show RNAseq type in sample hover

[jklugherz]

[hanars]

So you should be able to do this using the related field on the sample model without having to write query objects with outer refs - something like When(rnaseqtpm__isnull=False)

[jklugherz]

I originally had the __isnull syntax but the query was timing out with more than 1 rnaseq type. I could add indexes on the foreign keys to the rnaseq tables on the sample model?

[hanars]

There should be indices on the foreign keys already, so that sounds like a case where what django is doing is not what we want. You could try using the debug toolbar to see the actual executed SQL and figure out why its so inefficient and adjust it, but honestly that feels like overkill and its probably better not to try to use those fields in this case

That said, if you are going to make 3 distinct queries to each of the subtypes and then iterate through all the sample models and add annotations to them, using the additional_values does not really save us any compute and just makes the whole thing less readable in the end. In that case, you are better off with getting lists of sample Ids for each type and then adding to rnaSeqTypes based on that, something like

rna_type_samples = {
    'TPM': set(RnaSeqTpm.objects.filter(sample__in= sample_models).values_list('sample', flat=True),
    'Expression Outlier': set(RnaSeqOutlier.objects.filter(...
    'Splice Outlier': set(RnaSeqSpliceOutlier.objects.filter(...
}
...
sample['rnaSeqTypes'] = [type for type, sample_ids in rna_type_samples.items() if sample['id'] in sample_ids]

[jklugherz]

I implemented it this way ^ which does not require changes to the JS code as you suggested here.

Is there a general rule for what belongs in additional_values and when it's better to query related tables separately?

[hanars]

hasRnaOutlierData was based on whether there was rnaseqoutlier or rnaseqspliceoutlier data, for samples with only tpm data this would have been false, so we should update the behavior here to reflect that. Also it only considered active samples

[jklugherz]

I adjusted this condition check to match the behavior to show the link only if the sample is active and has either rnaseqoutlier or rnaseqspliceoutlier data

[hanars]

rnaSeqTypes will only have values for RNA samples so you can remove this part of the condition check

[jklugherz]

Without this part of the check some old(?) rna samples which do not have any matches in the 3 rnaseqtype tables will display like this:

There are 76 of these samples all created on 2016-12-05

WITH sample_rnaseq as (
    select s.guid, s.id, s.created_date,
           seqr_rnaseqtpm.id as tpm_id, seqr_rnaseqoutlier.id as outlier_id, sr.id as splice_outlier_id
    from seqr_sample s
    left join seqr_rnaseqtpm on s.id = seqr_rnaseqtpm.sample_id
    left join seqr_rnaseqoutlier on s.id = seqr_rnaseqoutlier.sample_id
    left join public.seqr_rnaseqspliceoutlier sr on s.id = sr.sample_id
    where s.sample_type in ('RNA')
)
SELECT * FROM sample_rnaseq
WHERE tpm_id is null
and outlier_id is null
and splice_outlier_id is null

[jklugherz]

they're all probably from the R0246_rna_seq_bc project, I've been using that for testing since I was certain it would have rna samples given the name

[hanars]

Oh we should only be showing this for active samples, not inactive sample