Closed SheenaTomko closed 3 years ago
Sheena, we added a default bin width of 1 to integer vars whose bin width (coming out of the database) was a floating point number. If we think that is resolved, I will take this out and throw and exception when we see binWidth and binWidthOverrides of floats on integer vars.
there was a plot.data bug, which has since been resolved. looking at that same term as in the screenshot, but in the viz tab, i see a bin width of 3000 w no filter applied. if i look in eda-inc attributegraph table for this term for this study also says 3000..
There is a fix for this committed to master, however we are holding off deploying it pending reload of all studies, which should provide valid values for this metadata. Subsetting service will then catch any missing values and hopefully throw exceptions with helpful messages on study read.
Jay just reloaded India ICEMR Fever Surveillance today. Looking at 'Plasmodium falciparum asexual stages, by microscopy result' in that study, I think it took >1 min to load and I got the google chrome popup telling me the page was unresponsive, did I want to wait or kill it. I waited and got the below screenshot, still with the bin width of 1. Looking in the vis tab, that bin width was 20,444.
Ryan, is the fix you have waiting supposed to fix this? So in theory, the metadata should be ok now and you'll remove the binWidth of 1?
Danielle said she would check out the studies individually using the code branch to see if any studies are still non-compliant. Then we can merge this in. @d-callan I am off Friday; if you test, feel free to merge, tag, and update tagger. Otherwise, I can do it Monday.
This looks good. Variables that previously had a bin width of 1 now have bin widths that make sense based on the range of the data.
The screenshot example is from India ICEMR Cohort, but I also saw the same thing in other studies like Score S. mansoni CRT, variable Birth year.
Basically, when you click on the variable, it takes a few seconds to load and seems to be defaulting to a bin width of 1, even if that doesn't make sense because of the range and volume of the data. You also see a red outline around the bin width box, which I think indicates that you're exceeding the plotting limits (not supposed to use a bin width that small). I checked, and these variables do not have a defaultBinWidth set by the ClinEpi Outreach team.