all current Combine Multiple example custom variables we have have blank 'relationship' values in their definitions

[Viqsi]

Every 'multiple' type custom variable currently in the d.m.o test database (all created by @WValenti, incidentally) that has more than one contributing variable (there's one of them with only one variable, go figure) has some kind of similar issue going on with the "relationship" field. Discovered this as part of testing fixes for #184 - the schema can't validate them because we keep getting relationship: "" showing up in those value definitions.

I'm not sure how API_ParseJSONDefinition let that fly, since the literal 'AND' and 'OR' are used in the generated SQL. Or maybe it doesn't and those variables never actually worked - they certainly don't have any valid AECs I've ever seen. So I'm not sure if this is a bug in diverweb's Combine Multiple form (most probable IMO), a bug in API_ParseJSONDefinition (can't be ruled out), a bug in the schema (doubt this), or something else entirely. Either way I can't see how they could have ever worked previously.

The affected variables are custom variable IDs 178 ('Pain_D2_v4'), 183 ('athrue'), and 185 ('Pain_D2_v5'). I have not yet attempted to reproduce this with the Combine Multiple form as it currently stands because frankly that thing's a hidden thing kept around only at @WValenti's insistence and so there are much higher priorities.

See also #72. I think that might be related. (Indeed, this started out as a comment to that issue.)

[Viqsi]

The schema's been updated (as of MathematicalMedicine/diverRPC@453c1a6) to assume that unless otherwise specified the relationship is "OR", which is how the code seems to operate. Did this because the same behavior is present in some cases of Combine Two Variables.

So I'm calling this done as a result.

[Viqsi]

Er, no, reopening. I misremembered what this issue is - it's an empty relationship, not a missing field. (I just want to sweep so much aside with the work I've put in; can you blame me?)