Genome Browser: [automatic categorization] South America appears as a different data-type to the other regions

[leehart]

Home > View genome > Add channel

1. Notice that "South America" has a different data-type icon to the others, i.e. a bar-chart rather than a line-graph Add the channels "South America" and "West Africa" (to compare)
2. Notice that the tracks for "South America" and "West Africa" are plotted differently ("South America" is plotted with bars, whereas "West Africa" is plotted with points) and the channels have different options, legend, etc.
3. The settings for the columns shows no difference in dataType

   - id: af_SAM
   dataType: Float
   name: South America
   - id: af_WAF
   dataType: Float
   name: West Africa

   I suspect/guess this might be caused by an internal algorithm for determining categorical-ness for number-type columns, maybe, with "South America" meeting the criteria but the data for other regions do not, maybe.

This issue is not present on the staging instance, so maybe my local version is out-of-date.

[benjeffery]

Yep, SAM has few enough samples that ratios of them don't count to more than 50 different ones. Fixable buy isCategorical : false

[leehart]

I gather testing with fewer variants locally explains the difference with staging. (I suspect that might also invalidate this issue... but I suppose adding isCategorical: false to the settings would prevent this from cropping up again...)

[leehart]

Closing because isCategorical: false has been set for af_SAM for pf_variants, and this was only an issue for incomplete data in testing scenarios. It might be worth noting that data are being categorized by default (effectively isCategorical: true), which is a "hidden feature", which might be potentially unexpected or unwanted. Documentation currently states:

IsCategorical
Boolean. Instructs Panoptes to treat the property as a categorical variable. For example, a combo box with the possible states is automatically shown in queries for this property. Categorical properties are automatically indexed.

(I couldn't see any mention of the default behaviour.)

[leehart]

Reopening because the underlying issue might fester if left swept under the carpet. [Maybe transfer this issue to Panoptes core? Perhaps the solution there is to either document the default behaviour, or to only categorize data when isCategorical is explicitly set true.]