Inconsistent LIMIT operations

[randomJoe211]

Nebula Graph version: 2021.10.18-nightly

LOOKUP ON player LIMIT 2 and LOOKUP ON player | LIMIT 2 return totally different numbers of rows.

1. The newly added LIMIT clause in LOOKUP (without a pipe) is based on the partition level and the returned number of rows is not as specified in the LIMIT clause. This is inappropriate.

(root@nebula) [basketballplayer]> LOOKUP ON player LIMIT 2;
+-------------+
| VertexID    |
+-------------+
| "player100" |
| "player101" |
| "player102" |
| "player103" |
| "player104" |
| "player105" |
| "player106" |
| "player107" |
| "player108" |
| "player109" |
| "player112" |
| "player113" |
| "player114" |
| "player117" |
| "player122" |
| "player124" |
| "player127" |
| "player132" |
| "player150" |
+-------------+
Got 19 rows (time spent 1315/1901 us)

Tue, 19 Oct 2021 09:34:17 UTC

(root@nebula) [basketballplayer]> LOOKUP ON player | LIMIT 2;
+-------------+
| VertexID    |
+-------------+
| "player100" |
| "player101" |
+-------------+
Got 2 rows (time spent 1146/1713 us)

Tue, 19 Oct 2021 09:34:20 UTC

2. Nebula Graph has the LIMIT syntax, the newly added LIMIT syntax conflicts with that of the old one. Now it is only added to LOOKUP, what if it is added to MATCH? Now MATCH required no pipe before LIMIT, if this new syntax is added, there will be a conflict.

(root@nebula) [basketballplayer]> MATCH (v:player) RETURN v | LIMIT 2;
+--------------------------------------------------------+
| v                                                      |
+--------------------------------------------------------+
| ("player105" :player{age: 31, name: "Danny Green"})    |
| ("player109" :player{age: 34, name: "Tiago Splitter"}) |
+--------------------------------------------------------+
Got 2 rows (time spent 2422/2940 us)

Tue, 19 Oct 2021 09:34:44 UTC

(root@nebula) [basketballplayer]> MATCH (v:player) RETURN v LIMIT 2;
+--------------------------------------------------------+
| v                                                      |
+--------------------------------------------------------+
| ("player105" :player{age: 31, name: "Danny Green"})    |
| ("player109" :player{age: 34, name: "Tiago Splitter"}) |
+--------------------------------------------------------+
Got 2 rows (time spent 2105/2599 us)

Tue, 19 Oct 2021 09:34:52 UTC

It is recommended to optimize LIMIT so that its behaviors can be reasonable, logical, and consistent throughout nGQL (openCypher-compatible syntax included).

[bright-starry-sky]

lookup is based on the partition level and returned rows number is partition_num * limit . this is the original design. It's not a bug.
In addition, the purpose of this design is related to the next index paging scan.
And match statement same with generic index scan , that's done in the Graph layer is to deal with the internal limits and the pipe limits.
doc need to be updated if necessary.
close this one.

[jievince]

It is recommended to optimize LIMIT so that its behaviors can be reasonable, logical, and consistent throughout nGQL (openCypher-compatible syntax included).

I totally agree with it！The syntax of the inconsistency will make the user difficult to understand.

[czpmango]

Nebula Graph has the LIMIT syntax, the newly added LIMIT syntax conflicts with that of the old one. Now it is only added to LOOKUP, what if it is added to MATCH? Now MATCH required no pipe before LIMIT, if this new syntax is added, there will be a conflict.

I think that's a reasonable consideration.
We should be careful in grammatical design. Reopen this issue and really hope we can discuss this some more.