user_y / fit_obj within for xdds {}

RWI

Hi everyone,

I wanted to be smart and adopt the for xdds {} structure for some simultaneous refinements. Now I seem to run into a problem when trying to refine the contribution of some experimental background:

`
for xdds {
....

xo_Is
	user_y empty_xy { _xy #include "empty.xy" }           
	local  empty_scale  1 min 0.5 max 1.5
	fit_obj fit_empty =  empty_scale  * empty_xy;           
	Plot_Fit_Obj(fit_empty, "empty")

}
`

With this I get a error message that "Parameter empty_xy belongs to a parameter type that must have a unique name."

Any idea how to get around this?

rowlesmr

I haven't used a user_y construct before, but this sounds like an Alan question.

johnsoevans

I've used user_y, but I'm not sure if I've ever applied it across multiple datasets.
I think you can define user_y once outside your xdds:
user_y empty_xy empty.xy
You can then have a fit_obj using the same empty_xy in each xdd

xdd
    prm empty_scale1 0.1
    fit_obj = empty_scale1 * empty_xy;

That will get around the "unique parameter name" error. However I think that you have to use a "prm" not a "local" for empty_scale. Try it and see. If so, you can't hide everything in for_xdds. Even if you could use local it wouldn't save you much space in the INP file as you'd need a "local empty_scale 1" for each xdd, it would just mean you don't have to set up multiple prm names. The simplest I can see at the moment is two lines in each xdd as above.

RWI

Thanks for the reply, I can confirm it behaves as you predicted, except for the part about the user_y, it looks to me as if I have to define it inside xdd.

pam_whitfield

Is there any particular reason you're using a xo_Is for a background contribution?
You can do it that way but xo_Is in a for_xdds loop can get a bit painful in general.
I have run into some issues with user_y inside for_xdds in the past - I don't remember the exact details off the top of my head but the order of the phases can sometimes matter. There's a regression in v7 that Alan said will only be fixed in v8. I'd have to do some digging to see what the exact problem was.
Pam

RWI

@pam_whitfield no idea just always had the xo_Is copied from one .inp to the next over the years, but it certainly works without xo_Is..

Now I naively thought about a way of generating a unique string for within each xdd, and the first thing I thought of is the xdd name itself. This maybe a bit weird but it seems to work:

xdd DATA/scan-0000.xy
xdd DATA/scan-0001.xy
xdd DATA/scan-0002.xy
for xdds  {
	local  scale  1 min  0.95 max 1.05
	macro fit_name
	{
		macro Variable_Name_From_String(Get(xdd_path_name)) {Get(xdd_path_name)}
	}
 	fit_obj fit_name =  scale * empty;       
}	
user_y empty{ _xy #include "empty.xy" }

Only the Plot_Fit_Obj part is now missing, and it wouldn't take
Plot_Fit_Obj(fit_name , "empty"), but I can live without this.

Any thoughts? or ways of making this more elegant?