About "conflicting directions specified for collection"

[Beansprouts]

This is a small part of my graph,every vertex may has a parent and children.

my case : we have a start vertex,then we need to find other verties have the same parent with the start vertex.

In old version,we can do this by an aql: for sv in subject filter v.name="C3" for v,e,p in 2 any sv outbound r_parent,inbound r_parent return p

But,in version 3.0.4 the same aql cannot work as before,just got a exception message "conflicting directions specified for collection".

It's a new feature! 

But,I think it's an bug.

[Jan]

Hi,

I have tried the query and got some parse errors in the part before the traversal.
I think the first part should read

    FOR sv IN subject FILTER sv.name == "C3"

(filtering on "sv" and not "v", and using a double equal sign to make the expression a comparison and not an assignment).

Apart from that, the traversal part specifies "sv" as the edge collection and "any" as the traversal direction. Then it specifies "r_parent" for outbound traversal and "r_parent" again for inbound traversal.
Though the combination of the "inbound" and "outbound" directions is the same as "any", I think the traversal part can be simplified to just

  FOR v,e,p IN 2 ANY sv r_parent RETURN p

The query parser just checks for whether multiple directions are specified for the same collection and will report an error if this is so.

Best regards
Jan

[Beansprouts]

Sorry,I made a misstake on the first part of my aql.

if i just use an "any" direction,the set of results will be bigger than that i need.

在 2016年8月26日星期五 UTC+8下午3:04:03，Jan写道：

[Jan]

Hi,

I think the results of

    FOR sv IN subject FILTER sv.name == "C3" FOR v,e,p IN 2 ANY sv r_parent RETURN p

should be identical to the results of

    FOR sv IN subject FILTER sv.name == "C3" FOR v,e,p IN 2 ANY sv INBOUND r_parent OUTBOUND r_parent RETURN p

apart from order of results inside the result set.
This also assumes that there are no edges in r_parent that have identical values in _from and _to in the same edge (self-referencing edges).

Can you provide some example data for which the query just using the collection once with "any" produces too many results?

Best regards
Jan

[Beansprouts]

Oh,I am quite at fault. Thanks.

在 2016年8月26日星期五 UTC+8下午3:36:33，Jan写道：

[Simran Brucherseifer]

Shouldn't there be two traversals? One to get from the start node to the parent node (outbound), and then another from that parent node down to all children (inbound)?