[Fix #192]: Add a new "Export to TXT for AI" button by kgibm · Pull Request #193 · eclipse-mat/mat

kgibm · 2026-06-30T20:21:24Z

Fixes #192

Add a new API AIDetailsProvider class that may be optionally set in a query's getResultMetaData.
Bump MAT to 1.18.0 for the new API.
Add an AIDetailsProvider to the ThreadOverviewQuery so that 2 depths of the tree are exported including all stack frames and all stack frame locals.
Add a new AITextOutputter and associated export button (Export to TXT for AI...). Uses the extension .text to differentiate it from the old TextOutputter that uses .txt. AITextOutputter inherits from TextOutputter and is largely the same with the addition of a prefix to help prompt AI consumers by explaining the format of the output (e.g. nesting with pipe character prefixes) and optionally an additional prefix through AIDetailsProvider to further prompt AI consumers by explaining the output.
The other differences between AITextOutputter and TextOutputter are dictated by the logic in ExportActions: the totals row is suppressed as that adds a lot of fluff especially with expanded rows, and, most importantly, check if the ResultMetaData includes a non-default tree export depth. By default, we don't expand trees. Unlike the TextOutputter, the AITextOutputter ignores current expansion in the Tree and always has a uniform output.
I installed a fresh Eclipse 2026-06 and after loading the project and without doing anything, lots of Eclipse .project, .classpath, .prefs, etc. files were updated. I reviewed the updates and they seemed benign/reasonable, so I guess since we're still keeping Eclipse project files checked in, I'm guessing it's reasonable to commit all of these changes. I'm not sure why Eclipse changed org.eclipse.jdt.core.compiler.release from enabled to disabled.
Add a test that checks that the AITextOutputter works and includes the thread overview query prefix in its output.

kgibm · 2026-06-30T20:22:33Z

@jasonk000 @krumts Hey guys, if you have a chance, please review.

Example output (snippet):

List of all known threads and key metadata such as the thread name, if available in the dump. In addition, if available in the dump, nested below each thread is a list of the stack frames of the thread at the time of the dump and below each stack frame, each stack frame local actively referenced by that stack frame.

The result is printed in an ASCII table form with rows and columns. There is a header line with columns separated by a pipe character, followed by a dividing line, and followed by each row of the output with columns separated by a pipe character. If the result is a tree, then the level of nesting is represented by the number of times a row is prefixed by a pipe character.

Object / Stack Frame                                                                                                                                                                                                                                                                                                 |Name                                                                                | Shallow Heap | Retained Heap | Max. Locals' Retained Heap |Context Class Loader                                          |Is Daemon| Priority |State                                                 | State value
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
java.lang.Thread @ 0xfe144cd8                                                                                                                                                                                                                                                                                        |Default Executor-thread-2                                                           |          112 |        61,040 |                            |org.eclipse.osgi.internal.framework.ContextFinder @ 0xe01f13f0|true     |        5 |[alive, in object wait, waiting, waiting indefinitely]|       0x191
|- at java.lang.Object.wait(J)V (Object.java(Native Method))                                                                                                                                                                                                                                                         |                                                                                    |              |               |                            |                                                              |         |          |                                                      |            
|- at com.ibm.ws.threading.internal.BoundedBuffer.waitGet_(J)V (BoundedBuffer.java:189)                                                                                                                                                                                                                              |                                                                                    |              |               |                         16 |                                                              |         |          |                                                      |            
|  |- <Java Local> com.ibm.ws.threading.internal.BoundedBuffer$GetQueueLock @ 0xfe144e48                                                                                                                                                                                                                             |                                                                                    |           16 |            16 |                            |                                                              |         |          |                                                      |            
[...]

kgibm · 2026-06-30T20:26:06Z

1. Add a new API AIDetailsProvider class that may be optionally set in a query's getResultMetaData. 1. Bump MAT to 1.18.0 for the new API. 1. Add an AIDetailsProvider to the ThreadOverviewQuery so that 2 depths of the tree are exported including all stack frames and all stack frame locals. 1. Add a new AITextOutputter and associated export button (Export to TXT for AI...). Uses the extension .text to differentiate it from the old TextOutputter that uses .txt. AITextOutputter inherits from TextOutputter and is largely the same with the addition of a prefix to help prompt AI consumers by explaining the format of the output (e.g. nesting with pipe character prefixes) and optionally an additional prefix through AIDetailsProvider to further prompt AI consumers by explaining the output. 1. The other differences between AITextOutputter and TextOutputter are dictated by the logic in ExportActions: the totals row is suppressed as that adds a lot of fluff especially with expanded rows, and, most importantly, check if the ResultMetaData includes a non-default tree export depth. By default, we don't expand trees. Unlike the TextOutputter, the AITextOutputter ignores current expansion in the Tree and always has a uniform output. 1. I installed a fresh Eclipse 2026-06 and after loading the project and without doing anything, lots of Eclipse .project, .classpath, .prefs, etc. files were updated. I reviewed the updates and they seemed benign/reasonable, so I guess since we're still keeping Eclipse project files checked in, I'm guessing it's reasonable to commit all of these changes. I'm not sure why Eclipse changed org.eclipse.jdt.core.compiler.release from enabled to disabled. 1. Add a test that checks that the AITextOutputter works and includes the thread overview query prefix in its output. Signed-off-by: Kevin Grigorenko <kevin.grigorenko@us.ibm.com>

Signed-off-by: Kevin Grigorenko <kevin.grigorenko@us.ibm.com>

jasonk000 · 2026-07-01T20:41:02Z

@kgibm , I haven't looked in detail yet - but I suggest we name it "Export to Markdown (.md)" or similar, and emit standard markdown. Markdown is lingua franca standard text format for AI tooling.

kgibm · 2026-07-02T12:50:26Z

That seems reasonable although will require some larger changes. I don't think Markdown has a concept of nested tables (for IResultTree) which is the main thing needed for the threads export. I guess we could do a nested bulleted list, with each bullet item having all non-blank column values as a comma separated list with column name prefix, e.g.:

Object / Stack Frame: java.lang.Thread @ 0xfe144cd8, Name: Default Executor-thread-2, Shallow Heap: 112, Retained Heap: 61,040 [...]
- Object / Stack Frame: at java.lang.Object.wait(J)V (Object.java(Native Method))
  - Object / Stack Frame: com.ibm.ws.threading.internal.BoundedBuffer$GetQueueLock @ 0xfe144e48, Shallow Heap: 16, Retained Heap: 16

jasonk000

Mostly, looks OK - only question is are we missing a null check at at result.getMetadata(), and should the file list include .txt?.

jasonk000 · 2026-09-19T16:36:10Z

@@ -0,0 +1,4 @@
+activeProfiles=


This and related files, to 1.18 bump, are required for this PR?

Yeah because of the new addAIDetailsProvider method added to the ResultMetaData API

jasonk000 · 2026-09-19T16:37:58Z

 	</projects>
 	<buildSpec>
+		<buildCommand>
+			<name>org.eclipse.m2e.core.maven2Builder</name>


Should be different PR, not blocking.

These changes came as part of the newer Eclipse automatically. They seem benign as far as I can tell

jasonk000 · 2026-09-19T16:40:59Z

 org.eclipse.jdt.core.compiler.problem.reportPreviewFeatures=warning
-org.eclipse.jdt.core.compiler.release=disabled
-org.eclipse.jdt.core.compiler.source=1.8
+org.eclipse.jdt.core.compiler.release=enabled


We could align these across all plugins -- any reason not to? Not sure if J9 impact.

Sure, I see no reason not to switch everything explicitly over to Java 21

jasonk000 · 2026-09-19T16:42:13Z

 eclipse.preferences.version=1
-org.eclipse.jdt.core.compiler.codegen.targetPlatform=1.8
-org.eclipse.jdt.core.compiler.compliance=1.8
+org.eclipse.jdt.core.compiler.codegen.targetPlatform=17


Similar here, we could make them all 21 for consistency unless there is a blocker.

jasonk000 · 2026-09-19T16:44:00Z

+    @Override
+    public void embedd(Context context, IResult result, Writer writer) throws IOException
+    {
+        AIDetailsProvider aiDetailsProvider = result.getResultMetaData().getAIDetailsProvider();


Null check required here for result.getResultMetadata().

jasonk000 · 2026-09-19T16:50:36Z

+        {
+            ExportDialog dialog = new ExportDialog(control.getShell(), //
+                            new String[] { Messages.ExportActions_PlainAiText }, //
+                            new String[] { "*.text" });//$NON-NLS-1$


this is a bit odd to me, should be .txt, or both .txt and .text? :)

I'm about to commit new changes and it'll be .md

jasonk000 · 2026-09-19T16:56:02Z

re: Markdown nested. Good point that it does not support nested tables. I looked online, this stackoverflow says that we should put html inside md table cell, but that does not seem like an improvement in end user readability.

kgibm · 2026-09-23T15:02:28Z

Thanks @jasonk000 for the review. I do still plan on converting this PR to markdown but just haven't found the cycles yet. Do you think it's okay to use nested bulleted list to simulate the table? HTML tables would work but one of the use cases I wanted for this was a simple, human-readable output (as well as LLM readable).

jasonk000 · 2026-09-23T15:31:28Z

one of the use cases I wanted for this was a simple, human-readable output (as well as LLM readable).

Yes I think bullets are a good choice. If it is for human + LLM consumption (vs structured like CSV!) then we have flexibility too. The structure can be tweaked over time as well, since the reader (human or LLM) can adapt.

Signed-off-by: Kevin Grigorenko <kevin.grigorenko@us.ibm.com>

kgibm · 2026-10-08T19:35:28Z

PR updated with new Export to Markdown for AI... export option. Example snippet for the thread stacks view:

Test.md

Export of a list of all known threads and key metadata such as the thread name, if available in the dump. In addition, if available in the dump, nested below each thread is a list of the stack frames of the thread at the time of the dump and below each stack frame, each stack frame local actively referenced by that stack frame.

The following Eclipse Memory Analyzer Tool markdown output represents a tree of result data from a query. Each root element in the tree is represented by a bulleted list item. If a tree item has children, then each child will be a bulleted list item with an additional level of indentation logically underneath that item. Each tree item is printed as a comma separated list of column name and value pairs separated by a colon.

* Object / Stack Frame: org.eclipse.osgi.framework.eventmgr.EventManager$EventThread @ 0x600805b58, Name: Start Level: Equinox Container: 4ed5060b-ff91-4852-a83b-ec794e84c2a8, Shallow Heap: 120, Retained Heap: 4,952, Context Class Loader: org.eclipse.osgi.internal.framework.ContextFinder @ 0x6001448b0, Is Daemon: true, Priority: 5, State: [alive, in object wait, waiting, waiting indefinitely], State value: 0x191
    * Object / Stack Frame: at java.lang.Object.wait0(J)V (Object.java(Native Method))
    * Object / Stack Frame: at java.lang.Object.wait(J)V (Object.java:366)
    * Object / Stack Frame: at java.lang.Object.wait()V (Object.java:339)
    * Object / Stack Frame: at org.eclipse.osgi.framework.eventmgr.EventManager$EventThread.getNextEvent()Lorg/eclipse/osgi/framework/eventmgr/EventManager$EventThread$Queued; (EventManager.java:396), Max. Locals' Retained Heap: 4,952
        * Object / Stack Frame: org.eclipse.osgi.framework.eventmgr.EventManager$EventThread @ 0x600805b58  Start Level: Equinox Container: 4ed5060b-ff91-4852-a83b-ec794e84c2a8, Shallow Heap: 120, Retained Heap: 4,952
    * Object / Stack Frame: at org.eclipse.osgi.framework.eventmgr.EventManager$EventThread.run()V (EventManager.java:337), Max. Locals' Retained Heap: 4,952
        * Object / Stack Frame: org.eclipse.osgi.framework.eventmgr.EventManager$EventThread @ 0x600805b58  Start Level: Equinox Container: 4ed5060b-ff91-4852-a83b-ec794e84c2a8, Shallow Heap: 120, Retained Heap: 4,952

Signed-off-by: Kevin Grigorenko <kevin.grigorenko@us.ibm.com>

kgibm requested review from jasonk000 and krumts as code owners June 30, 2026 20:21

kgibm added 3 commits July 1, 2026 08:04

Fix formatting

363f256

Signed-off-by: Kevin Grigorenko <kevin.grigorenko@us.ibm.com>

Fix new AITextOutputter test

14b92f7

Signed-off-by: Kevin Grigorenko <kevin.grigorenko@us.ibm.com>

kgibm force-pushed the issue192 branch from 9318f66 to 14b92f7 Compare July 1, 2026 13:04

jasonk000 approved these changes Sep 19, 2026

View reviewed changes

Convert AI export to Markdown

fad93a2

Signed-off-by: Kevin Grigorenko <kevin.grigorenko@us.ibm.com>

kgibm requested a review from jasonk000 October 8, 2026 19:29

Update jcstress to Java 21 as well

85c0636

Signed-off-by: Kevin Grigorenko <kevin.grigorenko@us.ibm.com>

Conversation

kgibm commented Jun 30, 2026

Uh oh!

kgibm commented Jun 30, 2026

Uh oh!

kgibm commented Jun 30, 2026

Uh oh!

jasonk000 commented Jul 1, 2026

Uh oh!

kgibm commented Jul 2, 2026

Uh oh!

jasonk000 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jasonk000 commented Sep 19, 2026

Uh oh!

kgibm commented Sep 23, 2026

Uh oh!

jasonk000 commented Sep 23, 2026

Uh oh!

kgibm commented Oct 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants