avoid jumping around wd in the psichomics modules

[hsun3163]

I wonder if the following script can be changed to get rid of setwd so that the wd will not be messed up so easily and give us a headache.

   for (f in c(${_input:ar,})){
      filename = gsub("^.*/", "", f)
      directory = gsub(filename, "", f)
      if (length(files[[directory]]) == 0){
        files[[directory]] = filename
        } else {
      files[[directory]] = append(files[[directory]], filename)
      }
    }
  
    if (length(files) == 1) {
      setwd(names(files)[1])
      res = prepareJunctionQuant(files[[1]])
    } else {
    res = list()
    for (i in 1:(length(files))) {
        d = names(files[i])
        setwd(d)
        res[[d]] = prepareJunctionQuant(files[[d]])
        }
    res = res %>% reduce(full_join, by = "Junction ID")
    }
    
    res[is.na(res)] <- 0

[gaow]

I think i know the answer to this but let me defer that to @Rhopala . The lesson here is that for unconventional behavior like this, @Rhopala you should add a line of comment in the code so others don't have to raise this question but instead can see how they can help ipmrove it.

[gaow]

@hsun3163 we can always let SoS change paths to absolute path eg some_dir = f"{some_dir:a}"

[Rhopala]

@hsun3163 This is for solving another headache. I am not sure if you know there once was a too-long-command problem: R R have an input length limit and the psichomics teams just write their prepareJunctionQuant() function input as (full/path/file1.bam, full/path/file2.bam, full/path/file3.bam ... full/path/file1n.bam) so the real data on BU's end can't be read in the same time. So the setwd codes there get rid of the "full/path"s and allows taking more input since now it only read file names.

[Rhopala]

@gaow Yeah I'll add comments for that problem