case3 :alter schema with index

[whitewum]

These are migration from internal jira reports.

case 1: the same property name, drop it , then re-create it.

case2: alter schema change property to a different type (string->int)

create tag test(col1 string , col2 string);

insert vertex test(col1,col2) values "11":("aa","bb");

alter tag test change (col2 int)

(root@nebula) [test2]> fetch prop on test "11"
+----------+-----------+-----------+
| VertexID | test.col1 | test.col2 |
+----------+-----------+-----------+
| "11" | "aa" | {color:#de350b}"bb"{color} |
+----------+-----------+-----------+

 

insert vertex test(col1,col2) values "22":("aa",99)
create tag index test_index on test(col2);

rebuild tag index test_index
+------------+
| New Job Id |
+------------+
| 49 |
+------------+
 show jobs;
+--------+---------------------+------------+------------+------------+
| Job Id | Command | Status | Start Time | Stop Time |
+--------+---------------------+------------+------------+------------+
| 49 | "REBUILD_TAG_INDEX" | "FINISHED" | 1607570799 | 1607570799 |

 

(root@nebula) [test2]> lookup on test where test.col2 < 99 yield test.col2;
+------+----------------------+
| _vid | col2 |
+------+----------------------+
| "11" | {color:#de350b}-2134143272594898944{color} |
+------+----------------------+
Got 1 rows (time spent 1957/3046 us)

case 3: alter schema with index

create tag test4(col1 int,col2 int);

insert vertex test4(col1,col2) values "a":(1,2)

insert vertex test4(col1,col2) values "b":(2,3)

alter tag test4 add (col3 int)

create tag index test4_index on test4(col1,col2,col3);

rebuild tag index test4_index

(root@nebula) [test3]> lookup on test4 where test4.col1 == 2;
Empty set (time spent 2026/2789 us)

case4:default value

# v2.0
(root@nebula) [test1]> create tag tag1(col1 timestamp default 0)
 Execution succeeded (time spent 18185/18647 us)
Tue, 26 Jan 2021 10:35:13 CST
(root@nebula) [test1]> insert vertex tag1() values "1":()
 Execution succeeded (time spent 2850/3427 us)
Tue, 26 Jan 2021 10:35:30 CST
(root@nebula) [test1]> insert vertex tag1() values "2":(111)
 [ERROR (-12)]: SemanticError: Column count doesn't match value count.

v1.0

(root@nebula) [test1]> create tag tag1(col1 timestamp default 0)
Execution succeeded (Time spent: 1.495/2.28 ms)Tue Jan 26 
Tue Jan 26 10:41:22 2021(root@nebula) [test1]> insert vertex tag1() values 1:()
Execution succeeded (Time spent: 1.487/2.372 ms)Tue Jan 26 10:41:32 2021(root@nebula) [test1]> insert vertex tag1() values 2:(10)
Execution succeeded (Time spent: 1.415/2.195 ms)Tue Jan 26 
Tue Jan 26 10:41:45 2021(root@nebula) [test1]> fetch prop on tag1 1
==================================
| VertexID | tag1.col1           |
==================================
| 1        | 1970-01-01 08:00:00 |
----------------------------------
Got 1 rows (Time spent: 1.511/2.596 ms)Tue Jan 26 10:41:49 2021(root@nebula) [test1]> fetch prop on tag1 2
==================================
| VertexID | tag1.col1           |
==================================
| 2        | 1970-01-01 08:00:00 |
----------------------------------
Got 1 rows (Time spent: 1.012/1.632 ms)Tue Jan 26 10:41:51 2021

[CPWstatic]

Prepare the space

(root@nebula) [(none)]> create space test1(vid_type=fixed_string(30))
Execution succeeded (time spent 39512/41812 us)

Tue, 14 Sep 2021 10:27:53 CST

(root@nebula) [(none)]> use test1
[ERROR (-1005)]: SpaceNotFound:

Tue, 14 Sep 2021 10:27:56 CST

(root@nebula) [(none)]> use test1
Execution succeeded (time spent 48813/51626 us)

Tue, 14 Sep 2021 10:27:58 CST

Case 1 : the same property name, drop it , then re-create it.

(root@nebula) [test1]> create tag test3(col1 string, col2 string);
Execution succeeded (time spent 46958/49297 us)

Tue, 14 Sep 2021 10:28:24 CST

(root@nebula) [test1]> insert vertex test3(col1, col2) value "11":("col1", "col2");
Execution succeeded (time spent 8589/11521 us)

Tue, 14 Sep 2021 10:29:04 CST

(root@nebula) [test1]> alter tag test3 drop (col2);
Execution succeeded (time spent 31366/36053 us)

Tue, 14 Sep 2021 10:29:53 CST

(root@nebula) [test1]> fetch prop on test3 "11";
+-----------------------------+
| vertices_                   |
+-----------------------------+
| ("11" :test3{col1: "col1"}) |
+-----------------------------+
Got 1 rows (time spent 6965/10054 us)

Tue, 14 Sep 2021 10:30:08 CST

(root@nebula) [test1]> alter tag test3 add (col2 string default "xxxx");
Execution succeeded (time spent 25945/28989 us)

Tue, 14 Sep 2021 10:30:52 CST

(root@nebula) [test1]> fetch prop on test3 "11";
+-----------------------------+
| vertices_                   |
+-----------------------------+
| ("11" :test3{col1: "col1"}) |
+-----------------------------+
Got 1 rows (time spent 5199/7972 us)

Tue, 14 Sep 2021 10:30:57 CST

(root@nebula) [test1]> fetch prop on test3 "11";
+-------------------------------------------+
| vertices_                                 |
+-------------------------------------------+
| ("11" :test3{col1: "col1", col2: "col2"}) |
+-------------------------------------------+
Got 1 rows (time spent 6125/9092 us)

Tue, 14 Sep 2021 10:31:00 CST

This problem is due to that the columns have no version, so if the column is deleted first and then added back to the tags/edges, we can still read the value after all. This is a known issue, we could add it to v3.0 roadmap. @Sophie-Xie

Case 2: alter schema change property to a different type (string->int)

(root@nebula) [test1]> create tag test(col1 string , col2 string);
Execution succeeded (time spent 83144/85189 us)

Tue, 14 Sep 2021 10:38:20 CST

(root@nebula) [test1]> insert vertex test(col1,col2) values "11":("aa","bb");
Execution succeeded (time spent 6843/9300 us)

Tue, 14 Sep 2021 10:38:42 CST

(root@nebula) [test1]> fetch prop on test "11"
+--------------------------------------+
| vertices_                            |
+--------------------------------------+
| ("11" :test{col1: "aa", col2: "bb"}) |
+--------------------------------------+
Got 1 rows (time spent 11830/14468 us)

Tue, 14 Sep 2021 10:38:58 CST

(root@nebula) [test1]> alter tag test change (col2 int)
Execution succeeded (time spent 40178/44064 us)

Tue, 14 Sep 2021 10:41:12 CST

(root@nebula) [test1]> fetch prop on test "11"
+--------------------------------------+
| vertices_                            |
+--------------------------------------+
| ("11" :test{col1: "aa", col2: "bb"}) |
+--------------------------------------+
Got 1 rows (time spent 9940/13559 us)

Tue, 14 Sep 2021 10:41:18 CST

(root@nebula) [test1]> insert vertex test(col1,col2) values "22":("aa",99)
Execution succeeded (time spent 15569/18790 us)

Tue, 14 Sep 2021 10:42:04 CST

(root@nebula) [test1]> create tag index test_index on test(col2);
Execution succeeded (time spent 68325/71791 us)

Tue, 14 Sep 2021 10:42:08 CST

(root@nebula) [test1]> rebuild tag index test_index
+------------+
| New Job Id |
+------------+
| 30         |
+------------+
Got 1 rows (time spent 61606/66416 us)

Tue, 14 Sep 2021 10:42:13 CST

(root@nebula) [test1]> show jobs
+--------+---------------------+------------+-------------------------+-------------------------+
| Job Id | Command             | Status     | Start Time              | Stop Time               |
+--------+---------------------+------------+-------------------------+-------------------------+
| 30     | "REBUILD_TAG_INDEX" | "FINISHED" | 2021-09-14T02:42:13.000 | 2021-09-14T02:42:13.000 |
+--------+---------------------+------------+-------------------------+-------------------------+
Got 1 rows (time spent 17940/20785 us)

Tue, 14 Sep 2021 10:42:17 CST

(root@nebula) [test1]> lookup on test where test.col2 < 99 yield test.col2;
+----------+----------------------+
| VertexID | test.col2            |
+----------+----------------------+
| "11"     | -2134143272594898944 |
+----------+----------------------+
Got 1 rows (time spent 13113/16100 us)

Tue, 14 Sep 2021 10:42:34 CST

The problem is that the tag/edge have versions, so we can read the older data even the datatype were changed. We should discuss this one. @Sophie-Xie

case 3:

(root@nebula) [test1]> create tag test4(col1 int,col2 int);
Execution succeeded (time spent 83118/86371 us)

Tue, 14 Sep 2021 10:47:19 CST

(root@nebula) [test1]> insert vertex test4(col1,col2) values "a":(1,2)
Execution succeeded (time spent 7692/11624 us)

Tue, 14 Sep 2021 10:47:23 CST

(root@nebula) [test1]> insert vertex test4(col1,col2) values "b":(2,3)
Execution succeeded (time spent 35314/47294 us)

Tue, 14 Sep 2021 10:47:27 CST

(root@nebula) [test1]> alter tag test4 add (col3 int)
Execution succeeded (time spent 152310/163985 us)

Tue, 14 Sep 2021 10:47:30 CST

(root@nebula) [test1]> create tag index test4_index on test4(col1,col2,col3);
Execution succeeded (time spent 118015/121077 us)

Tue, 14 Sep 2021 10:47:34 CST

(root@nebula) [test1]> rebuild tag index test4_index
[ERROR (-1009)]: SemanticError: Index test4_index not found in space test1

Tue, 14 Sep 2021 10:47:38 CST

(root@nebula) [test1]> rebuild tag index test4_index
[ERROR (-1009)]: SemanticError: Index test4_index not found in space test1

Tue, 14 Sep 2021 10:47:41 CST

(root@nebula) [test1]> rebuild tag index test4_index
+------------+
| New Job Id |
+------------+
| 33         |
+------------+
Got 1 rows (time spent 27239/30132 us)

Tue, 14 Sep 2021 10:47:48 CST

(root@nebula) [test1]> lookup on test4 where test4.col1 == 2;
Empty set (time spent 10488/13468 us)

Tue, 14 Sep 2021 10:50:39 CST

(root@nebula) [test1]> lookup on test4 where test4.col1 == 2 yield test4.col1;
Empty set (time spent 12313/14579 us)

Tue, 14 Sep 2021 11:05:37 CST

Seems like a bug which introduced in rebuild step. We could try to fix it in v2.6. @Sophie-Xie

case4: default value

The latest act should be the right one.

[Sophie-Xie]

case 1. create a new issue #2859

case 2. create a new issue #2873

case3. pls fix it in this issue.

other, #2641

[Sophie-Xie]

There are other high priority issues to do, this issue will be later to next version.

[critical27]

All schema problem, we'll try to fix in the version after 1230 by using Online DDL like Google F1

[Sophie-Xie]

Talked it offline, all schema problem will fixed in 5.x.

[wenhaocs]

For case 1, we need to check all the history schemas and make sure no same properties exist when adding properties to an existing schema.