8367530: The exhaustiveness errors could be improved #27256

lahodaj · 2025-09-12T11:45:33Z

Consider code like:

package test;
public class Test {
    private int test(Root r) {
        return switch (r) {
            case Root(R2(R1 _), R2(R1 _)) -> 0;
            case Root(R2(R1 _), R2(R2 _)) -> 0;
            case Root(R2(R2 _), R2(R1 _)) -> 0;
        };
    }
    sealed interface Base {}
    record R1() implements Base {}
    record R2(Base b1) implements Base {}
    record Root(R2 b2, R2 b3) {}
}

This is missing a case for Root(R2(R2 _), R2(R2 _)). javac will produce an error correctly, but the error is not very helpful:

$ javac test/Test.java
.../test/Test.java:4: error: the switch expression does not cover all possible input values
        return switch (r) {
               ^
1 error

The goal of this PR is to improve the error, at least in some cases to something along these lines:

$ javac test/Test.java 
.../test/Test.java:4: error: the switch expression does not cover all possible input values
        return switch (r) {
               ^
  missing patterns:
      Root(R2(R2 _), R2(R2 _))
1 error

The (very simplified) way it works in a recursive (or induction) way:

start with defining the missing pattern as the binding pattern for the selector type. This would certainly exhaust the switch.
for a current missing pattern, try to enhance it:
- if the current type is a sealed type, try to expand to its (direct) permitted subtypes. Remove those that are not needed.
- if the current (binding pattern) type is a record type, expand it to a record type, generate all possible combinations of its component types based on sealed hierarchies. Remove those that are not needed.

This approach relies heavily on our ability to compute exhaustiveness, which is evaluated repeatedly in the process.

There are some cases where the algorithm does not produce ideal results (see the tests), but overall seems much better than what we have now.

Another significant limitation is the speed of the process. Evaluating exhaustiveness is not a fast process, and this algorithm evaluates exhaustiveness repeatedly, potentially for many combinations of patterns (esp. for record patterns). So part of the proposal here is to have a time deadline for the computation. The default is 5s, and can be changed by -XDexhaustivityTimeout=<timeout-in-ms>.

There's also an open possibility for select tools to delay the more detailed computation to some later time, although that would need to be tried and evaluated.

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-8367530: The exhaustiveness errors could be improved (Enhancement - P2)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/27256/head:pull/27256
$ git checkout pull/27256

Update a local copy of the PR:
$ git checkout pull/27256
$ git pull https://git.openjdk.org/jdk.git pull/27256/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 27256

View PR using the GUI difftool:
$ git pr show -t 27256

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/27256.diff

Using Webrev

Link to Webrev Comment

…e class

bridgekeeper · 2025-09-12T11:46:59Z

👋 Welcome back jlahoda! A progress list of the required criteria for merging this PR into pr/27253 will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2025-09-12T11:47:34Z

❗ This change is not yet ready to be integrated.
See the Progress checklist in the description for automated requirements.

openjdk · 2025-09-12T11:48:06Z

⚠️ @lahodaj This pull request contains merges that bring in commits not present in the target repository. Since this is not a "merge style" pull request, these changes will be squashed when this pull request in integrated. If this is your intention, then please ignore this message. If you want to preserve the commit structure, you must change the title of this pull request to Merge <project>:<branch> where <project> is the name of another project in the OpenJDK organization (for example Merge jdk:master).

openjdk · 2025-09-12T11:48:29Z

@lahodaj The following label will be automatically applied to this pull request:

compiler

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing list. If you would like to change these labels, use the /label pull request command.

lahodaj · 2025-11-07T13:38:02Z

Thanks for the comments so far. I think I've updated the code where needed based on them, and tried to respond. Thanks!

biboudis · 2025-11-13T11:07:13Z

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/Flow.java

            tree.isExhaustive = tree.hasUnconditionalPattern ||
                                TreeInfo.isErrorEnumSwitch(tree.selector, tree.cases);
            if (exhaustiveSwitch) {
-                tree.isExhaustive |= exhaustiveness.exhausts(tree.selector, tree.cases);


This |= turned into =. Is that correct? I think so. Now it is guarded by tree.isExhaustive itself right?

Yes, this is intentional. There is one more additional if (!tree.isExhaustive) test below, which replaces the use of the or (but is faster in case we know the switch is exhaustive).

biboudis · 2025-11-13T11:14:37Z

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

+    }
+
+    /* The stricness of determining the equivalent of patterns, used in
+     * nestedComponentsEquivalent.


Also in computeCoverage. Can you provide an example/sentence of two strictly equivalent patterns and two loosely equivalent?

biboudis · 2025-11-13T11:21:36Z

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

+    /*
+     * Based on {@code basePattern} generate new {@code RecordPattern}s such that all
+     * components instead of {@code replaceComponent}th component, which is replaced
+     * with values from {@code updatedNestedPatterns}. Resulting {@code RecordPatterns}s
+     * are sent to {@code target}.
+     */


/* * Using {@code basePattern} as a starting point, generate new {@code * RecordPattern}s, such that all corresponding components but one, are the * same. The component described by the {@code replaceComponent} index is * replaced with all {@code PatternDescription}s taken from {@code * updatedNestedPatterns} and the resulting {@code RecordPatterns}s are sent * to {@code target}. */

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

biboudis · 2025-11-13T11:30:45Z

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

+                }
+            }
+
+            //assert?


Address this todo?

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

Co-authored-by: Aggelos Biboudis <biboudis@gmail.com>

mcimadamore · 2025-11-14T12:52:15Z

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

+
+                return currentMissingPatterns;
+            } else if ((bp.type.tsym.flags_field & Flags.RECORD) != 0 &&
+                       //only expand record types into record patterns if there's a chance it may change the outcome


I agree with this choice

mcimadamore · 2025-11-14T13:18:46Z

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

+
+                removeUnnecessaryPatterns(selectorType, bp, basePatterns, inMissingPatterns, combinatorialPatterns);
+
+                CoverageResult coverageResult = computeCoverage(targetType, combinatorialPatterns, PatternEquivalence.LOOSE);


How can the expansion of a record pattern not cover? E.g. if we start with an exhaustive pattern R r and we expand R into all its components R(A1, B1, C1), R(A2, B2, C2) ... -- how can we get into a situation where we're no longer exhaustive?

The purpose of the code here (and a few lines below) is to merge "unnecesarily" specific patterns inside the missing patterns into their more generic supertypes. Like if at a specific place in the record patterns, there are both A and B, permitted subtypes of Base, we could merge and replace with Base here. But I understand it is somewhat matter of opinion on what is better.

Note the "unnecessary" combinatorialPatterns (i.e. those that are covered by the user provided patterns) are already removed here, and computeCoverage is ran on combinatorialPatterns, so those may not cover the original type, because they no longer contain anything covered by the user-provided patterns.

One example where this changes the outcomes is:

jdk/test/langtools/tools/javac/patterns/ExhaustivenessConvenientErrors.java

Line 214 in da13071

public void testComplex1(Path base) throws Exception {

where the current outcome is:

test.Test.Root(test.Test.R2 _, test.Test.Base _, test.Test.Base _) test.Test.Root(test.Test.R3 _, test.Test.Base _, test.Test.Base _)

but if I disable this merging, we would get:

test.Test.Root(test.Test.R3 _, test.Test.R1 _, test.Test.R2 _) test.Test.Root(test.Test.R2 _, test.Test.R1 _, test.Test.R2 _) test.Test.Root(test.Test.R2 _, test.Test.R1 _, test.Test.R3 _) test.Test.Root(test.Test.R3 _, test.Test.R2 _, test.Test.R1 _) test.Test.Root(test.Test.R3 _, test.Test.R1 _, test.Test.R1 _) test.Test.Root(test.Test.R3 _, test.Test.R2 _, test.Test.R2 _) test.Test.Root(test.Test.R3 _, test.Test.R1 _, test.Test.R3 _) test.Test.Root(test.Test.R3 _, test.Test.R3 _, test.Test.R3 _) test.Test.Root(test.Test.R2 _, test.Test.R2 _, test.Test.R3 _) test.Test.Root(test.Test.R2 _, test.Test.R3 _, test.Test.R1 _) test.Test.Root(test.Test.R3 _, test.Test.R2 _, test.Test.R3 _) test.Test.Root(test.Test.R3 _, test.Test.R3 _, test.Test.R1 _) test.Test.Root(test.Test.R3 _, test.Test.R3 _, test.Test.R2 _) test.Test.Root(test.Test.R2 _, test.Test.R1 _, test.Test.R1 _) test.Test.Root(test.Test.R2 _, test.Test.R2 _, test.Test.R2 _) test.Test.Root(test.Test.R2 _, test.Test.R2 _, test.Test.R1 _) test.Test.Root(test.Test.R2 _, test.Test.R3 _, test.Test.R2 _) test.Test.Root(test.Test.R2 _, test.Test.R3 _, test.Test.R3 _)

There may be a way to better control this merging (given the original user-provided pattern has _ on both the second and third component), but given how coverage works for record patterns, it is a bit tricky.

Thanks, I think my question originated from a wrong assumption on how your code behaved -- after some debugging sessions I've seen cases where the set of patterns does not cover the target

mcimadamore · 2025-11-14T13:28:29Z

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

+                                                    incompletePatterns,
+                                                    Set.of(defaultPattern));
+        } catch (TimeoutException ex) {
+            return ex.missingPatterns != null ? ex.missingPatterns : Set.of();


Instead of a timeout, I wonder if you could instead cut the recursion at a specific threshold. It seems to me that recursing more will provide more precision at the nested level, so it's a trade off between when do we want to stop.

Overload resolution provides some kind of precedent:

error: incompatible types: String cannot be converted to int m("Hello"); ^ Note: Some messages have been simplified; recompile with -Xdiags:verbose to get full output 1 error

(We "compress" the diagnostic whenever we can somehow figure out if an overload is "better" than the others). Then if you provide the option, you get the full thing:

error: no suitable method found for m(String) m("Hello"); ^ method Test.m() is not applicable (actual and formal argument lists differ in length) method Test.m(int) is not applicable (argument mismatch; String cannot be converted to int) method Test.m(int,int) is not applicable (actual and formal argument lists differ in length) method Test.m(int,int,int) is not applicable (actual and formal argument lists differ in length) method Test.m(int,int,int,int) is not applicable (actual and formal argument lists differ in length) method Test.m(int,int,int,int,int) is not applicable (actual and formal argument lists differ in length) method Test.m(int,int,int,int,int,int) is not applicable (actual and formal argument lists differ in length)

But, also, maybe putting an upper bound on the recursion, no matter what, might be a good idea?

While I agree that having a timeout in weird/non-standard in javac, I believe it is the first time when we have a process that a) can run for a very long time; b) is not required for correctness. In all other cases I can recall, if some process is slow (including e.g. verifying exhaustiveness), then it is required for correctness. And so we cannot skip it based on criteria like time.

The timeout here provides a way to say how much real-world time is the user willing to invest to get the outcome - if more time is invested, more detailed missing patterns may possibly be computed. It is a somewhat weird approach for javac, but it is a limit that (I think) the user and javac can agree on.

We could introduce a limit on e.g. the depth to which we expand, but adding one level of nesting may affect performance significantly or almost not at all, depending on e.g. the record type form/shape.

E.g. having a record with many components, where each component is a sealed type with many permitted subtypes, one level of nesting may lead to a high number of newly generated patterns possibly taking a lot of time to go through them. But having a record that has a single component that is not a sealed type should only generate one pattern, and so have minimal impact.

mcimadamore · 2025-11-14T13:31:01Z

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

+                                Symbol enumType = bp.type.tsym;
+                                return enum2Constants.get(enumType).stream().map(c -> enumType.toString() + "." + c.name);
+                            } else {
+                                return Stream.of(pd.toString());


As discussed offline, eager string generation will render the diagnostic arguments completely opaque to the formatter. This would mean that no where clauses will be generated, and unambiguous qualifiers will not be omitted.

Also, don't forget JCDiagnostic.MultilineDiagnostic, which we use in overloads to render repetitive fragments in tabular format.

mcimadamore · 2025-11-14T13:35:01Z

test/langtools/tools/javac/patterns/ExhaustivenessConvenientErrors.java

+                           case Triple(_, A _, _) -> 0;
+                           case Triple(_, _, A _) -> 0;
+                           case Triple(A p, C(Nested _, NestedBaseA _), _) -> 0;
+                           case Triple(A p, C(Nested _, NestedBaseB _), C(Underneath _, NestedBaseA _)) -> 0;


Where is Underneath defined?

That's a mistake, it should be Nested. Fixed here:
08fdb6d
Thanks!

mcimadamore · 2025-11-14T14:11:39Z

test/langtools/tools/javac/patterns/ExhaustivenessConvenientErrors.java

+               }
+               """,
+               "test.Test.Root(test.Test.R2 _, test.Test.R2(test.Test.Base _, test.Test.R2 _), test.Test.R2(test.Test.R2 _, test.Test.Base _))");
+               //ideally, the result would be as follow, but it is difficult to split Base on two distinct places:


With recursion it should be possible -- but I guess the problem becomes, when do you stop?

e.g.
{ Base } -> { R1, R2 } -> { R1, R2(Base, Base) } -> ...

Perhaps an idea here (not sure how crazy -- possibly a lot) would be to run an analysis on the original patterns as defined in the source code, and determine the max depth, and then expand "up to that depth".

So, in the above case, you would know that when you get to {R1, R2}, it's "good enough" for the level of depth that is present in the code.

The problem here for the current approach is that the current approach only expands one pattern at a time, but to get the optimal result here, both the test.Test.Base _ patterns would need to be expanded at the same time (as some combinations of these two are required and some are not).

I was thinking of doing a full expansion originally, but I was worried a bit that for fairly small inputs, the number of combinations could be too high. For example, a newly added:

jdk/test/langtools/tools/javac/patterns/ExhaustivenessConvenientErrors.java

Line 512 in 08fdb6d

return switch (b) {

a relatively naive expansion (expanding components that have record patterns in the original pattern) would lead to a few hundreds of patterns, unless I am mistaken. There may be a way to limit the expansions, though.

We can look into this deeper, surely.

mcimadamore

I think this is impressive work. In "normal" situations (e.g. switches not too big, or too nested) I can easily imagine the new diagnostics to be a life saver.

There's of course a lot of tinkering and followup work that might be possible, to improve the performance of the analysis, or to fine tune the expansion more to the shape of the code.

For now, I think I see two more general issues that stick out:

the early flattening to string, which bypasses the diagnostic formatter -- but that should be easy to fix
the timeout-based strategy. We don't have anything like that anywhere else in the compiler. I think it would be preferrable to have a "variable rate" of accuracy, and maybe limit how the analysis is ran, unless the user really wants to discover every detail. But I'm not sure it's always possible. Cutting on recursion might be a good way to put a ceiling on complexity. Another avenue might be to refuse to expand sealed types that have more than N ermitted subclasses.

mcimadamore · 2025-11-14T15:24:20Z

src/jdk.compiler/share/classes/com/sun/tools/javac/comp/ExhaustivenessComputer.java

+                //if there is a binding pattern at a place where the original based patterns
+                //have a record pattern, try to expand the binding pattern into a record pattern
+                //create all possible combinations of record pattern components:
+                Type[] componentTypes = ((ClassSymbol) bp.type.tsym).getRecordComponents()


This same "pattern" of code seems repeated in makePatternDescription -- ideally there should be a way to compute the "instantiated" component types

openjdk-notifier · 2025-11-18T14:06:09Z

The parent pull request that this pull request depends on has now been integrated and the target branch of this pull request has been updated. This means that changes from the dependent pull request can start to show up as belonging to this pull request, which may be confusing for reviewers. To remedy this situation, simply merge the latest changes from the new target branch into this pull request by running commands similar to these in the local repository for your personal fork:

git checkout JDK-8367530
git fetch https://git.openjdk.org/jdk.git master
git merge FETCH_HEAD
# if there are conflicts, follow the instructions given by git merge
git commit -m "Merge master"
git push

openjdk · 2025-11-18T14:08:37Z

@lahodaj this pull request can not be integrated into master due to one or more merge conflicts. To resolve these merge conflicts and update this pull request you can run the following commands in the local repository for your personal fork:

git checkout JDK-8367530
git fetch https://git.openjdk.org/jdk.git master
git merge FETCH_HEAD
# resolve conflicts and follow the instructions given by git merge
git commit -m "Merge master"
git push

davidalayachew · 2025-11-20T04:17:11Z

WOAH

Is this fabled "Example Generator" that was talked about in the past?

Does this mean that we will be granted at least one example from now on when our Switch is non-exhaustive?

If so, this is amazing news! Wow!

…ternDescriptions through the diagnostics to format it at the end.

openjdk · 2025-11-25T17:59:05Z

@lahodaj build has been added to this pull request based on files touched in new commit(s).

lahodaj added 17 commits August 4, 2025 17:42

A attempt to create more user-friendly errors for switch exaustiveness.

3d6728c

Improvements.

cad665e

8364991: Incorrect not-exhaustive error

2cfbc71

Merge branch 'JDK-8364991' into exhaustiveness-errors

36e337f

Improving test debuggability.

e4f655d

Better name.

8647a74

Merge branch 'JDK-8364991' into exhaustiveness-errors

2f569e3

Avoid computing exhaustiveness unnecessarily, cleanup.

5f70d8b

Merge branch 'master' into JDK-8364991

74c72e6

Adjusting to spec.

8735a3d

Merge branch 'JDK-8364991' into exhaustiveness-errors

7320a5a

Fixing the exhaustiveness search.

29428b3

Adding test.

bed37c4

8367499: Refactor exhaustiveness computation from Flow into a separat…

0d77887

…e class

Adding doc comment.

6963835

Merge branch 'JDK-8367499' into JDK-8367530

d54e53d

Cleanup, making timeout work.

31d3990

openjdk bot added the compiler compiler-dev@openjdk.org label Sep 12, 2025

lahodaj added 8 commits September 25, 2025 17:32

Enabling disabled test.

ce83803

Simplifying the code as suggested.

0ee6086

Cleanup.

c78f696

Cleanup.

9066c51

Cleanup.

789f09e

Fixing tests.

78af109

Merge branch 'JDK-8364991' into JDK-8367499

e69c20e

Merge branch 'JDK-8367499' into JDK-8367530

d35773c

Reflecting review comments.

50ddab0

biboudis reviewed Nov 13, 2025

View reviewed changes

lahodaj and others added 2 commits November 13, 2025 18:05

Apply suggestions from code review

75c3cb0

Co-authored-by: Aggelos Biboudis <biboudis@gmail.com>

Cleanup, reflecting review feedback.

8c48cf8

openjdk bot removed the rfr Pull request is ready for review label Nov 13, 2025

Fixing trailing whitespaces.

da13071

openjdk bot added the rfr Pull request is ready for review label Nov 13, 2025

mcimadamore reviewed Nov 14, 2025

View reviewed changes

openjdk-notifier bot changed the base branch from pr/27247 to master November 18, 2025 14:06

openjdk bot added the merge-conflict Pull request has merge conflict with target branch label Nov 18, 2025

Merge branch 'master' into JDK-8367530

3e4af72

openjdk bot removed the merge-conflict Pull request has merge conflict with target branch label Nov 18, 2025

Fixing cases that are based on review feedback.

08fdb6d

lahodaj added 3 commits November 25, 2025 15:15

Reflecting review feedback: using multi-line diagnostics; passing Pat…

cee029c

…ternDescriptions through the diagnostics to format it at the end.

Merge branch 'master' into JDK-8367530-3

cfa2efa

Fixing test.

1e6a018

openjdk bot added the build build-dev@openjdk.org label Nov 25, 2025


		removeUnnecessaryPatterns(selectorType, bp, basePatterns, inMissingPatterns, combinatorialPatterns);

		CoverageResult coverageResult = computeCoverage(targetType, combinatorialPatterns, PatternEquivalence.LOOSE);

8367530: The exhaustiveness errors could be improved #27256

Are you sure you want to change the base?

8367530: The exhaustiveness errors could be improved #27256

Conversation

lahodaj commented Sep 12, 2025 • edited by openjdk bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewing

Uh oh!

bridgekeeper bot commented Sep 12, 2025

Uh oh!

openjdk bot commented Sep 12, 2025

Uh oh!

openjdk bot commented Sep 12, 2025

Uh oh!

openjdk bot commented Sep 12, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lahodaj commented Nov 7, 2025

Uh oh!

biboudis Nov 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

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!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

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!

mcimadamore Nov 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

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!

mcimadamore left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

openjdk-notifier bot commented Nov 18, 2025

Uh oh!

openjdk bot commented Nov 18, 2025

Uh oh!

davidalayachew commented Nov 20, 2025

Uh oh!

openjdk bot commented Nov 25, 2025

Uh oh!

Reviewers

Assignees

Labels

lahodaj commented Sep 12, 2025 •

edited by openjdk bot

Loading

openjdk bot commented Sep 12, 2025 •

edited

Loading

biboudis Nov 13, 2025 •

edited

Loading

mcimadamore Nov 14, 2025 •

edited

Loading