parquet與schema的那點事
最近做了一個從數據庫遷移到parquet報異常的案例,錯誤如下:
java.lang.IllegalArgumentException:No enum constant org.apache.parquet.schema.OriginalType.nihao total entry:11 java.lang.Enum.valueOf(Unknown Source)
org.apache.parquet.schema.OriginalType.valueOf(OriginalType.java:21)
org.apache.parquet.schema.MessageTypeParser.addPrimitiveType(MessageTypeParser.java:163)
org.apache.parquet.schema.MessageTypeParser.addType(MessageTypeParser.java:112)
org.apache.parquet.schema.MessageTypeParser.addGroupTypeFields(MessageTypeParser.java:100)
org.apache.parquet.schema.MessageTypeParser.parse(MessageTypeParser.java:93)
org.apache.parquet.schema.MessageTypeParser.parseMessageType(MessageTypeParser.java:83)
根據錯誤內容,很明顯就是報No enum constant org.apache.parquet.schema.OriginalType.nihao 的錯誤。
查看OriginalType的源碼定義,支持的type裏沒有nihao,這就是爲什麼報錯:
https://github.com/apache/parquet-mr/blob/master/parquet-column/src/main/java/org/apache/parquet/schema/OriginalType.java
那麼nihao是從哪裏來的呢?接下來就要查看數據表結構的shema,schema如下:
name
age
test(nihao)
…
原來nihao來源於test(nihao),那麼爲什麼呢?
繼續分析源碼:
https://github.com/apache/parquet-mr/blob/master/parquet-column/src/main/java/org/apache/parquet/schema/MessageTypeParser.java
/**
*
* @param input the text representation of the schema to parse
* @return the corresponding object representation
*/
public static MessageType parseMessageType(String input) {
return parse(input);
}
private static MessageType parse(String schemaString) {
Tokenizer st = new Tokenizer(schemaString, " ;{}()\n\t");
原來對schemaString做了parse,new了Tokenizer對象,Tokenizer哪裏來的啊?
public Tokenizer(String schemaString, String string) {
st = new StringTokenizer(schemaString, " ,;{}()\n\t=", true);
}
原來來源於StringTokenizer:https://docs.oracle.com/javase/7/docs/api/java/util/StringTokenizer.html#StringTokenizer(java.lang.String,%20java.lang.String,%20boolean)
The set of delimiters (the characters that separate tokens)
,;{}()\n\t=字符就是分隔符啊,分隔符()就是問題的關鍵,原來如此:
test(nihao)爲schema,parseMessageType要parse出type,就是用的分隔符 ,;{}()\n\t=來parse,這裏用到的分隔符是(), 由於nihao不是OriginalType中的member,所以報錯啊。
解決方法就是避免schema用特殊符號 ,;{}()\n\t=