Closed bennahugo closed 1 year ago
I suspect I have no choice but to split out the data into a single field database. I think the selection mechanism is not working as expected because it is not using TaQL underneath the hood and then tries to infer shapes based on first row? Why not do that from the meta data keyword tables and the coldescriptors and/or nrow?
Yup as I expected splitting to a single field database works (or at least it runs past that point...). @JSKenyon probably best to implement the selection mechanisms via taql for this reason.
confirmed single field database runs through without a hitch
I made a decision to excise all TAQL from QuartiCal to avoid compatibility issues with the various backends. I appreciate that this does need to be fixed. Could you please point me at the data so that I can reproduce @bennahugo?
I have manage to reproduce - the issue is not with the selection but with the assert statement. The assert is relying on the output of xds_from_storage_ms
which will omit the names of columns which are not initialised (i.e. columns which exist but haven't been filled, if the exemplar row it reads is in unfilled region). I am working on a fix.
@bennahugo This should be on the stimelation
branch now if you want to check that it has resolved your problem.
Running Commit #a2bb83c (stimelation)
input_ms.select_fields is set to [2] in this case (there are 4 fields available). I have filled DE_MODEL{1,2,3} with data for this field through crystalball, e.g.
I get
the following shows the columns are correctly filled
full command (quotes needed to run in a z-shell without escaping)