对于parquet格式， 修改了字段类型又如何

注：图来自info官网

[slave01:21000] > create table ml_123 (a int,b varchar(10)) STORED AS PARQUET ;
Query: create table ml_123 (a int,b varchar(10)) STORED AS PARQUET

Fetched 0 row(s) in 0.07s 
[slave01:21000] > insert into ml_123(a,b) values(3,cast('孟梁' as varchar(10)));
Query: insert into ml_123(a,b) values(3,cast('孟梁' as varchar(10)))
Query submitted at: 2018-03-14 11:30:44 (Coordinator: http://slave01:25000)
Query progress can be monitored at: http://slave01:25000/query_plan?query_id=35482310a072c5c2:f2c507ea00000000
Modified 1 row(s) in 0.21s
[slave01:21000] > select * from ml_123;
Query: select * from ml_123
Query submitted at: 2018-03-14 11:30:47 (Coordinator: http://slave01:25000)
Query progress can be monitored at: http://slave01:25000/query_plan?query_id=ea450028b742c082:25c024e100000000
+---+------+
| a | b    |
+---+------+
| 3 | 孟梁 |
+---+------+
Fetched 1 row(s) in 0.13s
[slave01:21000] > 

下面开始修改int为bigint，看看是什么情况

[slave01:21000] > alter table ml_123 change a a bigint;
Query: alter table ml_123 change a a bigint

Fetched 0 row(s) in 0.18s
[slave01:21000] > select * from ml_123;
Query: select * from ml_123
Query submitted at: 2018-03-14 11:32:30 (Coordinator: http://slave01:25000)
Query progress can be monitored at: http://slave01:25000/query_plan?query_id=85412aa2816c7a71:dd93c58200000000
WARNINGS: 
File 'hdfs://myha/user/hive/warehouse/telecom_xdr.db/ml_123/35482310a072c5c2-f2c507ea00000000_309295272_data.0.parq' has an incompatible Parquet schema for column 'telecom_xdr.ml_123.a'. Column type: BIGINT, Parquet schema:
optional int32 a [i:0 d:1 r:0]



File 'hdfs://myha/user/hive/warehouse/telecom_xdr.db/ml_123/35482310a072c5c2-f2c507ea00000000_309295272_data.0.parq' has an incompatible Parquet schema for column 'telecom_xdr.ml_123.a'. Column type: BIGINT, Parquet schema:
optional int32 a [i:0 d:1 r:0]

[slave01:21000] > 

出现上面情况，就是parquert在存储是，做了加密，二级制存储压缩。如果改为bigint，那么占用的空间会大些，所以之前的压缩不可用， 如果生产环境出现误操作，怎么恢复呢？

看下面：

[slave01:21000] > alter table ml_123 change a a int;
Query: alter table ml_123 change a a int

Fetched 0 row(s) in 0.10s
[slave01:21000] > select * from ml_123;
Query: select * from ml_123
Query submitted at: 2018-03-14 11:34:52 (Coordinator: http://slave01:25000)
Query progress can be monitored at: http://slave01:25000/query_plan?query_id=68401ff9b0414c60:2eb2c5e400000000
+---+------+
| a | b    |
+---+------+
| 3 | 孟梁 |
+---+------+
Fetched 1 row(s) in 0.13s
[slave01:21000] > 

对，把字段类型恢复到原样即可。是不是有点像外部表的感觉。 壳和躯体分开的。

我们在看看改了varchar长度会导致乱码吗？

[slave01:21000] > desc ml_123;
Query: describe ml_123
+------+-------------+---------+
| name | type        | comment |
+------+-------------+---------+
| a    | int         |         |
| b    | varchar(10) |         |
+------+-------------+---------+
Fetched 2 row(s) in 0.01s
[slave01:21000] > alter table ml_123 change b b varchar(30);
Query: alter table ml_123 change b b varchar(30)

Fetched 0 row(s) in 0.16s
[slave01:21000] > select * from ml_123;
Query: select * from ml_123
Query submitted at: 2018-03-14 11:37:00 (Coordinator: http://slave01:25000)
Query progress can be monitored at: http://slave01:25000/query_plan?query_id=c74eda9401c455d7:1252cec600000000
+---+------+
| a | b    |
+---+------+
| 3 | 孟梁 |
+---+------+
Fetched 1 row(s) in 0.13s
[slave01:21000] > desc ml_123;
Query: describe ml_123
+------+-------------+---------+
| name | type        | comment |
+------+-------------+---------+
| a    | int         |         |
| b    | varchar(30) |         |
+------+-------------+---------+
Fetched 2 row(s) in 0.01s
[slave01:21000] > 

那么改短了，我们试试乱码吗？

[slave01:21000] > desc ml_123;
Query: describe ml_123
+------+-------------+---------+
| name | type        | comment |
+------+-------------+---------+
| a    | int         |         |
| b    | varchar(30) |         |
+------+-------------+---------+
Fetched 2 row(s) in 0.01s
[slave01:21000] > alter table ml_123 change b b varchar(2);
Query: alter table ml_123 change b b varchar(2)

Fetched 0 row(s) in 0.09s
[slave01:21000] > select * from ml_123;
Query: select * from ml_123
Query submitted at: 2018-03-14 11:38:17 (Coordinator: http://slave01:25000)
Query progress can be monitored at: http://slave01:25000/query_plan?query_id=af45bd6629b19d8d:c662468200000000
+---+---+
| a | b |
+---+---+
| 3 | � |
+---+---+
Fetched 1 row(s) in 0.13s
[slave01:21000] > 

那么我们试试，生成环境出现这种异常，我们怎么恢复？ 改回来能正常显示吗？

看：

Query progress can be monitored at: http://slave01:25000/query_plan?query_id=af45bd6629b19d8d:c662468200000000
+---+---+
| a | b |
+---+---+
| 3 | � |
+---+---+
Fetched 1 row(s) in 0.13s
[slave01:21000] > alter table ml_123 change b b varchar(12);
Query: alter table ml_123 change b b varchar(12)

Fetched 0 row(s) in 0.09s
[slave01:21000] > select * from ml_123;
Query: select * from ml_123
Query submitted at: 2018-03-14 11:39:14 (Coordinator: http://slave01:25000)
Query progress can be monitored at: http://slave01:25000/query_plan?query_id=d5449e9748082434:b0fc657200000000
+---+------+
| a | b    |
+---+------+
| 3 | 孟梁 |
+---+------+
Fetched 1 row(s) in 0.13s
[slave01:21000] > 

可以的。 哈哈，一样的。都是壳和躯体分开。

如果生成环境遇到问题，不要惊慌。恢复即可。

--end ---

老农民