我想在命令行上使用与编写器架构不同的读取器架构反序列化Avro数据。我可以在序列化时指定编写器模式,但不能在反序列化期间指定。
record.json
(数据文件):
{"test1": 1, "test2": 2}
writer.avsc
(作者架构):
{
"type": "record",
"name": "pouac",
"fields": [
{
"name": "test1",
"type": "int"
},
{
"name": "test2",
"type": "int"
}
]
}
reader.avsc
(读者架构):
{
"type": "record",
"name": "pouac",
"fields": [{
"name": "test2",
"type": "int",
"aliases": ["test1"]
}]
}
序列化数据:
$ java -jar avro-tools-1.8.2.jar fromjson --schema-file writer.avsc record.json > record.avro
为了反序列化数据,我尝试了以下方法:
$ java -jar avro-tools-1.8.2.jar tojson --schema-file reader.avsc record.avro
Exception in thread "main" joptsimple.UnrecognizedOptionException: 'schema-file' is not a recognized option
...
我主要是寻找命令行指令,因为我编写Java代码并不那么容易,但我很乐意使用Java代码来编译自己。实际上,我感兴趣的是确切的反序列化结果。 (this conversation在我为实现别名而打开的一个快速公关PR中描述了更为根本的问题
答案 0 :(得分:2)
avro-tools tojson
目标仅用作将二进制编码的Avro文件转换为JSON的转储工具。架构始终伴随Avro文件中的记录,如下面的链接所示。因此,avro-tools无法覆盖它。
http://avro.apache.org/docs/1.8.2/#compare
我不知道可以用来实现你想要的独立工具。我认为你需要做一些编程才能达到预期的效果。 Avro有许多支持的语言,包括Python,但跨语言的功能并不统一。 Java在我的经验中是最先进的。例如,Python缺乏在DataFileReader上指定读者架构的能力,这有助于实现您的目标:
https://github.com/apache/avro/blob/master/lang/py/src/avro/datafile.py#L224
您可以在Python中获得的最接近的内容如下;
import avro.schema as avsc
import avro.datafile as avdf
import avro.io as avio
reader_schema = avsc.parse(open("reader.avsc", "rb").read())
# need ability to inject reader schema as 3rd arg
with avdf.DataFileReader(open("record.avro", "rb"), avio.DatumReader()) as reader:
for record in reader:
print record
就您所概述的架构和数据而言。预期的行为应未定义,因此会发出错误。
可以使用以下Java代码验证此行为;
package ca.junctionbox.soavro;
import org.apache.avro.Schema;
import org.apache.avro.SchemaValidationException;
import org.apache.avro.SchemaValidationStrategy;
import org.apache.avro.SchemaValidator;
import org.apache.avro.SchemaValidatorBuilder;
import java.util.ArrayList;
public class Main {
public static final String V1 = "{\n" +
" \"type\": \"record\",\n" +
" \"name\": \"pouac\",\n" +
" \"fields\": [\n" +
" {\n" +
" \"name\": \"test1\",\n" +
" \"type\": \"int\"\n" +
" },\n" +
" {\n" +
" \"name\": \"test2\",\n" +
" \"type\": \"int\"\n" +
" }\n" +
" ]\n" +
"}";
public static final String V2 = "{\n" +
" \"type\": \"record\",\n" +
" \"name\": \"pouac\",\n" +
" \"fields\": [{\n" +
" \"name\": \"test2\",\n" +
" \"type\": \"int\",\n" +
" \"aliases\": [\"test1\"]\n" +
" }]\n" +
"}";
public static void main(final String[] args) {
final SchemaValidator sv = new SchemaValidatorBuilder()
.canBeReadStrategy()
.validateAll();
final Schema sv1 = new Schema.Parser().parse(V1);
final Schema sv2 = new Schema.Parser().parse(V2);
final ArrayList<Schema> existing = new ArrayList<>();
existing.add(sv1);
try {
sv.validate(sv2, existing);
System.out.println("Good to go!");
} catch (SchemaValidationException e) {
e.printStackTrace();
}
}
}
这会产生以下输出:
org.apache.avro.SchemaValidationException: Unable to read schema:
{
"type" : "record",
"name" : "pouac",
"fields" : [ {
"name" : "test2",
"type" : "int",
"aliases" : [ "test1" ]
} ]
}
using schema:
{
"type" : "record",
"name" : "pouac",
"fields" : [ {
"name" : "test1",
"type" : "int"
}, {
"name" : "test2",
"type" : "int"
} ]
}
at org.apache.avro.ValidateMutualRead.canRead(ValidateMutualRead.java:70)
at org.apache.avro.ValidateCanBeRead.validate(ValidateCanBeRead.java:39)
at org.apache.avro.ValidateAll.validate(ValidateAll.java:51)
at ca.junctionbox.soavro.Main.main(Main.java:47)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at org.codehaus.mojo.exec.ExecJavaMojo$1.run(ExecJavaMojo.java:294)
at java.lang.Thread.run(Thread.java:748)
别名通常用于模式演进中的向后兼容性,允许从不同/遗留密钥到公共密钥名称的映射。鉴于您的作者架构并未将test1和test2字段视为&#34;可选&#34;通过使用工会,我无法看到你想要这种转变的场景。如果你想&#34;掉落&#34; test1字段然后可以通过从v2模式规范中排除它来实现。任何可以应用读者方案的读者都会使用v2模式定义忽略test1。
说明我的进化意义;
v1架构
{
"type": "record",
"name": "pouac",
"fields": [
{
"name": "test1",
"type": "int"
}]
}
v2架构
{
"type": "record",
"name": "pouac",
"fields": [
{
"name": "test2",
"type": "int",
"aliases": ["test1"]
}]
}
您可以使用v1格式的TB级数据,并引入将test1字段重命名为test2的v2格式。别名允许您对v1和v2数据执行map-reduce作业,Hive查询等,而无需先主动重写所有旧的v1数据。请注意,这假设字段的类型和语义没有变化。
答案 1 :(得分:-2)
您可以运行java -jar avro-tools-1.8.2.jar tojson
查看帮助,它告诉您可以使用以下命令:
java -jar avro-tools-1.8.2.jar tojson record.avro > tost.json
,这将输出到文件:
{"test1":1,"test2":2}
您也可以使用--pretty
argumment:
java -jar avro-tools-1.8.2.jar tojson --pretty record.avro > tost.json
并且输出很漂亮:
{
"test1" : 1,
"test2" : 2
}