Is there an implicit limit on the number of series() calls a script can make?
Would it be a number somewhere between 255 and 259 ?

I have a script with 259 series calls. It crashes in Train mode in unpredictable ways.
1. A Windows dialog tells me the program has crashed, or
2. I get a pile of runtime errors, followed by (1), or
3. Zorro just freezes, requiring task kill.

If the failure mode is 1 or 2, at least, the diag file has something in it. Zorro's last breath is


If I bring the number of series calls down to 255, Train succeeds.

There is no fixed limit, but of course all data on a PC is limited by the available memory. However even on a very old PC this limit is far higher than 259. Exceeding it will produce an error message, not a crash.

Writing or reading past the length of a series however will cause a crash - that's where I'd look first.

Come to think of it, I have executed a script containing 652 distinct series() calls, successfully. At least, without apparent error. This script wrote values to a file for offline feature selection.
Now I'm looping back with the chosen features and trying to get a DTREE on them.

Maybe there is a bad interaction when there are >259 series calls followed by calls
adviseLong(DTREE, 0, ...) and
adviseShort()

Or maybe the limit only appears in Train mode, not Test.

No, I'm afraid a bad interaction or a limit that only appears in Train mode are also unlikely.

Thanks.
I'll dig deeper.
In general, is this allowed?

Train.

After many hours isolating the problem, I am able to reproduce at least one of the crash conditions with a script that has only 16 series() calls, and sets only set(RULES).
I'll post the example script in a new thread in the Scripts forum.
http://www.opserver.de/ubb7/ubbthreads.php?ubb=showflat&Number=431279&#Post431279