How to make a group by in cypher query language?

[naughtyGitCat]

I have these relations:
NodeType1 A {ts:11} =r1=> NodeType2 m
NodeType1 B {ts: 22} =r1=> NodeType2 m
NodeType1 Y {ts: 22} =r1=> NodeType2 n
NodeType1 Z {ts: 32} =r1=> NodeType2 n
NodeType1 X {ts: 23} =r1=> NodeType2 o
.......

For each NodeType2, I wish to get max/min ts property NodeType1 who relate to it by r1

[naughtyGitCat]

try use call and unwind,but not fit

[intouch_vivek]

[quote="naughtyGitCat, post:1, topic:13725"]
why not you write parameterized query?
Could u please let me know ts is property of node or relationship?

[naughtyGitCat]

ts => timestamp
means NodeType1 node create time

[intouch_vivek]

naughtyGitCat:

NodeType1 Z {ts: 32} =r1=> NodeType2 n
NodeType1 Z {ts: 23} =r1=> NodeType2 o

How can your node Z has two 2 create time?

[naughtyGitCat]

I'm sorry it's a mistake

NodeType1 A {ts:11} =r1=> NodeType2 m
NodeType1 B {ts: 22} =r1=> NodeType2 m
NodeType1 Y {ts: 22} =r1=> NodeType2 n
NodeType1 Z {ts: 32} =r1=> NodeType2 n
NodeType1 X {ts: 23} =r1=> NodeType2 o

[intouch_vivek]

I tried below
try below:

1. Load:
   Merge(:NodeType1 {name:'A', ts:11}) -[:r1]-> (n1:NodeType2 {name:'m'});
   Merge(:NodeType1 {name:'B', ts: 22}) -[:r1]-> (n1:NodeType2 {name:'m'});
   merge(:NodeType1 {name:'Y', ts: 22}) -[:r1]-> (n1:NodeType2 {name:'n'});
   merge(:NodeType1 {name:'Z', ts: 32}) -[:r1]-> (n1:NodeType2 {name:'n'});
   merge(:NodeType1 {name:'X', ts: 23}) -[:r1]-> (n1:NodeType2 {name:'o'});

2. Merge Nodes:
   MATCH (n1:NodeType2)
   WITH n1.name as name, collect(n1) as nodes
   CALL apoc.refactor.mergeNodes(nodes, {properties: "combine"}) YIELD node
   RETURN node

MATCH (n1:NodeType1)
WITH n1.name as name, collect(n1) as nodes
CALL apoc.refactor.mergeNodes(nodes, {properties: "combine"}) YIELD node
RETURN node

3. Aggregation Query: Example for node m
   match (n1:NodeType1 ) -[r1]-> (n2:NodeType2 {name:'m'})
   with n2, max(n1.ts) as maxval, min(n1.ts) as minval
   return n2, maxval,minval

[naughtyGitCat]

Yeah,

for NodeType2 {
   // run cypher to find max(ts) NodeType1;
}

is a method,

but need many times net io, and cause inconsistency

I mean do run cypher one time, to find group by type2 max(type1)

[naughtyGitCat]

your code means

select max(type1) where type2.name='xxx'

No offense, but I want to do this way,

select max(type1) group by type2.name

[intouch_vivek]

I did not understand where did u get
select max(type1) where type2.name='xxx'

In my query as below, with n2 means you group by n2 node given in the match

match (n1:NodeType1 ) -[r1]-> (n2:NodeType2 {name:'m'})
with n2, max(n1.ts) as maxval, min(n1.ts) as minval
return n2, maxval,minval

[naughtyGitCat]

yes, but the NodeType2 name is not specified at first, in database, there is a lot of NodeType2 node but not all named by m

[intouch_vivek]

As I mentioned us parameterized query. You can pass any value in parameter for an example i passed m

1. :param node2:'m'

2. match (n1:NodeType1 ) -[r1]-> (n2:NodeType2 {name:$node2)
   with n2, max(n1.ts) as maxval, min(n1.ts) as minval
   return n2, maxval,minval

[naughtyGitCat]

Yes。We can send many times parameterized query to neo4j server, But this need many times net io, and cause inconsistency, unless do a explict trx.begin(), Is there any way to send a non parameterized query to finish it

[intouch_vivek]

Try match (n1:NodeType1 ) -[r1]-> (n2:NodeType2)
with n2, max(n1.ts) as maxval, min(n1.ts) as minval
return n2, maxval,minval

[naughtyGitCat]

why not you write parameterized query?

you mean do forloop or foreach in program logic?

[stefan_armbrust]

Your question's title is fairly misleading:

1. there's no "sql group" in Neo4j
2. there's also no CQL.

The language you're using is called cypher. I suspect your asking for aggregation functions, see https://neo4j.com/docs/cypher-manual/current/functions/aggregating/

[naughtyGitCat]

cypher query language => CQL?
GROUP BY is a calc method, dont care about the name

[naughtyGitCat]

and agggregate only run on a specific node,but not unknown nodes

[tms]

I understood what he meant, I have the same question.

I've blown a "merge" statement, so now I have too many nodes (with a specific label) to "DETACH DELETE" in one transaction. So I want to do an operation that in SQL I would accomplish with a "GROUP BY" -- I want to somehow match a group of nodes based on an attribute, "DETACH DELETE" each of those (in a single transaction), then iterate until all the nodes are gone.

I'm new to Cypher, and so I'm still struggling to understand the semantics of these group operators.

On a related note, is there any way to perform multiple transactions (inside a subquery or something) from the neobrowser (the tool that comes up on port 7474)?

[tard_gabriel]

MATCH (n:Label {key:value}) <- This match a group of nodes based on an attribute and a label
RETURN n