Order of labels in signifPlot

[ellispatrick]

I think signifPlot changes the order of the cell type labels. This might have something to do when it turns from and to into factors. Mabe this info needs to be retained in the spicy call?

[swbioinf]

Confirmed. I was just typing up an issue for this. Should be an easy fix for a quite sneaky frustrating little bug!

When the df.shape object is constructed for building the semicircle legend label, it looses the factor-ness of condition.

df.shape <- data.frame(
    cellTypeA = c(NA, NA), cellTypeB = c(NA, NA), size = c(1, 1),
    condition = c(
      levels(test$condition)[1], levels(test$condition)[coef]
    )
  )

Then when its plotted, the circles are plotted after a cast to a factor, not the presented order. So sometimes its wrong. At least if your test condition is alphabetically before 'Control'!

    ggplot2::geom_point(
      data = df.shape, ggplot2::aes(shape = condition), x = 10000, y = 10000,
      key_glyph = spicyR:::draw_key_half_circle 
    ) 

Checked this the values of my two groups vs the L values extracted by the 'bind' function.

I've tested this by mucking around in the bubblePlot function and turning df.shape$condition into an appropriately ordered factor. And that works, but can't control it from the outside. Fix might look like:

df.shape$condition <- factor(df.shape$condition, levels=df.shape$condition)

[swbioinf]

Ah! Looks like this was a duplicate of #73 which was already fixed in commit 4c8c1da .

I was using v1.22.0 from bioconductor.

I've resinstalled from github and it all works :)

NB: This is reported as version 1.21.8.

Thanks!