Revisiting an earlier request

This is something just to put on the ultra low priority which might be looked at when it’s raining and there is nothing else to do. I have no immediate requirement, but it seems to me it would open up quite a number of possibilities. A similar request (but more complicated) was floated in April 21, and various work rounds were proposed.

The suggestion is that on completion of processing a ‘record’ (for want of a better word) the values are cached. Thus when a new record is read in, the values of the prior record would still be held in memory. Then as well as being able to reference the content of a current row using $(x), #(x) {picking a symbol at random…} would return the prior cached value of that column.

Data analysts will be gritting their collective teeth at the very notion of juxtaposition having any meaning (so apologies) but the perfect is the enemy of the good, and I can think of many uses where serial / pseudo-serial processing could be handled by this. {e.g. time-series processing like computing exponential averages and movement deltas).

Just a thought, but this seems to be far less implementation heavy than the April 21 request of having access to any prior value, but pretty useful nonetheless.

I’m not sure using special symbol for the row below is an improvement over using Copy Cols and Offset. Also it wouldn’t handle when you want to, say, average the 10 rows above.

A better approach might be a new transform that can create a new column with a moving sum/average/minimum/maximum etc (perhaps exponential moving average) of a selected column over N rows. I’m not sure what you would call such a transform. Running ? Moving ? Rolling?

Can you link to the April 21 post? I couldn’t find it.

I have added a new Moving transform which does moving average/median/sum/min/max. Please try it and let me know what you think.

You may need to refresh the page.