8289389: Fix warnings: type should also implement hashCode() since it overrides Object.equals() #821

andy-goryachev-oracle · 2022-07-08T19:46:20Z

added missing hashCode() methods

Progress

Change must be properly reviewed (1 review required, with at least 1 Reviewer)
Change must not contain extraneous whitespace
Commit message must refer to an issue

Issue

JDK-8289389: Fix warnings: type should also implement hashCode() since it overrides Object.equals()

Reviewers

Kevin Rushforth (@kevinrushforth - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jfx pull/821/head:pull/821
$ git checkout pull/821

Update a local copy of the PR:
$ git checkout pull/821
$ git pull https://git.openjdk.org/jfx pull/821/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 821

View PR using the GUI difftool:
$ git pr show -t 821

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jfx/pull/821.diff

bridgekeeper · 2022-07-08T19:48:03Z

👋 Welcome back andy-goryachev-oracle! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

mlbridge · 2022-07-08T19:54:10Z

Webrevs

kevinrushforth

This is not a complete review, just one thing I spotted quickly.

modules/javafx.fxml/src/main/java/javafx/fxml/FXMLLoader.java

hjohn · 2022-07-08T23:44:15Z

Just a general note, shouldn't most of these use Objects.hash?

I see a lot of code in the form of:

  if(x == null) return 0;

  return x.hashCode();

And even a few where hashes are calculated manually for two values:

  x.hashCode() * 31 + y.hashCode();

These could be written as:

  Objects.hash(x);  // null check for free
  Objects.hash(x, y);  // free null checks and hashes are merged

hjohn · 2022-07-08T23:45:58Z

modules/javafx.graphics/src/main/java/com/sun/javafx/css/CalculatedValue.java

+            h = 31 * h + value.hashCode();
+        }
+        return h;
+    }


Just an example, but wouldn't: Objects.hash(relative, origin, value) here work just as well?

Honestly, I would just turn this class into a record and everything will be taken care of. I'm not sure it is within the scope of the fix.

Such a change is definitely out of scope. More importantly, we can't use record anywhere in JavaFX until we bump the minimum version of the JDK needed to run JavaFX. I plan to start a discussion on the openjfx-dev mailing list, since it has been something I've wanted to do for a while now.

@hjohn: yes, but at a price: Object.hash(Object ...) incurs overhead by creating a temporary Object[] + boxing of a Boolean.

Ah, I didn't realize you checked how this is optimized by the JIT.

In order to reduce collisions, the hash of each component is typically added to h * 31 even when that hash is 0, whereas you skip the h = 31 * h in the case of null. It might not be a problem in practice, since value and origin are unlikely to collide, being of different types, but you might want to look at it.

In any case, I need to time to look at it, which I won't have until after JavaFX 19 RDP1, so let's leave this until then.

you bring a good point, Kevin, thank you!

I would use Boolean.hashCode(relative); for a boolean.

Kevin, I checked Effective Java 3rd Edition and it also says to use 0 (or some other constant) for null.

Using 0 for a null component of the hash is fine, as long as the accum variable is initialized to something > 0, and as long as you accumulate all values, even a 0. By which I mean that the following is not OK:

int h = 1; if (comp1 != null) h = 31 * h + comp1.hashCode(); if (comp2 != null) h = 31 * h + comp2.hashCode();

nor is this:

int h = (comp1 == null) ? 0 : comp1.hashCode(); h = 31 * h + (comp2 == null) ? 0 : comp2.hashCode();

but this is:

int h = 1; h = 31 * h + (comp1 == null) ? 0 : comp1.hashCode(); h = 31 * h + (comp2 == null) ? 0 : comp2.hashCode();

Anyway, the latest change in this PR is good.

nlisker

Except for one case, looks good. I think that nearly all of the equals implementations are not the best (or just wrong), but it's outside the scope of the PR.

nlisker · 2022-07-19T03:43:51Z

modules/javafx.controls/src/test/java/test/javafx/scene/control/ListViewTest.java

            if (obj == null) return false;
            return id == ((RT22599_DataType)obj).id;


I note that the equals method here is wrong. The hashCode implementation is fine.

This does look dubious. Since this was written for a specific unit test, it might have been intentional to have a "wrong" equals method, but if so, a comment should have been added to that effect. In any case, fixing equals is out of scope for this PR. And given what equals does, the proposed hashCode method is correct.

nlisker · 2022-07-19T03:44:57Z

modules/javafx.controls/src/test/java/test/javafx/scene/control/TableViewTest.java

            if (obj == null) return false;
            return id == ((RT22599_DataType)obj).id;


Same as ListViewTest.

nlisker · 2022-07-19T03:45:37Z

modules/javafx.controls/src/test/java/test/javafx/scene/control/TreeTableViewTest.java

            if (obj == null) return false;
            return id == ((RT22599_DataType)obj).id;


Same as ListViewTest.

nlisker · 2022-07-19T03:46:17Z

modules/javafx.controls/src/test/java/test/javafx/scene/control/TreeViewTest.java

            if (obj == null) return false;
            return id == ((RT22599_DataType)obj).id;


Same as ListViewTest.

this and ListViewTest - the logic in hashCode() follows the logic in equals().
it is fairly safe change, as these objects are never put into a hashtable and never accessed outside of the test context.

nlisker · 2022-07-19T03:50:47Z

modules/javafx.fxml/src/main/java/javafx/fxml/FXMLLoader.java

        return false;
    }


Same as ListViewTest.

in the case of FXMLoader, hashCode() also follows the logic of equals().
However, we could mix something specific to FXMLoader to the hash, to avoid collision with origin URL.

I think this is fine as you now have it. If anyone feels it is worth it, we could file a follow-on issue to look into whether equals should be changed, but I don't know whether it is worth it.

nlisker · 2022-07-19T04:17:18Z

modules/javafx.graphics/src/main/java/com/sun/javafx/css/CalculatedValue.java

+            h = 31 * h + value.hashCode();
+        }
+        return h;
+    }


I would use Boolean.hashCode(relative); for a boolean.

Kevin, I checked Effective Java 3rd Edition and it also says to use 0 (or some other constant) for null.

nlisker · 2022-07-19T04:21:26Z

modules/javafx.media/src/main/java/com/sun/media/jfxmediaimpl/NativeMediaAudioClipPlayer.java

+    public int hashCode() {
+        return sourceClip.hashCode();
+    }


Why are you ignoring the significant fields? equals compares priority, loopCount, volume, balance, rate and pan in addition.

nlisker · 2022-07-19T04:24:42Z

modules/javafx.media/src/main/java/com/sun/media/jfxmediaimpl/NativeMediaAudioClipPlayer.java

+            if (player == null) {
+                return 0;
+            }
+            return player.hashCode();


This is fine, but can also be a ternary return player == null ? 0 : player.hashCode();

hjohn · 2022-07-19T19:55:34Z

I really must wonder at some of these changes. They seem... unnecessary. Just because equals is overridden, doesn't mean hashCode must be as well, especially not if the objects involved are mutable -- it might in fact be better to leave it alone or return a constant value from it (or only use immutable fields).

For example, let's take FXMLLoader. I put it in a HashMap. I modify its location. With a traditionally implemented hashCode that checks the same fields as equals, the object will now fail HashMap contains check, even though I am actually using the same instance. If we had left the default hashCode inplace, it would be found correctly (as the identity hash code will be unchanged). However, if we had used a different instance, then the default identity hashCode would not work either. Returning a constant from hashCode then would "solve" the problem again (at the expense of poor performance in large hash maps).

TLDR; don't use mutable fields in hashCode, if there are only mutable fields, return a constant. Or turn off these too broad warnings that don't really grasp the full impact. I know this is considered a bit of a "holy grail", but it fails in the face of mutable objects that it might be better not to implement hashCode at all and give the wrong impression.

andy-goryachev-oracle · 2022-07-19T20:18:14Z

John, thank you for your comments!

I fully agree with the idea of not putting objects with mutable fields that participate in either equals() or hashCode() into Hashtable. This scenario might still be legitimate, for example by not modifying the fields in question after instantiation.

I also agree that some of the changes might look unnecessary, or perhaps we should either return super.hashCode() or maybe getClass().hashCode() -- not a constant, as hashCode() guarantees some variability, while a constant does not.

I do think, however, that suppressing this warning or turning it off is not a good idea. it forces the developer to think about exactly this issue.

Do you have any specific objections?

hjohn · 2022-07-19T20:27:23Z

John, thank you for your comments!

I fully agree with the idea of not putting objects with mutable fields that participate in either equals() or hashCode() into Hashtable. This scenario might still be legitimate, for example by not modifying the fields in question after instantiation.

Risky, but that's on the user... they'll have to check the source code to see how hashCode is implemented before doing this kind of thing.

I also agree that some of the changes might look unnecessary, or perhaps we should either return super.hashCode() or maybe getClass().hashCode() -- not a constant, as hashCode() guarantees some variability, while a constant does not.

Yes, getClass().hashCode() is recommended, it is constant for all instances of that class.

I do think, however, that suppressing this warning or turning it off is not a good idea. it forces the developer to think about exactly this issue.

Yeah, I suppose you can "surpress" the warning by overriding it with getClass().hashCode() if there are no immutable fields. Even though performance would suffer, it would at least be working correctly all the time if someone accidently put these in a hashed collection. Once the problem is discovered and the developer realizes their mistake, they can always wrap the object to provide a hash code more suitable to their purpose.

I think many of these have equals implemented primarily because they're used as values for properties. These need equals because of how changes are detected.

tomsontom · 2022-07-21T12:29:38Z

I really must wonder at some of these changes. They seem... unnecessary. Just because equals is overridden, doesn't mean hashCode must be as well, especially not if the objects involved are mutable -- it might in fact be better to leave it alone or return a constant value from it (or only use immutable fields).

This not true - there's a contract between hashCode and equals quoting the JavaDoc of hashCode():

If two objects are equal according to the equals(Object) method, then calling the hashCode method on each of the two objects must produce the same integer result.

hjohn · 2022-07-21T13:11:29Z

I really must wonder at some of these changes. They seem... unnecessary. Just because equals is overridden, doesn't mean hashCode must be as well, especially not if the objects involved are mutable -- it might in fact be better to leave it alone or return a constant value from it (or only use immutable fields).

This not true - there's a contract between hashCode and equals quoting the JavaDoc of hashCode():

Thanks, I'm well aware. The issue here is that once a mutable object is stored in a map, you can have either:

Have it find and match the same instance of the object (by not overriding hashCode or calling System.identityHashCode()), regardless of whether that object was mutated in the mean time
Have it find and match any instance (by overriding hashCode and using the same mutable fields as equals uses), but only if the original instance wasn't mutated in the mean time (its bucket will not be adjusted by hash map)
Have it work in all cases, but with terrible performance (by overriding hashCode with the suggested getClass().hashCode() which is effectively a constant)

This isn't mentioned anywhere aside from the small note ("provided no information used in {@code equals} comparisons on the object is modified").

andy-goryachev-oracle · 2022-07-21T18:28:35Z

John:

You do bring a good point, I am sure someone have tripped this trap by adding a mutable object to a hashtable and then mutating it afterwards. We do have some words of caution in the java.util.Map interface javadoc:

"Note: great care must be exercised if mutable objects are used as map keys..."

At the same time, I feel like this discussion goes beyond the scope of this PR, as it falls under the rubric of design decisions in the client code.

Is there anything specific you think should be changed in this PR, or are you against these code changes in principle?

hjohn · 2022-07-21T22:20:30Z

Yes, it's beyond the scope of PR. I suppose people will have to be careful as with all classes that are mutable and have decided to override equals/hashCode -- no reason to stop now I suppose.

kevinrushforth

Looks pretty good now. I noted a couple places that might need a change.

kevinrushforth · 2022-07-22T16:10:18Z

modules/javafx.controls/src/test/java/test/javafx/scene/control/ListViewTest.java

            if (obj == null) return false;
            return id == ((RT22599_DataType)obj).id;


This does look dubious. Since this was written for a specific unit test, it might have been intentional to have a "wrong" equals method, but if so, a comment should have been added to that effect. In any case, fixing equals is out of scope for this PR. And given what equals does, the proposed hashCode method is correct.

kevinrushforth · 2022-07-22T16:11:48Z

modules/javafx.controls/src/test/java/test/javafx/scene/control/TreeViewTest.java

            if (obj == null) return false;
            return id == ((RT22599_DataType)obj).id;


kevinrushforth · 2022-07-22T16:14:54Z

modules/javafx.fxml/src/main/java/javafx/fxml/FXMLLoader.java

        return false;
    }


I think this is fine as you now have it. If anyone feels it is worth it, we could file a follow-on issue to look into whether equals should be changed, but I don't know whether it is worth it.

kevinrushforth · 2022-07-22T16:23:02Z

modules/javafx.graphics/src/main/java/com/sun/javafx/css/CalculatedValue.java

+            h = 31 * h + value.hashCode();
+        }
+        return h;
+    }


Using 0 for a null component of the hash is fine, as long as the accum variable is initialized to something > 0, and as long as you accumulate all values, even a 0. By which I mean that the following is not OK:

int h = 1; if (comp1 != null) h = 31 * h + comp1.hashCode(); if (comp2 != null) h = 31 * h + comp2.hashCode();

nor is this:

int h = (comp1 == null) ? 0 : comp1.hashCode(); h = 31 * h + (comp2 == null) ? 0 : comp2.hashCode();

but this is:

int h = 1; h = 31 * h + (comp1 == null) ? 0 : comp1.hashCode(); h = 31 * h + (comp2 == null) ? 0 : comp2.hashCode();

Anyway, the latest change in this PR is good.

kevinrushforth · 2022-07-22T16:25:41Z

modules/javafx.media/src/main/java/com/sun/media/jfxmediaimpl/NativeMediaAudioClipPlayer.java


+    @Override
+    public int hashCode() {
+        int h = sourceClip.getLocator().getURI().hashCode();


Is URI::hashCode() guaranteed to return a non-zero value? If not, then initializing h to 1 and accumulating this would be better.

kevinrushforth · 2022-07-22T16:27:18Z

modules/javafx.web/src/main/java/com/sun/webkit/graphics/WCRectangle.java


+    @Override
+    public int hashCode() {
+        int h = Float.floatToIntBits(x);


Best to initialize h to 1 and accumulate x.

andy-goryachev-oracle · 2022-07-22T17:00:53Z

excellent suggestions, Kevin - many thanks!

kevinrushforth

The changes look good.

I note that by initializing the hash to CLASSNAME.class.hashCode(), instead of a constant, the hash code produced will not be predictable or stable across different invocations of the JVM. This is entirely permissible by the spec of Object.hashCode(), so it is just a (somewhat) interesting note.

openjdk · 2022-07-22T17:39:57Z

@andy-goryachev-oracle This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

8289389: Fix warnings: type should also implement hashCode() since it overrides Object.equals()

Reviewed-by: kcr

You can use pull request commands such as /summary, /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been 12 new commits pushed to the master branch:

9f636da: Merge
6da05c2: 8290348: TreeTableView jumping to top
54e689c: 8288450: Update attribution in gstreamer.md file
146dfeb: Merge
25e8605: Merge
69ae421: Merge
929c1c7: Merge
e8b9ff1: Merge
a7c345a: 8285691: Additional fix for JDK-8282121
bee2e55: Merge
... and 2 more: https://git.openjdk.org/jfx/compare/e38979b90bd024946c6aac0e5ae32dcc121d2284...master

As there are no conflicts, your changes will automatically be rebased on top of these commits when integrating. If you prefer to avoid this automatic rebasing, please check the documentation for the /integrate command for further details.

As you do not have Committer status in this project an existing Committer must agree to sponsor your change. Possible candidates are the reviewers of this PR (@kevinrushforth, @nlisker) but any other Committer may sponsor as well.

➡️ To flag this PR as ready for integration with the above commit message, type /integrate in a new comment. (Afterwards, your sponsor types /sponsor in a new comment to perform the integration).

andy-goryachev-oracle · 2022-07-22T18:03:09Z

The changes look good.

I note that by initializing the hash to CLASSNAME.class.hashCode(), instead of a constant, the hash code produced will not be predictable or stable across different invocations of the JVM. This is entirely permissible by the spec of Object.hashCode(), so it is just a (somewhat) interesting note.

I'd say this is a desired outcome.

There is a group of vulnerabilities caused by predictive nature of hashCode, for example, a DOS attack which places large number of objects into the same hashtable bin, see
https://stackoverflow.com/questions/8669946/application-vulnerability-due-to-non-random-hash-functions

I know this has no impact on this PR, just wanted to add add to your interesting note, Kevin.

andy-goryachev-oracle · 2022-07-22T18:31:47Z

If there are no objections, I'd like to integrate this PR on Monday (25 July)

andy-goryachev-oracle · 2022-07-25T22:12:12Z

/integrate

openjdk · 2022-07-25T22:13:25Z

@andy-goryachev-oracle
Your change (at version 530bdd9) is now ready to be sponsored by a Committer.

nlisker · 2022-07-25T22:13:27Z

/sponsor

openjdk · 2022-07-25T22:15:00Z

Going to push as commit 075cc80.
Since your change was applied there have been 12 commits pushed to the master branch:

9f636da: Merge
6da05c2: 8290348: TreeTableView jumping to top
54e689c: 8288450: Update attribution in gstreamer.md file
146dfeb: Merge
25e8605: Merge
69ae421: Merge
929c1c7: Merge
e8b9ff1: Merge
a7c345a: 8285691: Additional fix for JDK-8282121
bee2e55: Merge
... and 2 more: https://git.openjdk.org/jfx/compare/e38979b90bd024946c6aac0e5ae32dcc121d2284...master

Your commit was automatically rebased without conflicts.

openjdk · 2022-07-25T22:15:18Z

@nlisker @andy-goryachev-oracle Pushed as commit 075cc80.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

8289389: implemented missing hashCode()

8a37130

openjdk bot added the rfr Ready for review label Jul 8, 2022

kevinrushforth reviewed Jul 8, 2022

View reviewed changes

modules/javafx.fxml/src/main/java/javafx/fxml/FXMLLoader.java Outdated Show resolved Hide resolved

8289389: toExternalForm()

d82e683

hjohn reviewed Jul 8, 2022

View reviewed changes

8289389: minimize the impact of collision

ed53414

kevinrushforth self-requested a review July 12, 2022 11:38

nlisker suggested changes Jul 19, 2022

View reviewed changes

andy-goryachev-oracle added 2 commits July 19, 2022 08:12

Merge remote-tracking branch 'origin/master' into 8289389.hash

58208e2

8289389: review comments

dd9e60a

kevinrushforth reviewed Jul 22, 2022

View reviewed changes

8289389: review comments

530bdd9

kevinrushforth approved these changes Jul 22, 2022

View reviewed changes

openjdk bot added the ready Ready to be integrated label Jul 22, 2022

openjdk bot added the sponsor Ready to sponsor label Jul 25, 2022

openjdk bot added the integrated Pull request has been integrated label Jul 25, 2022

openjdk bot closed this Jul 25, 2022

openjdk bot removed ready Ready to be integrated rfr Ready for review sponsor Ready to sponsor labels Jul 25, 2022

andy-goryachev-oracle deleted the 8289389.hash branch July 25, 2022 22:17

		if (obj == null) return false;
		return id == ((RT22599_DataType)obj).id;

8289389: Fix warnings: type should also implement hashCode() since it overrides Object.equals() #821

8289389: Fix warnings: type should also implement hashCode() since it overrides Object.equals() #821

Uh oh!

Conversation

andy-goryachev-oracle commented Jul 8, 2022 • edited by openjdk bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewers

Reviewing

Uh oh!

bridgekeeper bot commented Jul 8, 2022

Uh oh!

mlbridge bot commented Jul 8, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Webrevs

Uh oh!

kevinrushforth left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

hjohn commented Jul 8, 2022

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!

nlisker 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!

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!

hjohn commented Jul 19, 2022

Uh oh!

andy-goryachev-oracle commented Jul 19, 2022

Uh oh!

hjohn commented Jul 19, 2022

Uh oh!

tomsontom commented Jul 21, 2022

Uh oh!

hjohn commented Jul 21, 2022

Uh oh!

andy-goryachev-oracle commented Jul 8, 2022 •

edited by openjdk bot

Loading

mlbridge bot commented Jul 8, 2022 •

edited

Loading

openjdk bot commented Jul 22, 2022 •

edited

Loading