csiro-crop-informatics / granular Goto Github PK

On the output data tab, put some indication of where the analysis is up to

Need peaks to be able to be entered ad hoc, so sorting happens on the server

Need to iterate over dataframe

Now that we can input the peak x values, when mixdist is run it would be useful to estimate the peak heights for each distribution and use them to calculate prior weights for each underlying dist. Can just get proportion for each peak_val. Need to work out how that matches a weight.

the output data still exists after refresh

This is messing up bind_rows in server.R

all the javascript code is in html and the variables are global. change the javascript so that it is more modular.

The values returned from the line-hover functions are incorrect (in terms of matching the underlying data). It's probably due to the transformation that's happening with the zooming, which is also affecting the axis after brushing.

@Seanli52 could you have a look at that?

Will be more maintainable and distributable as a package. Also... documentation!

With the new functionality to select example data, multiple datasets can be passed to the browser as different selections are made. each time an example set is chosen, a new svg is added in the browser. The paths for the new data are added on top of the existing ones.

This needs to be reconfigured so that the old graph is removed before the new one is drawn.

@Seanli52 or @rajgaire would you be able to have a look?

Need to be able to distribute without requiring shinytest

mix_dist should be able to exist on the end of a magrittr chain. Needs to handle column names for size and dist.

after f827d6e clicking the reset button is triggering the analysis. Need to look at: rstudio/shiny#167

Tidy dataframes will be the easiest default data structure. Having the size and distribution data stored as a list column could be useful. Checking for whether the column is list type and unlisting if necessary should be straightforward

The lines from samples (inst/shiny/www/index.html) are correct (they stop at the end of the x axis) prior to zoom, but they extend too far afterwards

At the moment, a dataframe is processed in a loop, so that only the full results are visible to shiny.

Would probably be neater to structure in a list, so that each sample is an element, and the output is built up as the list is iterated over. The output can then be visible in the browser as the analysis is happening.

Need to be more flexible with data formats, or at least explicitly describe requirements and test for them, giving a useful response if they're wrong. Crashing atm.

mixDist currently takes a dataframe as its second argument. Would work better when iterating with list if this was just a vector, then there is no looping in the function. Also change argument order so dist data is first (helps with %>%)