基本概念
- Table(表),与其他数据库的表的概念一致,即为数据的集合。
- Items(数据项),代表一行数据。
- Attributes(属性),一个数据项可由多个属性构成,一个属性由属性名、属性值类型和属性值构成。
- Partition Key(分区键),即是最简单的主键,由一个属性构成,一张表有且只有一个分区键。由于 DynamoDB 内部在存储数据时使用分区键的 Hash 值实现跨多个分区的数据项目平均分布,故分区键又称之为 Hash Key。
- Sort Key(排序键),排序键和分区键构成另一种主键,用于在分区中按排序键排序。由于 DynamoDB 内部按照排序键值有序地将具有相同分区键的项目存储在互相紧邻的物理位置,故排序键又称之为 Range Key。
- Secondary Indexes(二级索引),这样命名应该是将主键视为一级索引。DynamoDB 的二级索引有两种:
- Global secondary index(GSI),创建跟主键不同的分区键和排序键的索引,DynamoDB 有最多创建 20 个 GSI 的限制;
- Local secondary index(LSI),创建跟主键的分区键相同但是排序键相异的索引,当且仅当创建表时可用,DynamoDB 有最多创建 5 个 LSI 的限制;
配置本地开发环境
下载 DynamoDB 本地版本
从 计算机上的 DynamoDB(可下载版本)- aws 下载,并放置到合适的位置。
配置
> aws configure --profile local AWS Access Key ID [None]: fake-ak
AWS Secret Access Key [None]: fake-sk
Default region name [None]:
Default output format [None]:
启动
java -Djava.library.path=./DynamoDBLocal_lib -jar DynamoDBLocal.jar -sharedDb
CLI 操作
注意:如果是本地版本需要添加
--endpoint-url http://localhost:8000
# 创建表格
aws dynamodb create-table \
--table-name ab-debug-proxy \
--attribute-definitions \
AttributeName=host,AttributeType=S \
AttributeName=port,AttributeType=N \
AttributeName=channel,AttributeType=S \
--key-schema AttributeName=host,KeyType=HASH AttributeName=port,KeyType=RANGE \
--global-secondary-indexes IndexName=channel-index,KeySchema=["{AttributeName=channel,KeyType=HASH}"],Projection="{ProjectionType=ALL}",ProvisionedThroughput="{ReadCapacityUnits=1,WriteCapacityUnits=1}" \
--provisioned-throughput ReadCapacityUnits=1,WriteCapacityUnits=1 \
--endpoint-url http://localhost:8000
# 删除表格
aws dynamodb delete-table \
--table-name ab-debug-proxy \
--endpoint-url http://localhost:8000
# 列出所有表格
aws dynamodb list-tables --endpoint-url http://localhost:8000
# 添加数据项
aws dynamodb put-item \
--table-name ab-debug-proxy \
--item '{"host": {"S": "192.168.1.0"},"port": {"N": "9090"},"channel": {"S": "sw"} }' \
--return-consumed-capacity TOTAL \
--endpoint-url http://localhost:8000
# 扫描数据项
aws dynamodb scan \
--table-name ab-debug-proxy \
--endpoint-url http://localhost:8000
Java 操作
使用 DynamoDBMapper
AmazonDynamoDB amazonDynamoDB = builder.build();
final String prefix = "project-" + AppConfig.getVariant() + "-";
final DynamoDBMapperConfig.TableNameOverride tableNameOverride =
DynamoDBMapperConfig.TableNameOverride.withTableNamePrefix(prefix);
DynamoDBMapperConfig dynamoDBMapperConfig = DynamoDBMapperConfig.builder()
.withTableNameOverride(tableNameOverride)
.build();
DynamoDBMapper DynamoDBMapper(amazonDynamoDB, dynamoDBMapperConfig);
批量加载(Batch Load)
List keyPairList = new ArrayList<>();
for (Proxy proxy : proxyList) {
KeyPair keyPair = new KeyPair();
keyPair.setHashKey(proxy.getHost());
keyPair.setRangeKey(proxy.getPort());
keyPairList.add(keyPair);
}
Map, List> keyPairForTable = new HashMap<>();
keyPairForTable.put(Proxy.class, keyPairList);
Map> stringListMap = dynamoDBMapper.batchLoad(keyPairForTable);
排错
DynamoDBMappingException: [class name] :no mapping for HASH key
确保对主键设置了以下的注解:
@DynamoDBHashKey
查询数据时使用 limit 字段但是没有返回预期的数据
原因是 limit 限制的是去查询集合而非结果集合,查看注释可知:
/**
* Sets the limit of items to scan and returns a pointer to this object for
* method-chaining. Please note that this is not the same as the
* number of items to return from the scan operation -- the operation will
* cease and return as soon as this many items are scanned, even if no
* matching results are found.
*
* @see DynamoDBScanExpression#getLimit()
*/
public DynamoDBScanExpression withLimit(Integer limit) {
this.limit = limit;
return this;
}
那么怎么做到对结果集合的 limit 呢?采用其带 page
的接口分页查询,如下:
// page query
do {
QueryResultPage proxyQueryResultPage =
dynamoDBMapper.queryPage(Proxy.class, queryExpression);
proxies.addAll(proxyQueryResultPage.getResults());
queryExpression.setExclusiveStartKey(proxyQueryResultPage.getLastEvaluatedKey());
} while (queryExpression.getExclusiveStartKey() != null && proxies.size() < limit);
设置 SaveBehavior.UPDATE_SKIP_NULL_ATTRIBUTES
但在调用 batchSave
方法时并没有跳过为 null 的字段
在创建 DynamoDBMapperConfig 时设置 UPDATE_SKIP_NULL_ATTRIBUTES,如下:
DynamoDBMapperConfig dynamoDBMapperConfig = DynamoDBMapperConfig.builder()
.withTableNameOverride(tableNameOverride)
// 设置 UPDATE_SKIP_NULL_ATTRIBUTES
.withSaveBehavior(DynamoDBMapperConfig.SaveBehavior.UPDATE_SKIP_NULL_ATTRIBUTES)
.build();
- 关于 SaveBehavior 参考 Using the SaveBehavior Configuration for the DynamoDBMapper
查看 SaveBehavior
源码,发现:
/**
* UPDATE_SKIP_NULL_ATTRIBUTES is similar to UPDATE, except that it
* ignores any null value attribute(s) and will NOT remove them from
* that item in DynamoDB. It also guarantees to send only one single
* updateItem request, no matter the object is key-only or not.
*/
UPDATE_SKIP_NULL_ATTRIBUTES,
也就是说 不支持 批量的操作!
Scan 过滤 Boolean 类型字段时返回为空
Map eav = new HashMap<>();
Map ean = new HashMap<>();
StringBuilder filterExpressionBuilder = new StringBuilder();
eav.put(":valid", new AttributeValue().withBOOL(proxy.getLocked()));
ean.put("#valid","valid");
filterExpressionBuilder.append("#valid = :valid");
DynamoDBScanExpression scanExpression = new DynamoDBScanExpression().withConsistentRead(false)
.withExpressionAttributeValues(eav)
.withExpressionAttributeNames(ean)
.withFilterExpression(filterExpressionBuilder.toString());
PaginatedScanList result = dynamoDBMapper.scan(Proxy.class, scanExpression);
DynamoDB 存储 Boolean 类型时其实是使用 Number 类型存储,进行过滤时不能使用布尔值进行过滤,而是要使用 “0” (false)和 “1”(true),更改如下:
eav.put(":valid", new AttributeValue().withN(proxy.getLocked() ? "1" : "0"));
DynamoDBMappingException
调用 DynamoDB 的 batchSave 接口发生以下错误:
com.amazonaws.services.dynamodbv2.datamodeling.DynamoDBMappingException: not supported; requires @DynamoDBTyped or @DynamoDBTypeConverted
at com.amazonaws.services.dynamodbv2.datamodeling.StandardModelFactories$Rules$NotSupported.set(StandardModelFactories.java:664)
at com.amazonaws.services.dynamodbv2.datamodeling.StandardModelFactories$Rules$NotSupported.set(StandardModelFactories.java:650)
at com.amazonaws.services.dynamodbv2.datamodeling.StandardModelFactories$AbstractRule.convert(StandardModelFactories.java:709)
at com.amazonaws.services.dynamodbv2.datamodeling.StandardModelFactories$AbstractRule.convert(StandardModelFactories.java:691)
at com.amazonaws.services.dynamodbv2.datamodeling.DynamoDBMapperFieldModel.convert(DynamoDBMapperFieldModel.java:138)
at com.amazonaws.services.dynamodbv2.datamodeling.DynamoDBMapper.batchWrite(DynamoDBMapper.java:1107)
at com.amazonaws.services.dynamodbv2.datamodeling.AbstractDynamoDBMapper.batchSave(AbstractDynamoDBMapper.java:173)
原来在注解为 @DynamoDBTable
的实体类中使用了自定义类型的字段,DynamoDB 并不支持。
解决方案:在该自定义类型的字段上添加注解 @DynamoDBTypeConvertedJson
即可,原理是使用 Jackson 将该字段的值 json 化再存储。如果不想要 Jackson
,可以自定义转化器,如下:
// CustomTypeConverter.java
public class CustomTypeConverter implements DynamoDBTypeConverter {
private Class clazz;
public CustomTypeConverter(Class clazz) {
this.clazz = clazz;
}
@Override
public String convert(T object) {
return Json.toJson(object);
}
@Override
public T unconvert(String json) {
return Json.fromJson(json, clazz);
}
}
// JobTreeConverter.java
public class JobTreeConverter extends CustomTypeConverter {
public JobTreeConverter(Class clazz) {
super(clazz);
}
}
// JobGroup.java
@DynamoDBTable(tableName = "job-group")
public class JobGroup implements Serializable {
@DynamoDBHashKey(attributeName = "id")
@SerializedName("id")
private String id;
@SerializedName("name")
private String name;
@DynamoDBTypeConverted(converter = JobTreeConverter.class)
@DynamoDBAttribute(attributeName = "default_job_tree")
@SerializedName("default_job_tree")
private JobTree defaultJobTree;
//...
}
参考
- 计算机上的 DynamoDB(可下载版本)- aws
- Instroduction - Amazon DynamoDB
- DynamoDB API 文档 -aws