer. It recognizes tokens only, without any knowledge of their context. It does not know anything regarding the language grammar. There is no way, with the current design to work around this. This will has to wait a while. Could be a long one :-( Status: ForFuture Owner: ayman.alsairafi Labels: Type-Enhancement Dec 04, 2009 issue 127 (Incorrect syntax highlighting for variables that are also ty...) reported by javlopez - What steps will reproduce the problem? 1. be in text/java mode 2. type Object Object = null; 3. both Object strings are recognized as types even though only the first is the type What is the expected output? What do you see instead? First 'Object' should be highlighted as a type, second 'Object' should be an identifier. What version of the product are you using? On what operating system? verified on jsyntaxpane-0.9.4 and jsyntaxpane-0.9.5-b29, WinXP 64 Please provide any additional information below. What steps will reproduce the problem? 1. be in text/java mode 2. type Object Object = null; 3. both Object strings are recognized as types even though only the first is the type What is the expected output? What do you see instead? First 'Object' should be highlighted as a type, second 'Object' should be an identifier. What version of the product are you using? On what operating system? verified on jsyntaxpane-0.9.4 and jsyntaxpane-0.9.5-b29, WinXP 64 Please provide any additional information below. Earlier this year Nov 27, 2009 Building Wiki page commented on by bob.tantlinger - Wish you guys would just package the source as a zip file with an ant build file, or at least include jflex in the svn. Hate having to jump through hoops and and mess around with maven just to try this out. Wish you guys would just package the source as a zip file with an ant build file, or at least include jflex in the svn. Hate having to jump through hoops and and mess around with maven just to try this out. Nov 27, 2009 UsingNew Wiki page commented on by bob.tantlinger - "You can configure a single setting by calling the DefaultSyntaxKit.setProperty method, or you can supply a Properties instance to merge it with the default properties." How? The DefaultSyntaxKit has no such method... "You can configure a single setting by calling the DefaultSyntaxKit.setProperty method, or you can supply a Properties instance to merge it with the default properties." How? The DefaultSyntaxKit has no such method... Nov 19, 2009 issue 106 (Undo/Redo on Mac OS X) Status changed by ayman.alsairafi - This is fixed with issue 88 and the latest binary release. Status: Fixed This is fixed with issue 88 and the latest binary release. Status: Fixed Nov 19, 2009 issue 111 (All Text is removed when clicking on ReplaceAll button while...) Status changed by ayman.alsairafi - This is actually fixed already in latest binary build. I just forgot to update the issue. Status: Fixed This is actually fixed already in latest binary build. I just forgot to update the issue. Status: Fixed Nov 14, 2009 issue 109 (IndentAction for Python Code) Status changed by ayman.alsairafi - dded PythonIndent action to r123 Status: Fixed dded PythonIndent action to r123 Status: Fixed Nov 14, 2009 r123 (FIx for issue 109 - Python Auto Indentation) committed by ayman.alsairafi - FIx for issue 109 - Python Auto Indentation FIx for issue 109 - Python Auto Indentation Nov 14, 2009 issue 115 (JavaScript support enhancements) Status changed by ayman.alsairafi - Added in r122 Status: Fixed Added in r122 Status: Fixed Nov 14, 2009 r122 (FIx for issue 115 - Better JavaScript support) committed by ayman.alsairafi - FIx for issue 115 - Better JavaScript support FIx for issue 115 - Better JavaScript support Nov 14, 2009 issue 116 (Recompile in java5) Labels changed by ayman.alsairafi - Labels: JDK5 Labels: JDK5 Nov 14, 2009 issue 120 (Problem with SQL statement highlighting) Labels changed by ayman.alsairafi - This works in JDK6, but needs to be retrofitted to JDK 5 branch, which I cannot maintain. Labels: JDK5 This works in JDK6, but needs to be retrofitted to JDK 5 branch, which I cannot maintain. Labels: JDK5 Nov 12, 2009 issue 126 (Add OSGi manifest to existing jar manifest) reported by c.cavey - A basic OSGi manifest will make jsyntaxpane behave as a drop-in component for an OSGi environment with no impact on other usage. The easiest method of achieving this is to integrate the "bnd" tool (http://www.aqute.biz/Code/Bnd) into your build process. Be sure that ALL of the imports are optional and that all internal packages are exported. Additionally include a meaningful symbolic id, name, and version I've used the "wrap" tool (ant/maven task, cli) within bnd to create a plugin with the following properties file -nouses:true Bundle-Version: 0.9.5.b29 Bundle-Name: JSyntaxPane Bundle-SymbolicName: jsyntaxpane Import-Package: !jsyntaxpane, ;resolution:=optional A basic OSGi manifest will make jsyntaxpane behave as a drop-in component for an OSGi environment with no impact on other usage. The easiest method of achieving this is to integrate the "bnd" tool (http://www.aqute.biz/Code/Bnd) into your build process. Be sure that ALL of the imports are optional and that all internal packages are exported. Additionally include a meaningful symbolic id, name, and version I've used the "wrap" tool (ant/maven task, cli) within bnd to create a plugin with the following properties file -nouses:true Bundle-Version: 0.9.5.b29 Bundle-Name: JSyntaxPane Bundle-SymbolicName: jsyntaxpane Import-Package: !jsyntaxpane, ;resolution:=optional Nov 12, 2009 issue 125 (SyntaxKits need to support alternate classloaders) reported by c.cavey - When using jsyntaxpane in the context of OSGi, classloader issues arise. OSGi does not take kindly to Class.forName as this circumvents the class loader dependency heirarchy that OSGi builds for dependency management. Without getting too deep into OSGi, the easiest solution is to allow for an optional classloader to be passed into the constructors of all SyntaxKits so that the base class (DefaultSyntaxKit) can optionally use this classloader for creating components. When using jsyntaxpane in the context of OSGi, classloader issues arise. OSGi does not take kindly to Class.forName as this circumvents the class loader dependency heirarchy that OSGi builds for dependency management. Without getting too deep into OSGi, the easiest solution is to allow for an optional classloader to be passed into the constructors of all SyntaxKits so that the base class (DefaultSyntaxKit) can optionally use this classloader for creating components. Nov 11, 2009 issue 112 (All text disappear if ctrl-z (undo) is pressed after a Repla...) commented on by ayman.alsairafi - binary b29 is already out. binary b29 is already out. Nov 10, 2009 issue 113 (Replace (single occurrence) is missed in the Find&Replace di...) commented on by eliboni - great, let me try out! Thanks! great, let me try out! Thanks! Nov 10, 2009 issue 112 (All text disappear if ctrl-z (undo) is pressed after a Repla...) commented on by eliboni - thx - I'll try it out as soon as the binary is released thx - I'll try it out as soon as the binary is released Nov 08, 2009 issue 111 (All Text is removed when clicking on ReplaceAll button while...) commented on by ayman.alsairafi - An NPE is occuring somewhere, but I can't find it yet. I might just automatically remove highlights while replaceAll is done to avoid the error. An NPE is occuring somewhere, but I can't find it yet. I might just automatically remove highlights while replaceAll is done to avoid the error. Nov 08, 2009 issue 111 (All Text is removed when clicking on ReplaceAll button while...) changed by ayman.alsairafi - Status: Accepted Owner: ayman.alsairafi Status: Accepted Owner: ayman.alsairafi Nov 08, 2009 jsyntaxpane-0.9.5-b29.jar (JSyntaxPane Release 0.9.5-b29) file uploaded by ayman.alsairafi Nov 08, 2009 issue 113 (Replace (single occurrence) is missed in the Find&Replace di...) Status changed by ayman.alsairafi - Almost forgot about that. Just added in r121 and b29 binary download. Status: Fixed Almost forgot about that. Just added in r121 and b29 binary download. Status: Fixed Nov 08, 2009 r121 (FIx for issue 113 - Replace Single occurance.) committed by ayman.alsairafi - FIx for issue 113 - Replace Single occurance. FIx for issue 113 - Replace Single occurance. Nov 08, 2009 jsyntaxpane-0.9.5-b28.jar (JSyntaxPane Release 0.9.5-b28) file uploaded by ayman.alsairafi - Labels: Featured Type-Executable OpSys-All Labels: Featured Type-Executable OpSys-All Nov 08, 2009 issue 124 (Undo grouping doesn't include first action in the chain) commented on by komoku - That's really great. Thanks for the quick reaction and for this nice project, it's very useful to me. I hope Issue 113 is also fixed in the new binary, that one would be nice too :) That's really great. Thanks for the quick reaction and for this nice project, it's very useful to me. I hope Issue 113 is also fixed in the new binary, that one would be nice too :) Nov 08, 2009 r120 (FIx for issue 122 - QuickFind in TabbedPane - Ignore previou...) committed by ayman.alsairafi - FIx for issue 122 - QuickFind in TabbedPane - Ignore previous commit. FIx for issue 122 - QuickFind in TabbedPane - Ignore previous commit. Nov 08, 2009 r119 (FIx for issue 122 - QuickFind in TabbedPane) committed by ayman.alsairafi - FIx for issue 122 - QuickFind in TabbedPane FIx for issue 122 - QuickFind in TabbedPane Nov 08, 2009 issue 122 (Quick Find generates error when editor not hosted in JViewPo...) Status changed by ayman.alsairafi - Fixed in r119. I Changed it to a Container, and avoided the cast altogether. Status: Fixed Fixed in r119. I Changed it to a Container, and avoided the cast altogether. Status: Fixed Nov 08, 2009 issue 121 (Find and Replace dialog can still modify when it's read-only) Status changed by ayman.alsairafi - Fixed in r118 Status: Fixed Fixed in r118 Status: Fixed Nov 08, 2009 r118 (FIx for issue 121 - Replace all modifies read-only editor) committed by ayman.alsairafi - FIx for issue 121 - Replace all modifies read-only editor FIx for issue 121 - Replace all modifies read-only editor Nov 08, 2009 issue 124 (Undo grouping doesn't include first action in the chain) Status changed by ayman.alsairafi - I did more changes in r117. Will create a binary download shortly. See issue 88 Status: Fixed I did more changes in r117. Will create a binary download shortly. See issue 88 Status: Fixed Nov 08, 2009 issue 123 (Undo broken on edits modifying multiple lines) changed by ayman.alsairafi - I did more changes in r117. Will create a binary download shortly. See issue 88 Status: Fixed Owner: ayman.alsairafi I did more changes in r117. Will create a binary download shortly. See issue 88 Status: Fixed Owner: ayman.alsairafi Nov 08, 2009 issue 117 (Replace-all and then undo will erase buffer.) changed by ayman.alsairafi - Something similar done. I did more changes in r117. Will create a binary download shortly. See issue 88 Status: Fixed Owner: ayman.alsairafi Something similar done. I did more changes in r117. Will create a binary download shortly. See issue 88 Status: Fixed Owner: ayman.alsairafi Nov 08, 2009 issue 112 (All text disappear if ctrl-z (undo) is pressed after a Repla...) changed by ayman.alsairafi - I did more changes in r117. Will create a binary download shortly. See issue 88 This seem to be fixed. But need more verification Status: Fixed Owner: ayman.alsairafi I did more changes in r117. Will create a binary download shortly. See issue 88 This seem to be fixed. But need more verification Status: Fixed Owner: ayman.alsairafi Nov 08, 2009 issue 106 (Undo/Redo on Mac OS X) commented on by ayman.alsairafi - I did more changes in r117. Will create a binary download shortly. See issue 88 I did more changes in r117. Will create a binary download shortly. See issue 88 Nov 08, 2009 issue 88 (Undo grouping is strange and deconcerting) commented on by ayman.alsairafi - See fix in r117. Will be do a binary release shortly. See fix in r117. Will be do a binary release shortly. Nov 08, 2009 issue 88 (Undo grouping is strange and deconcerting) Status changed by ayman.alsairafi - I managed to find a new, better UndoManager that does not depend on time, but the position of edits. http://tips4java.wordpress.com/2008/10/27/compound-undo-manager/ This seem to be much better now for most of the Undo Issues. Status: Fixed I managed to find a new, better UndoManager that does not depend on time, but the position of edits. http://tips4java.wordpress.com/2008/10/27/compound-undo-manager/ This seem to be much better now for most of the Undo Issues. Status: Fixed Nov 08, 2009 r117 (Fix for issue 88 and other Undos - using better UndoManager) committed by ayman.alsairafi - Fix for issue 88 and other Undos - using better UndoManager Fix for issue 88 and other Undos - using better UndoManager Nov 07, 2009 issue 124 (Undo grouping doesn't include first action in the chain) changed by ayman.alsairafi - Hmm.. The change seem to work for this instance. I'll see if it sorts out some other issues. Thanks a million! Status: Accepted Owner: ayman.alsairafi Labels: Type-Enhancement Priority-Medium OpSys-All Component-UI Usability Hmm.. The change seem to work for this instance. I'll see if it sorts out some other issues. Thanks a million! Status: Accepted Owner: ayman.alsairafi Labels: Type-Enhancement Priority-Medium OpSys-All Component-UI Usability Nov 07, 2009 issue 124 (Undo grouping doesn't include first action in the chain) reported by komoku - What steps will reproduce the problem? 1. Type a word fast in the editor, for example "testing". 2. Press Ctrl+Z to undo. 3. "esting" will be deleted, but the "t" will remain. I think undo grouping generally works OK (despite issue 88 ), except for this problem. It seems that the first action in a group is not added to a compound undoable action. I think it is highly probable that this is the main reason behind the perceived "unintuitiveness" of undos reported in issue 88 , but I don't report it there since this is more specific and perhaps the reporter of issue 88 would still want more improvements. Off the top of my head, I think the following fix would work (too late in the night to test it now): in CompoundUndoManager, change this: if (comp == null) { comp = new CompoundEdit(); } To this: if (comp == null) { comp = new CompoundEdit(); startMillis = now; } Note that with the current code, if the last action performed was a long enough time ago, you create a new CompoundEdit (which is fine) but then you enter the if (now - startMillis > IDLE_DELAY_MS) block (since you haven't refreshed startMillis yet by that time), and this makes you immediately END the compound edit, causing the bug. With the added line you trivially prevent this without any side effects. What steps will reproduce the problem? 1. Type a word fast in the editor, for example "testing". 2. Press Ctrl+Z to undo. 3. "esting" will be deleted, but the "t" will remain. I think undo grouping generally works OK (despite issue 88 ), except for this problem. It seems that the first action in a group is not added to a compound undoable action. I think it is highly probable that this is the main reason behind the perceived "unintuitiveness" of undos reported in issue 88 , but I don't report it there since this is more specific and perhaps the reporter of issue 88 would still want more improvements. Off the top of my head, I think the following fix would work (too late in the night to test it now): in CompoundUndoManager, change this: if (comp == null) { comp = new CompoundEdit(); } To this: if (comp == null) { comp = new CompoundEdit(); startMillis = now; } Note that with the current code, if the last action performed was a long enough time ago, you create a new CompoundEdit (which is fine) but then you enter the if (now - startMillis > IDLE_DELAY_MS) block (since you haven't refreshed startMillis yet by that time), and this makes you immediately END the compound edit, causing the bug. With the added line you trivially prevent this without any side effects. Nov 07, 2009 Building Wiki page commented on by ian.phillips - The lexers package is generated by JFlex based on the grammars in the `src/main/jflex/jsyntaxpane/lexers` folder. The maven build will generate the lexers for you, or there is a JFlex plugin for IDEA (which I haven't used, so I don't know how good it is just that it exists ;-) The lexers package is generated by JFlex based on the grammars in the `src/main/jflex/jsyntaxpane/lexers` folder. The maven build will generate the lexers for you, or there is a JFlex plugin for IDEA (which I haven't used, so I don't know how good it is just that it exists ;-) Nov 05, 2009 issue 123 (Undo broken on edits modifying multiple lines) reported by un.compte.pour.tester - 1. Type some lines of text in the default editor. 2. Select those lines and either tab on comment them 3. Undo (Ctrl-z or button) The lines disappear. (they come back when undoing one more time.) using jsyntaxpane-0.9.5-b27.jar (on Debian unstable but it should not be relevent) Imho, the remove and insertString steps of the replace should be grouped together. (Btw, why are you doing a replace on the whole selected part instead of inserts at the beginning of the lines is it more efficient to do one much larger replace spanning many lines ?). Btw, the bug is worse for multiple lines untab (shift-tab) or uncomment : another level of undo does no make the lines reappear ! 1. Type some lines of text in the default editor. 2. Select those lines and either tab on comment them 3. Undo (Ctrl-z or button) The lines disappear. (they come back when undoing one more time.) using jsyntaxpane-0.9.5-b27.jar (on Debian unstable but it should not be relevent) Imho, the remove and insertString steps of the replace should be grouped together. (Btw, why are you doing a replace on the whole selected part instead of inserts at the beginning of the lines is it more efficient to do one much larger replace spanning many lines ?). Btw, the bug is worse for multiple lines untab (shift-tab) or uncomment : another level of undo does no make the lines reappear ! Nov 04, 2009 issue 122 (Quick Find generates error when editor not hosted in JViewPo...) reported by gtanuel - What steps will reproduce the problem? 1. Host the editor in a JTabbedPane 2. Run it, press Ctrl-F, it will complain that it can't cast to JViewPort. I changed the code in actions/gui/QuickFindDialog.java to use JComponent instead of JViewPort, it's working fine now. What steps will reproduce the problem? 1. Host the editor in a JTabbedPane 2. Run it, press Ctrl-F, it will complain that it can't cast to JViewPort. I changed the code in actions/gui/QuickFindDialog.java to use JComponent instead of JViewPort, it's working fine now.

issue 124 (Undo grouping doesn't include first action in the chain) reported by komoku - What steps will reproduce the problem? 1. Type a word fast in the editor, for example "testing". 2. Press Ctrl+Z to undo. 3. "esting" will be deleted, but the "t" will remain. I think undo grouping generally works OK (despite issue 88 ), except for this problem. It seems that the first action in a group is not added to a compound undoable action. I think it is highly probable that this is the main reason behind the perceived "unintuitiveness" of undos reported in issue 88 , but I don't report it there since this is more specific and perhaps the reporter of issue 88 would still want more improvements. Off the top of my head, I think the following fix would work (too late in the night to test it now): in CompoundUndoManager, change this: if (comp == null) { comp = new CompoundEdit(); } To this: if (comp == null) { comp = new CompoundEdit(); startMillis = now; } Note that with the current code, if the last action performed was a long enough time ago, you create a new CompoundEdit (which is fine) but then you enter the if (now - startMillis > IDLE_DELAY_MS) block (since you haven't refreshed startMillis yet by that time), and this makes you immediately END the compound edit, causing the bug. With the added line you trivially prevent this without any side effects.

What steps will reproduce the problem? 1. Type a word fast in the editor, for example "testing". 2. Press Ctrl+Z to undo. 3. "esting" will be deleted, but the "t" will remain. I think undo grouping generally works OK (despite issue 88 ), except for this problem. It seems that the first action in a group is not added to a compound undoable action. I think it is highly probable that this is the main reason behind the perceived "unintuitiveness" of undos reported in issue 88 , but I don't report it there since this is more specific and perhaps the reporter of issue 88 would still want more improvements. Off the top of my head, I think the following fix would work (too late in the night to test it now): in CompoundUndoManager, change this: if (comp == null) { comp = new CompoundEdit(); } To this: if (comp == null) { comp = new CompoundEdit(); startMillis = now; } Note that with the current code, if the last action performed was a long enough time ago, you create a new CompoundEdit (which is fine) but then you enter the if (now - startMillis > IDLE_DELAY_MS) block (since you haven't refreshed startMillis yet by that time), and this makes you immediately END the compound edit, causing the bug. With the added line you trivially prevent this without any side effects.

Yesterday

Last 7 days

Last 30 days

Earlier this year