Add type adapters for `java.time` classes by eamonnmcmanus · Pull Request #2948 · google/gson

eamonnmcmanus · 2025-11-25T00:30:06Z

Add type adapters for java.time classes.

These adapters essentially freeze the JSON representation thatReflectiveTypeAdapterFactory established by default, based on the private fields of java.time classes. That's not a great representation, but it is understandable. Changing it to anything else would break compatibility with systems that are expecting the current format.

If Gson had been updated with java.time adapters at the time the java.time API was added, the representation would surely have been something else, probably ISO standard formats. We can potentially supply non-default adapters for that, but we'll still need to have these legacy adapters by default.

I've been meaning to make this change for a while, but the need to do so becomes more urgent with this JDK
commit which makes a number of java.time fields transient. That means that, in JDK 26, the reflective-based adapter will no longer work with the classes that were changed.

This addresses #1059 and #739. All relevant classes in the java.time package are covered.

(Google internal bug reference: b/463237567.)

These adapters essentially freeze the JSON representation that `ReflectiveTypeAdapterFactory` established by default, based on the private fields of `java.time` classes. That's not a great representation, but it is understandable. Changing it to anything else would break compatibility with systems that are expecting the current format. If Gson had been updated with `java.time` adapters at the time the `java.time` API was added, the representation would surely have been something else, probably ISO standard formats. We can still supply non-default adapters for that, but we'll still need to have these legacy adapters by default. I've been meaning to make this change for a while, but the need to do so becomes more urgent with [this JDK commit](openjdk/jdk@cc05530) which makes a number of `java.time` fields `transient`. That means that the reflective-based adapter will no longer work with the classes that were changed.

I think it is time to do this, but obviously it should be a separate PR.

cpovirk · 2025-11-25T17:03:28Z

+        }
+
+        @Override
+        @SuppressWarnings("JavaDurationGetSecondsGetNano")


Huh, I would have hoped that the getSeconds and getNano calls below were close enough to avoid this warning :(

I'm not sure what you mean? The warning is supposed to prevent you from calling getSeconds() when you meant toSeconds().

https://errorprone.info/bugpattern/JavaDurationGetSecondsGetNano:

If you call duration.getNano(), you must also call duration.getSeconds() in ‘nearby’ code.

I get the impression that the check would be happy with foo(duration.getSeconds(), duration.getNano()) but not the array case. But I see no other examples in my depot search for ([{]|new.*\[)[^\042]*getSeconds.*getNano lang:java, so I don't think I'll bother to file a feature request.

(I forget where we landed with getSeconds vs. toSeconds. I just know that I got sad about the various get* and to* methods, including the to*Part methods, in b/201676318.)

* Move the new `TypeAdapter` implementations into a new class `JavaTimeTypeAdapters`. This allows us to omit that class from Android builds where it otherwise triggers warnings about SDK levels. * Ensure that every `java.time` class that needs a `TypeAdapter` has one. This doesn't include "subpackages" `java.time.*` like `java.time.chrono`, where it doesn't appear that there are any classes that are likely to be serialized to JSON. * Fix the handling of `ZoneId` and its subclasses.

* refactor: slightly optimize ConstructorConstructor * Update comments Co-authored-by: Marcono1234 <Marcono1234@users.noreply.github.com> * spotless apply Co-authored-by: Marcono1234 <Marcono1234@users.noreply.github.com> --------- Co-authored-by: Marcono1234 <Marcono1234@users.noreply.github.com>

I copied the internal version that was running on Google's infrastructure, and some adjustments were needed.

This still doesn't pass with a local build, and I may need to debug an OSGi problem.

cpovirk · 2025-12-19T15:56:14Z

+        }
+
+        @Override
+        @SuppressWarnings("JavaDurationGetSecondsGetNano")


https://errorprone.info/bugpattern/JavaDurationGetSecondsGetNano:

If you call duration.getNano(), you must also call duration.getSeconds() in ‘nearby’ code.

I get the impression that the check would be happy with foo(duration.getSeconds(), duration.getNano()) but not the array case. But I see no other examples in my depot search for ([{]|new.*\[)[^\042]*getSeconds.*getNano lang:java, so I don't think I'll bother to file a feature request.

(I forget where we landed with getSeconds vs. toSeconds. I just know that I got sad about the various get* and to* methods, including the to*Part methods, in b/201676318.)

It appears that sometimes the clauses being checked for appear in the other order, possibly because of a `HashMap` or the like somewhere in the guts of OSGi. I haven't seen this on GitHub, but I do see it when running locally with Google's JDK, which has more hash randomization.

It's enough just to swap the two clauses of a line when they are not in the expected order. Also, unrelatedly, update `protobuf-maven-plugin`.

…ip ci] Bumps [com.google.code.gson:gson](https://github.com/google/gson) from 2.13.2 to 2.14.0. Release notes *Sourced from [com.google.code.gson:gson's releases](https://github.com/google/gson/releases).* > Gson 2.14.0 > ----------- > > What's Changed > -------------- > > * Add type adapters for `java.time` classes by [`@eamonnmcmanus`](https://github.com/eamonnmcmanus) in [google/gson#2948](https://redirect.github.com/google/gson/pull/2948) > > When the `java.time` API is available, Gson automatically can read and write instances of classes like `Instant` and `Duration`. The format it uses essentially freezes the JSON representation that `ReflectiveTypeAdapterFactory` established by default, based on the private fields of `java.time` classes. That's not a great representation, but it is understandable. Changing it to anything else would break compatibility with systems that are expecting the current format. > > With this change, Gson no longer tries to access private fields of these classes using reflection. So it is no longer necessary to run with `--add-opens` for these classes on recent JDKs. > * Remove `com.google.gson.graph` by [`@eamonnmcmanus`](https://github.com/eamonnmcmanus) in [google/gson#2990](https://redirect.github.com/google/gson/pull/2990). > > This package was not part of any released artifact and depended on Gson internals in potentially problematic ways. > * Validate that strings being parsed as integers consist of ASCII characters by [`@eamonnmcmanus`](https://github.com/eamonnmcmanus) in [google/gson#2995](https://redirect.github.com/google/gson/pull/2995) > > Previously, strings could contain non-ASCII Unicode digits and still be parsed as integers. That's inconsistent with how JSON numbers are treated. > * Fix duplicate key detection when first value is null by [`@andrewstellman`](https://github.com/andrewstellman) in [google/gson#3006](https://redirect.github.com/google/gson/pull/3006) > > This could potentially break code that was relying on the incorrect behaviour. For example, this JSON string was previously accepted but will no longer be: `{"foo": null, "foo": bar}`. > * Remove `Serializable` from internal `Type` implementation classes. by [`@eamonnmcmanus`](https://github.com/eamonnmcmanus) in [google/gson#3011](https://redirect.github.com/google/gson/pull/3011) > > The nested classes `ParameterizedTypeImpl`, `GenericArrayTypeImpl`, and `WildcardTypeImpl` in `GsonTypes` are implementations of the corresponding types (without `Impl`) in `java.lang.reflect`. For some reason, they were serializable, even though the `java.lang.reflect` implementations are not. Having unnecessarily serializable classes could *conceivably* have been a security problem if they were part of a larger exploit using serialization. (We do not consider this a likely scenario and do not suggest that you need to update Gson just to get this change.) > * Add `LegacyProtoTypeAdapterFactory`. by [`@eamonnmcmanus`](https://github.com/eamonnmcmanus) in [google/gson#3014](https://redirect.github.com/google/gson/pull/3014) > > This is not part of any released artifact, but may be of use when trying to fix code that is currently accessing the internals of protobuf classes via reflection. > * Make AppendableWriter do flush and close if delegation object supports by [`@MukjepScarlet`](https://github.com/MukjepScarlet) in [google/gson#2925](https://redirect.github.com/google/gson/pull/2925) > > Other less visible changes > -------------------------- > > * Add default capacity to EnumTypeAdapter maps by [`@MukjepScarlet`](https://github.com/MukjepScarlet) in [google/gson#2959](https://redirect.github.com/google/gson/pull/2959) > * refactor: move derived adapters from Gson to TypeAdapters by [`@MukjepScarlet`](https://github.com/MukjepScarlet) in [google/gson#2951](https://redirect.github.com/google/gson/pull/2951) > * Optimize `new Gson()` by [`@MukjepScarlet`](https://github.com/MukjepScarlet) in [google/gson#2864](https://redirect.github.com/google/gson/pull/2864) > > New Contributors > ---------------- > > * [`@ThirdGoddess`](https://github.com/ThirdGoddess) made their first contribution in [google/gson#2944](https://redirect.github.com/google/gson/pull/2944) > * [`@lmj798`](https://github.com/lmj798) made their first contribution in [google/gson#2988](https://redirect.github.com/google/gson/pull/2988) > * [`@Eng-YasminKotb`](https://github.com/Eng-YasminKotb) made their first contribution in [google/gson#3005](https://redirect.github.com/google/gson/pull/3005) > * [`@andrewstellman`](https://github.com/andrewstellman) made their first contribution in [google/gson#3006](https://redirect.github.com/google/gson/pull/3006) > > **Full Changelog**: <google/gson@gson-parent-2.13.2...gson-parent-2.14.0> Commits * [`3ff35d6`](google/gson@3ff35d6) [maven-release-plugin] prepare release gson-parent-2.14.0 * [`a3024fd`](google/gson@a3024fd) Bump the maven group with 13 updates ([#3002](https://redirect.github.com/google/gson/issues/3002)) * [`5689ffe`](google/gson@5689ffe) Bump the github-actions group across 1 directory with 3 updates ([#3018](https://redirect.github.com/google/gson/issues/3018)) * [`48db33c`](google/gson@48db33c) Add `LegacyProtoTypeAdapterFactory`. ([#3014](https://redirect.github.com/google/gson/issues/3014)) * [`53d703e`](google/gson@53d703e) Update outdated comment regarding serializable types ([#3012](https://redirect.github.com/google/gson/issues/3012)) * [`0189b72`](google/gson@0189b72) Remove `Serializable` from internal `Type` implementation classes. ([#3011](https://redirect.github.com/google/gson/issues/3011)) * [`f4d371d`](google/gson@f4d371d) Fix duplicate key detection when first value is null ([#3006](https://redirect.github.com/google/gson/issues/3006)) * [`27d9ba1`](google/gson@27d9ba1) Fix typo in README (JPMS dependencies section) ([#3005](https://redirect.github.com/google/gson/issues/3005)) * [`1fa9b7a`](google/gson@1fa9b7a) Validate that strings being parsed as integers consist of ASCII characters (#... * [`b7d5954`](google/gson@b7d5954) Add iterator fail-fast tests for LinkedTreeMap.clear() ([#2992](https://redirect.github.com/google/gson/issues/2992)) * Additional commits viewable in [compare view](google/gson@gson-parent-2.13.2...gson-parent-2.14.0) [![Dependabot compatibility score](https://dependabot-badges.githubapp.com/badges/compatibility\_score?dependency-name=com.google.code.gson:gson&package-manager=maven&previous-version=2.13.2&new-version=2.14.0)](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores) Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`. [//]: # (dependabot-automerge-start) [//]: # (dependabot-automerge-end) --- Dependabot commands and options You can trigger Dependabot actions by commenting on this PR: - `@dependabot rebase` will rebase this PR - `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it - `@dependabot show ignore conditions` will show all of the ignore conditions of the specified dependency - `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself) - `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself) - `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

## Release notes Sourced from com.google.code.gson:gson's releases. Gson 2.14.0 What's Changed Add type adapters for java.time classes by @eamonnmcmanus in google/gson#2948 When the java.time API is available, Gson automatically can read and write instances of classes like Instant and Duration. The format it uses essentially freezes the JSON representation that ReflectiveTypeAdapterFactory established by default, based on the private fields of java.time classes. That's not a great representation, but it is understandable. Changing it to anything else would break compatibility with systems that are expecting the current format. With this change, Gson no longer tries to access private fields of these classes using reflection. So it is no longer necessary to run with --add-opens for these classes on recent JDKs. Remove com.google.gson.graph by @eamonnmcmanus in google/gson#2990. This package was not part of any released artifact and depended on Gson internals in potentially problematic ways. Validate that strings being parsed as integers consist of ASCII characters by @eamonnmcmanus in google/gson#2995 Previously, strings could contain non-ASCII Unicode digits and still be parsed as integers. That's inconsistent with how JSON numbers are treated. Fix duplicate key detection when first value is null by @andrewstellman in google/gson#3006 This could potentially break code that was relying on the incorrect behaviour. For example, this JSON string was previously accepted but will no longer be: {"foo": null, "foo": bar}. Remove Serializable from internal Type implementation classes. by @eamonnmcmanus in google/gson#3011 The nested classes ParameterizedTypeImpl, GenericArrayTypeImpl, and WildcardTypeImpl in GsonTypes are implementations of the corresponding types (without Impl) in java.lang.reflect. For some reason, they were serializable, even though the java.lang.reflect implementations are not. Having unnecessarily serializable classes could conceivably have been a security problem if they were part of a larger exploit using serialization. (We do not consider this a likely scenario and do not suggest that you need to update Gson just to get this change.) Add LegacyProtoTypeAdapterFactory. by @eamonnmcmanus in google/gson#3014 This is not part of any released artifact, but may be of use when trying to fix code that is currently accessing the internals of protobuf classes via reflection. Make AppendableWriter do flush and close if delegation object supports by @MukjepScarlet in google/gson#2925 Other less visible changes Add default capacity to EnumTypeAdapter maps by @MukjepScarlet in google/gson#2959 refactor: move derived adapters from Gson to TypeAdapters by @MukjepScarlet in google/gson#2951 Optimize new Gson() by @MukjepScarlet in google/gson#2864 New Contributors @ThirdGoddess made their first contribution in google/gson#2944 @lmj798 made their first contribution in google/gson#2988 @Eng-YasminKotb made their first contribution in google/gson#3005 @andrewstellman made their first contribution in google/gson#3006 Full Changelog: google/gson@gson-parent-2.13.2...gson-parent-2.14.0 Gson 2.13.2 The main changes in this release are just newer dependencies. ... (truncated) ## Changelog Sourced from com.google.code.gson:gson's changelog. Change Log The change log for versions newer than 2.10 is available only on the GitHub Releases page. Version 2.10 Support for serializing and deserializing Java records, on Java ≥ 16. (google/gson#2201) Add JsonArray.asList and JsonObject.asMap view methods (google/gson#2225) Fix TypeAdapterRuntimeTypeWrapper not detecting reflective TreeTypeAdapter and FutureTypeAdapter (google/gson#1787) Improve JsonReader.skipValue() (google/gson#2062) Perform numeric conversion for primitive numeric type adapters (google/gson#2158) Add Gson.fromJson(..., TypeToken) overloads (google/gson#1700) Fix changes to GsonBuilder affecting existing Gson instances (google/gson#1815) Make JsonElement conversion methods more consistent and fix javadoc (google/gson#2178) Throw UnsupportedOperationException when JsonWriter.jsonValue is not supported (google/gson#1651) Disallow JsonObject Entry.setValue(null) (google/gson#2167) Fix TypeAdapter.toJson throwing AssertionError for custom IOException (google/gson#2172) Convert null to JsonNull for JsonArray.set (google/gson#2170) Fixed nullSafe usage. (google/gson#1555) Validate TypeToken.getParameterized arguments (google/gson#2166) Fix #1702: Gson.toJson creates CharSequence which does not implement toString (google/gson#1703) Prefer existing adapter for concurrent Gson.getAdapter calls (google/gson#2153) Improve ArrayTypeAdapter for Object[] (google/gson#1716) Improve AppendableWriter performance (google/gson#1706) ## Commits 3ff35d6 [maven-release-plugin] prepare release gson-parent-2.14.0 a3024fd Bump the maven group with 13 updates (#3002) 5689ffe Bump the github-actions group across 1 directory with 3 updates (#3018) 48db33c Add LegacyProtoTypeAdapterFactory. (#3014) 53d703e Update outdated comment regarding serializable types (#3012) 0189b72 Remove Serializable from internal Type implementation classes. (#3011) f4d371d Fix duplicate key detection when first value is null (#3006) 27d9ba1 Fix typo in README (JPMS dependencies section) (#3005) 1fa9b7a Validate that strings being parsed as integers consist of ASCII characters (# b7d5954 Add iterator fail-fast tests for LinkedTreeMap.clear() (#2992) Additional commits viewable in compare view ![Dependabot compatibility score](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores) Issue-ID: CIMAN-33 Signed-off-by: dependabot[bot] <support@github.com> Change-Id: I3707533b94e77d221389af97baae133d90bdb985 GitHub-PR: #37 GitHub-Hash: dc2f307754a500e1 Signed-off-by: onap.gh2gerrit <releng+onap-gh2gerrit@linuxfoundation.org>

eamonnmcmanus added 6 commits November 24, 2025 16:18

Apply a couple of Error Prone suggestions.

28393e1

Reformat a couple of source files.

9b12b2e

Fix another EP warning.

3a38f77

Fix a particularly annoying EP warning/error.

9b48171

Update the Android API level.

f20ca75

I think it is time to do this, but obviously it should be a separate PR.

eamonnmcmanus requested a review from cpovirk November 25, 2025 01:54

cpovirk reviewed Nov 25, 2025

View reviewed changes

cpovirk mentioned this pull request Nov 25, 2025

Change the minimum supported Android API to 26. #2949

Closed

Marcono1234 reviewed Nov 27, 2025

View reviewed changes

eamonnmcmanus and others added 9 commits December 18, 2025 16:09

Fix outdated comment (google#2954)

4ddb1e5

Compilation fixes.

6e13db7

I copied the internal version that was running on Google's infrastructure, and some adjustments were needed.

Merge branch 'main' into javatime

16955d4

Fix lint errors.

5b36906

Fix a particularly annoying lint error.

4d0e6fe

Fix the fix.

6ab3eda

Further fixes.

6146d57

This still doesn't pass with a local build, and I may need to debug an OSGi problem.

cpovirk approved these changes Dec 19, 2025

View reviewed changes

nithril-png mentioned this pull request Dec 23, 2025

v3.3.9 still has serialization problems argoproj/argo-workflows#9692

Open

eamonnmcmanus added 9 commits December 23, 2025 12:21

Revert to Android API 23, but with gummy-bears-api.

be7e42a

Annotate JavaTimeTypeAdapters to suppress animal-sniffer errors.

fa66bb9

Fix use of assignment expression.

957cfbe

Remove the parser, which was overkill.

e1cee31

It's enough just to swap the two clauses of a line when they are not in the expected order. Also, unrelatedly, update `protobuf-maven-plugin`.

Merge branch 'osgimakingworkforpeopleagain' into javatime

41b7490

Use a local @IgnoreJRERequirement.

f36a5c5

Configure the local animal-sniffer annotation.

e752f74

Suppress IdentifierName warning.

0ba4b48

eamonnmcmanus added 3 commits December 23, 2025 14:10

Update comments and undo an unnecessary split.

3d58365

Merge branch 'osgimakingworkforpeopleagain' into javatime

8ac01db

Restore Math.toIntExact.

dfa4883

eamonnmcmanus merged commit c47db7b into google:main Dec 23, 2025
14 of 15 checks passed

eamonnmcmanus deleted the javatime branch December 23, 2025 22:38

Marcono1234 mentioned this pull request Jan 13, 2026

Follow-up for java.time adapter changes #2972

Open

eamonnmcmanus mentioned this pull request Feb 12, 2026

com.google.gson.JsonIOException: Failed making field 'java.time.LocalDateTime#date' accessible; either increase its visibility or write a custom TypeAdapter for its declaring type. #2977

Closed

Conversation

eamonnmcmanus commented Nov 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cpovirk Nov 25, 2025

Choose a reason for hiding this comment

Uh oh!

eamonnmcmanus Dec 18, 2025

Choose a reason for hiding this comment

Uh oh!

cpovirk Dec 19, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cpovirk Dec 19, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

eamonnmcmanus commented Nov 25, 2025 •

edited

Loading