Dimension Optimization

Test every valid position for one specific dimension while keeping others fixed. It's the mirror of Position Optimization: instead of asking "what dimension should go in this slot?", it asks "where should this dimension go?".

When to use

• You suspect a specific dimension is in the wrong slot and want to find its best home.
• After a greedy run, you want to double-check that one specific dimension's position is justified.
• A new dimension was added to the cube and you want to see where it fits best.

JSON config example

{
  "instance": "tm1srv01",
  "cube": "Sales",
  "executions": 5,
  "output": "csv",
  "optimize_dimension": "Customer"
}

optimize_dimension must be a valid dimension name in the cube. OptimusPy validates this on startup, so typos fail fast with a clear error.

What it does

OptimusPy keeps every other dimension fixed in its current position and slides the target dimension through every position, measuring at each one.

For an 8-dimension cube that's 7 evaluations, one per position other than the one the dimension is in today (the current position is what the original-order baseline measures, so there's no point testing it again). That's a fraction of a full greedy run, which measures 27 orders on the same cube.

Interaction with the locked slot

The constraint is on the slot, not on the dimension: if the dimension sitting last in the storage order has string elements, that last slot is locked and nothing else can move into it. So on a cube with a locked slot, the sweep skips the last position and tests one fewer (6 positions on an 8-dimension cube). A dimension that carries string elements somewhere else in the order is tried in every open position except the last, because TM1 refuses to move a string dimension into the last slot. On a cube whose last dimension is numeric, optimizing such a dimension tests every position but the last.

One thing to bear in mind is that if optimize_dimension names the locked dimension itself, every move of it is refused, so the run evaluates nothing and the log reports the skipped candidates. See The Locked Slot.

CLI

optimuspy optimize sales_dimension.json

The repository includes samples/optimize_dimension.json as a starting point.