The problem with VM.builder

[sergeikharchenko]

Hi!

I found a bug with that function.
It doesn't work properly if you have a marker's name, which the grepl function finds twice or more in colnames.

For example: my dataset has such colnames:
Sample_name, Class_decontamination, S, K, Ca, Ti, V, Material

You have your own function VM.proportionate.prop.values here (inside VM.builder):
resu <- fingR::VM.proportionate.prop.values(data = df.sources, class = class, tracers = tracers, contributions = contributions, VM.name = VM.name)

And it contains such code:
for (var in tracers) { df.value[, which(grepl(var, colnames(df.value)))] <- round(df.value[, which(grepl(var, colnames(df.value)))], fingR::lvl.signif(data[[var]])) df.SD[, which(grepl(var, colnames(df.SD)))] <- round(df.SD[, which(grepl(var, colnames(df.SD)))], fingR::lvl.signif(data[[var]])) }
and
for (var in tracers) { df.value[[which(grepl(var, colnames(df.value)))]] <- round(df.value[[which(grepl(var, colnames(df.value)))]], fingR::lvl.signif(data[[var]])) }

In this code and maybe somewhere else grepl gives you two indicies for var == "S" ---- "Sample_name" and "S". And after that the code uses wrong column (first of the ones).

Of coarse I've renamed my variable from "S" to "Sulfur", and it works good.
But it can be useful for you.

Best regards and many thanks for the tool again.
Sergey Kharchenko

[tchalauxclergue]

Hi Sergey,

Thank you very much for pointing this out! I'll patch this issue as soon as possible.

Best regards,
Thomas Chalaux-Clergue

[tchalauxclergue]

Hey Sergey, it should now be patched with the latest update (ver. 2.1.2).
The function VM.proportionate.prop.values has been adapted to search for property and uncertainty labels more precisely.
However, if multiple property or uncertainty labels are identified, the function is stopped and the user is notified. An error message is generated to notify the user of these mistaken labels and ask them to modify them accordingly.
It is also now possible, although it's optional, to specify how uncertainty is indicated by a suffix OR a prefix in the label. If different suffixes or prefixes are used, these can be listed in a vector (e.g. c("_SD", "_SD", "_RMSE", etc)).
Could you confirm that it is now working? Other functions use the grepl function, so you may find other issues later on...
Best, Thomas