sum_next_n and roll don't accept parameters as the offset

[jmorrisnrel]

What happened?

Trying to use a user-defined parameter as the offset for both sum_next_n and roll fail with different error messages.

Below is an example of constraints for user-defined minimum up and downtime limits on technologies. The user defined parameter in this instance needs to control the number of timesteps the constraints sum on a per-tech/node basis.

parameters:
  minimum_uptime:
    default: 0
    resample_method: mean
    title: Minimum uptime.
    description: >-
      Minimum number of consecutive timesteps a unit must remain online after starting up.
      Only valid for technologies with integer dispatch and cap_method set to 'integer'.
    unit: timesteps.

  minimum_downtime:
    default: 0
    resample_method: mean
    title: Minimum downtime.
    description: >-
      Minimum number of consecutive timesteps a unit must remain offline after shutting down.
      Only valid for technologies with integer dispatch and cap_method set to 'integer'.
    unit: timesteps.

constraints:
  minumum_uptime_integer_dispatch:
    description: >
      Enforce minimum uptime for integer dispatch technologies.
    foreach: [nodes, techs, timesteps]
    where: "integer_dispatch AND minimum_uptime"
    equations:
      - expression: > 
           operating_units >= sum_next_n(unit_start, timesteps, 4)

  minumum_downtime_integer_dispatch:
    description: >
      Enforce minimum downtime for integer dispatch technologies.
    foreach: [nodes, techs, timesteps]
    where: "integer_dispatch AND minimum_downtime"
    equations:
      - expression: > 
           operating_units + sum_next_n(unit_stop, timesteps, minimum_downtime) <= purchased_units

Trying to actually pass an integer parameter value into sum_next_n results in a
TypeError: only length-1 arrays can be converted to Python scalars

It looks like the entire parameter DataArray gets passed in as N in the SumNexN.as_array() function, so it is parsing the parameter name in the kwarg, but it doesn't index it to get the value for the specific [node, tech] in the constraint foreach loop.

roll doesn't seem to be parsing the kwarg at all and just fails when it tries to convert the parameter name to an int.
ValueError: invalid literal for int() with base 10: 'test_offset'

I haven't tested the other helper functions but assume get_val_at_index would also have the issue.

Which operating systems have you used?

• macOS
• Windows
• Linux

Version

v0.7.0.dev7

Relevant log output

[brynpickering]

@jmorrisnrel Correct. Not a bug, just the current state of the feature. It would require applying the shift/sum/roll over an arbitrary grouping of the other dimensions in the helper function, which is not trivial.