How to speed up Data Table Operations (e.g Stack, Split, ...)?

[#1]

I often need to Stack very large tables with millions of rows, do some operation and then Split the table.
What options are there to speed all this up besides working with "invisible" data tables in the jsl code?

I'm familiar with the command Begin/End Data Update, however I'm not sure it's relevant here or has any improvement on top of invisible tables.

[#2]

Invisible (or Private) are good starting points and then it depends a lot on your tables, sometimes you might be even able to avoid split/stack. 

[#3]

My goal is to Coalesce Columns which represent the same column, meaning I want to merge columns which represent the same column. I attached an example table where each of the columns Age, Weight, Height is spread across 2 columns. the goal is to merge each couple of the columns.
originally I would stack the columns, cut out the relevant part (Age/Weight/Height) and then split back, however this becomes a problem with very large data tables.

[#4]

Have you tried Query Builder? 

Sorry, I misunderstood the question. I don't think Query Builder helps here. 

[#5 · Accepted solution]

Names Default To Here(1);

dt = Open("$DOWNLOADS/example_input.jmp");
valid_cols = 4;

m = dt[0,0];
r = Transform Each({vals}, m,
	nonmissing = Where(!IsMissing(vals));
	vals[nonmissing];
);

dt[0, 1::valid_cols] = r;
dt << Delete Columns(valid_cols::N Cols(dt));

[#6 · Accepted solution]

And if this is how the data really looks like

1. Pick your "duplicate" columns
2. Make a subset with them + Id column (might have to create one if you don't already have one)
3. Drop rows which have missing values
4. Rename your columns to be same in both tables (+drop extra columns)
5. Use Update on the original table with the ID column
6. Close the subset