英文:
spring data mongodb calling save twice leads to duplicate key exception
问题
@Test
void testUpdate() {
FooDto fooDto = getResource("/json/foo.json", new TypeReference<FooDto>() {
});
Foo foo = fooMapper.fromDTO(fooDto);
foo = fooService.save(foo);
log.info("Saved foo: " + foo);
foo.setState(FooState.Bar);
foo = fooService.save(foo);
log.info("Updated foo: " + foo);
}
Repository:
public interface FooRepository extends MongoRepository<Foo, String>
Entity:
@Document
public class Foo {
@Id
private String id;
private FooState state;
@DBRef
@Cascade
private Collection<Bar> bars = new ArrayList<>();
...
}
CascadeMongoEventListener:
public class CascadeMongoEventListener extends AbstractMongoEventListener<Object> {
private @Autowired
MongoOperations mongoOperations;
public @Override void onBeforeConvert(final BeforeConvertEvent<Object> event) {
final Object source = event.getSource();
ReflectionUtils.doWithFields(source.getClass(), new CascadeSaveCallback(source, mongoOperations));
}
private static class CascadeSaveCallback implements ReflectionUtils.FieldCallback {
// Implementation of the CascadeSaveCallback class...
}
private static class IdentifierCallback implements ReflectionUtils.FieldCallback {
// Implementation of the IdentifierCallback class...
}
}
Expected behavior:
From the docs in org.springframework.data.mongodb.core.MongoOperations#save(T):
Save the object to the collection for the entity type of the object to
save. This will perform an insert if the object is not already
present, that is an 'upsert'.
Update: The problem might be related to the index on the Bar child collection. (DbRef and Cascade lead to mongoOperations::save being called from the EventListener)
Update: I think I found the problem. Since I am using a custom serialization/deserialization in my Converter (Document.parse()) the id field is not mapped properly. This results in id being null and therefore this leads to an insert instead of an update.
MongoResultConversion:
public class MongoResultConversion {
@Component
@ReadingConverter
public static class ToResultConverter implements Converter<Document, Bar> {
// Implementation of the ToResultConverter class...
}
@Component
@WritingConverter
public static class ToDocumentConverter implements Converter<Bar, Document> {
// Implementation of the ToDocumentConverter class...
}
}
英文:
I try to save an entity with spring data mongodb repository. I have an EventListener that cascades saves.
The problem is, that I need to save an entity to get its internal id and perform further state mutations and saving the entity afterwards.
@Test
void testUpdate() {
FooDto fooDto = getResource("/json/foo.json", new TypeReference<FooDto>() {
});
Foo foo = fooMapper.fromDTO(fooDto);
foo = fooService.save(foo);
log.info("Saved foo: " + foo);
foo.setState(FooState.Bar);
foo = fooService.save(foo);
log.info("Updated foo: " + foo);
}
I have an index on a child collection of foo. It will not update children but will try to insert them twice which leads to org.springframework.dao.DuplicateKeyException.
Why does it not save but tries to insert it again?
Related:
Edit: versions:
mongodb 4,
spring boot 2.3.3.RELEASE
Edit more details:
Repository:
public interface FooRepository extends MongoRepository<Foo, String>
Entity:
@Document
public class Foo {
@Id
private String id;
private FooState state;
@DBRef
@Cascade
private Collection<Bar> bars = new ArrayList<>();
...
}
CascadeMongoEventListener:
//from https://mflash.dev/blog/2019/07/08/persisting-documents-with-mongorepository/#unit-tests-for-the-accountrepository
public class CascadeMongoEventListener extends AbstractMongoEventListener<Object> {
private @Autowired
MongoOperations mongoOperations;
public @Override void onBeforeConvert(final BeforeConvertEvent<Object> event) {
final Object source = event.getSource();
ReflectionUtils
.doWithFields(source.getClass(), new CascadeSaveCallback(source, mongoOperations));
}
private static class CascadeSaveCallback implements ReflectionUtils.FieldCallback {
private final Object source;
private final MongoOperations mongoOperations;
public CascadeSaveCallback(Object source, MongoOperations mongoOperations) {
this.source = source;
this.mongoOperations = mongoOperations;
}
public @Override void doWith(final Field field)
throws IllegalArgumentException, IllegalAccessException {
ReflectionUtils.makeAccessible(field);
if (field.isAnnotationPresent(DBRef.class) && field.isAnnotationPresent(Cascade.class)) {
final Object fieldValue = field.get(source);
if (Objects.nonNull(fieldValue)) {
final var callback = new IdentifierCallback();
final CascadeType cascadeType = field.getAnnotation(Cascade.class).value();
if (cascadeType.equals(CascadeType.PERSIST) || cascadeType.equals(CascadeType.ALL)) {
if (fieldValue instanceof Collection<?>) {
((Collection<?>) fieldValue).forEach(mongoOperations::save);
} else {
ReflectionUtils.doWithFields(fieldValue.getClass(), callback);
mongoOperations.save(fieldValue);
}
}
}
}
}
}
private static class IdentifierCallback implements ReflectionUtils.FieldCallback {
private boolean idFound;
public @Override void doWith(final Field field) throws IllegalArgumentException {
ReflectionUtils.makeAccessible(field);
if (field.isAnnotationPresent(Id.class)) {
idFound = true;
}
}
public boolean isIdFound() {
return idFound;
}
}
}
Edit: expected behaviour
From the docs in org.springframework.data.mongodb.core.MongoOperations#save(T):
> Save the object to the collection for the entity type of the object to
> save. This will perform an insert if the object is not already
> present, that is an 'upsert'.
Edit - new insights:
it might be related to the index on the Bar child collection. (DbRef and Cascade lead to mongoOperations::save being called from the EventListener)
I created another similar test with another entity and it worked.
The index on the child "Bar" entity (which is held as collection in parent "Foo" entity):
@CompoundIndex(unique = true, name = "fooId_name", def = "{'fooId': 1, 'name': 1}")
update: I think I found the problem. Since I am using a custom serialization/deserialization in my Converter (Document.parse()) the id field is not mapped properly. This results in id being null and therefore this leads to an insert instead of an update.
I will write an answer if I resolved this properly.
public class MongoResultConversion {
@Component
@ReadingConverter
public static class ToResultConverter implements Converter<Document, Bar> {
private final ObjectMapper mapper;
@Autowired
public ToResultConverter(ObjectMapper mapper) {
this.mapper = mapper;
}
public MeasureResult convert(Document source) {
String json = toJson(source);
try {
return mapper.readValue(json, new TypeReference<Bar>() {
});
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
}
protected String toJson(Document source) {
return source.toJson();
}
}
@Component
@WritingConverter
public static class ToDocumentConverter implements Converter<Bar, Document> {
private final ObjectMapper mapper;
@Autowired
public ToDocumentConverter(ObjectMapper mapper) {
this.mapper = mapper;
}
public Document convert(Bar source) {
String json = toJson(source);
return Document.parse(json);
}
protected String toJson(Bar source) {
try {
return mapper.writeValueAsString(source);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
}
}
}
答案1
得分: 5
根据我最后的编辑所述,问题出在自定义序列化/反序列化和Mongo文档转换上。这导致id为null,因此执行了插入操作,而不是upsert操作。
以下是我实现的自定义转换器来映射objectid的代码:
public class MongoBarConversion {
@Component
@ReadingConverter
public static class ToBarConverter implements Converter<Document, Bar> {
private final ObjectMapper mapper;
@Autowired
public ToBarConverter(ObjectMapper mapper) {
this.mapper = mapper;
}
public Bar convert(Document source) {
JsonNode json = toJson(source);
setObjectId(source, json);
return mapper.convertValue(json, new TypeReference<Bar>() {});
}
protected void setObjectId(Document source, JsonNode jsonNode) {
ObjectNode modifiableObject = (ObjectNode) jsonNode;
String objectId = getObjectId(source);
modifiableObject.put(ID_FIELD, objectId);
}
protected String getObjectId(Document source) {
String objectIdLiteral = null;
ObjectId objectId = source.getObjectId("_id");
if (objectId != null) {
objectIdLiteral = objectId.toString();
}
return objectIdLiteral;
}
protected JsonNode toJson(Document source) {
JsonNode node = null;
try {
String json = source.toJson();
node = mapper.readValue(json, JsonNode.class);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
return node;
}
}
@Component
@WritingConverter
public static class ToDocumentConverter implements Converter<Bar, Document> {
private final ObjectMapper mapper;
@Autowired
public ToDocumentConverter(ObjectMapper mapper) {
this.mapper = mapper;
}
public Document convert(Bar source) {
try {
JsonNode jsonNode = toJson(source);
setObjectId(source, jsonNode);
String json = mapper.writeValueAsString(jsonNode);
return Document.parse(json);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
}
protected void setObjectId(Bar source, JsonNode jsonNode) throws JsonProcessingException {
ObjectNode modifiableObject = (ObjectNode) jsonNode;
JsonNode objectIdJson = getObjectId(source);
modifiableObject.set("_id", objectIdJson);
modifiableObject.remove(ID_FIELD);
}
protected JsonNode getObjectId(Bar source) throws JsonProcessingException {
ObjectNode _id = null;
String id = source.getId();
if (id != null) {
_id = JsonNodeFactory.instance.objectNode();
_id.put("$oid", id);
}
return _id;
}
protected JsonNode toJson(Bar source) {
return mapper.convertValue(source, JsonNode.class);
}
}
}
所以总结一下:如果id非空,连续的两次保存肯定会导致upsert操作。bug出在我的代码中。
英文:
As stated in my last edit the problem was with the custom serialization/deserialization and mongo document conversion. This resulted in id being null and therefore an insert was done instead of an upsert.
The following code is my implementation of my custom converter to map the objectid:
public class MongoBarConversion {
@Component
@ReadingConverter
public static class ToBarConverter implements Converter<Document, Bar> {
private final ObjectMapper mapper;
@Autowired
public ToBarConverter(ObjectMapper mapper) {
this.mapper = mapper;
}
public Bar convert(Document source) {
JsonNode json = toJson(source);
setObjectId(source, json);
return mapper.convertValue(json, new TypeReference<Bar>() {
});
}
protected void setObjectId(Document source, JsonNode jsonNode) {
ObjectNode modifiableObject = (ObjectNode) jsonNode;
String objectId = getObjectId(source);
modifiableObject.put(ID_FIELD, objectId);
}
protected String getObjectId(Document source) {
String objectIdLiteral = null;
ObjectId objectId = source.getObjectId("_id");
if (objectId != null) {
objectIdLiteral = objectId.toString();
}
return objectIdLiteral;
}
protected JsonNode toJson(Document source) {
JsonNode node = null;
try {
String json = source.toJson();
node = mapper.readValue(json, JsonNode.class);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
return node;
}
}
@Component
@WritingConverter
public static class ToDocumentConverter implements Converter<Bar, Document> {
private final ObjectMapper mapper;
@Autowired
public ToDocumentConverter(ObjectMapper mapper) {
this.mapper = mapper;
}
public Document convert(Bar source) {
try {
JsonNode jsonNode = toJson(source);
setObjectId(source, jsonNode);
String json = mapper.writeValueAsString(jsonNode);
return Document.parse(json);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
}
protected void setObjectId(Bar source, JsonNode jsonNode) throws JsonProcessingException {
ObjectNode modifiableObject = (ObjectNode) jsonNode;
JsonNode objectIdJson = getObjectId(source);
modifiableObject.set("_id", objectIdJson);
modifiableObject.remove(ID_FIELD);
}
protected JsonNode getObjectId(Bar source) throws JsonProcessingException {
ObjectNode _id = null;
String id = source.getId();
if (id != null) {
_id = JsonNodeFactory.instance.objectNode();
_id.put("$oid", id);
}
return _id;
}
protected JsonNode toJson(Bar source) {
return mapper.convertValue(source, JsonNode.class);
}
}
}
So to conclude: two subsequent saves should (and will) definitely lead to an upsert if the id is non null. The bug was in my code.
答案2
得分: 3
所有的MongoDB驱动都包含在客户端生成ID的功能。如果您只是为了获取ID而进行保存,研究一下如何使用客户端生成的ID,并完全删除第一次保存。
英文:
All MongoDB drivers include functionality to generate ids on the client side. If you only save to get the id, research how to use client-side id generation and remove the first save entirely.
答案3
得分: 0
我相信您在尝试第二次保存而没有从数据库获取数据时会遇到此问题。您正在更改保存的对象,而不是保存到数据库中的对象。尝试使用类似于findById的方法检索现有的 foo,然后执行下一步操作并保存它。
英文:
I believe you are facing this issue as you try to save for the second time without fetching from db. You are changing the object returned by the save, not the object saved into the db. Try retrieving existing foo by using a method like findById and then perform next steps and saving it
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库,让每个人都能够通过互相帮助和分享经验来进步。
评论