marray iterator syntax not properly supported

[Dimitar Misev]

This works

select encode(
marray i in [0:100]
values i[0]
, "csv")

however this doesn't, but should work as well equivalent as the above:

select encode(
marray i in [0:100]
values i
, "csv")

Same with the condenser iterators.

In fact I'm not sure why and if we want to support the first case at all (aside of backwards-compatibility).

[Peter Baumann]

yes, we do need both (again, actually, as it used to work already):

• iteration vector is needed for

  marray p in [0:99, 0:99, 0:99]
  values p[0]+p[1]+p[2]
  

• single values are needed for

  marray x in [0:99], y in [0:99]
  values x+y
  

[Peter Baumann]

…with the following semantics check:

• an iteration variable without indexing is allowed only on 1D iterators
• indexing is required on iterators >1D

Also, forgot to say that we might have mixed cases, too:

marray p in [0:99, 0:99, 0:99], time[0:99]
values p[0]+p[1]+p[2]+time

[Dimitar Misev]

The

marray p in [0:99, 0:99, 0:99]
values p[0]+p[1]+p[2]

is quite confusing to newcomers from what I observed, they tend to think that [0:99, 0:99, 0:99] is therefore some sort of array. Is there some reason for this notation? Why not simply enforce

{{{
marray p in [0:99], q in [0:99], r in [0:99]
values p+q+r
}}}

[Peter Baumann]

no need to expose to newcomers, but it is already in the syntax and we should not loose it.

[Dimitar Misev]

I think we should close this (in favor of the mda syntax).