Added 'as.is' passed to reshape2::melt.array().

[fdetsch]

Numeric-style dimnames (see sample data) are currently being converted to numeric by reshape2::melt.array(). Since this is not always desirable, I believe a customization option should be made available to the user.

In my particular case, dimnames '01' to '12' are sample names that I want to appear as categorical rather than continuous features in the resulting plot. Current behavior:

out_tt1.1 = readRDS("C:/Users/florianD/Downloads/out_tt1.1.rds")

ggcorrplot::ggcorrplot(out_tt1.1, type = "upper", p.mat = out_tt1.1
                       , show.diag = TRUE) + 
  scale_fill_gradient2(bquote(bold("p-value")), limit = c(0, 1), midpoint = 0.5
                       , low = "#460000", high =  "#EBEAF7", mid = "#B5BD4C")

With adjustments:

ggcorrplot::ggcorrplot(out_tt1.1, type = "upper", p.mat = out_tt1.1
                       , show.diag = TRUE, as.is = TRUE) + 
  scale_fill_gradient2(bquote(bold("p-value")), limit = c(0, 1), midpoint = 0.5
                       , low = "#460000", high =  "#EBEAF7", mid = "#B5BD4C")

[kassambara]

Thank you for pointing this out! In correlation matrix plot, the axis texts should always be considered as character not continuous variable. Therefore, I think that we should, internally, specify as.is = TRUE in reshape2::melt(), without adding extra argument to ggcorrplot()

[fdetsch]

In general, I agree. However, I thought that for reasons of backward compatibility (ie. in the unlikely event that a user really wanted to have numeric axes), it would be a good idea to add a separate argument. Your call!

[kassambara]

merged now, thanks

[fdetsch]

Awesome, thanks!