[#882] Separate csv metadata map by type #891
Conversation
|
We just merged the PR fixing our licence block header. From |
| // The properties are incompatible to get a conversion error | ||
| // if the metadata is not separated | ||
| FlinkAsciiGraphLoader loader = getLoaderFromString( | ||
| "graph[" + |
There was a problem hiding this comment.
Pls add multiple entities with same label to test that case as well.
Your example Graph posted within your Issue #882 would fit.
test[
(v1:A {keya:1, keyb:2, keyc:\"Foo\"}),
(v2:A {keya:1.2f, keyb:\"Bar\", keyc:2.3f}),
(v3:A {keya:\"Bar\", keyb:true}),
(v1)-[e1:A {keya:14, keyb:3, keyc:\"Foo\"}]->(v1),
(v1)-[e2:a {keya:1.1f, keyb:\"Bar\", keyc:2.5f}]->(v1),
(v1)-[e3:a {keya:true, keyb:3.13f}]->(v1)
]
- Multiple vertex entities with label 'A'
- Multiple edge entities with label 'a'
There was a problem hiding this comment.
The problem with the test graph is it doesn't fail reliably. It only sometimes fails because edge e1 is compatible to the metadata of v:A but not the other way around. The 'e:a' edges don't have any 'v:a' vertices to collide with, only edge 'e1:A' conflicts with the vertices.
graph[
(v1:A {keya:true, keyb:1, keyc:\"Foo\"}),
(v2:A {keya:false, keyb:2.1f}),
(v3:B {keya:2.3f, keyb:\"Bar\"}),
(v4:B {keya:1.1d}),
(v1)-[e1:A {keya:1, keyb:\"Foo\", keyc:3.3d}]->(v2),
(v2)-[e2:A {keya:2, keyb:7.2d}]->(v1),
(v2)-[e3:B {keya:false}]->(v3),
(v3)-[e4:B {keya:true, keyb:13}]->(v4)
]
This should be guaranteed to fail on 'keya' with a conversion error and adds multiple entities.
There was a problem hiding this comment.
Allright, then use your proposed graph for testing.
| // The properties are incompatible to get a conversion error | ||
| // if the metadata is not separated | ||
| FlinkAsciiGraphLoader loader = getLoaderFromString( | ||
| "graph[" + |
A tuple of type and label is used as key in the metadata map to separate vertices and edges using the same label. Both Source and Sink failed/produced wrong results if edges and vertices use the same label.
Fixes #882