Code efficiency (R)

[kubu4]

Well, I took a version of this functional, but messy and redundant code which was ~190 lines of code (snippet):

for (comparison in comparisons) {
  
  # Create data frame with coefficients of variation for each Treatment group
  for (index in 1:length(transcripts_treatment_fpkm_list))  {
    if (comparison == "all") {
      # Set treatment column name
      treatment_col <- paste("Treatment", index, "_", "FPKM_Coeff_of_Var", sep = "")
    
      # Checks if data frame exists
      if (exists("transcripts_treatment_fpkm_CoV_df")) {
      
        # Add new column with coefficients of variation for each subsequent TreatmentN
        transcripts_treatment_fpkm_CoV_df$column <- rowcvs(transcripts_treatment_fpkm_list[[index]], unbiased = TRUE)
      } else
          # Creates initial data frame with coefficients of variation for FPKM in Treatment 1
          transcripts_treatment_fpkm_CoV_df <- as.data.frame(rowcvs(transcripts_treatment_fpkm_list[[index]], unbiased = TRUE))
      
      # Checks if data frame exists
      if (exists("genes_treatment_fpkm_CoV_df")) {
      
        # Add new column with coefficients of variation for each subsequent TreatmentN
        genes_treatment_fpkm_CoV_df$column <- rowcvs(genes_treatment_fpkm_list[[index]], unbiased = TRUE)
      } else
          # Creates initial data frame with coefficients of variation for FPKM in Treatment 1
          genes_treatment_fpkm_CoV_df <- as.data.frame(rowcvs(genes_treatment_fpkm_list[[index]], unbiased = TRUE))
    
      # Renames columns
      names(transcripts_treatment_fpkm_CoV_df)[index] <- treatment_col
      names(genes_treatment_fpkm_CoV_df)[index] <- treatment_col
    }
    
    # Treatments 2&4 (females)
    if (comparison == "f_v_m" & (index == "2" || index == "4")) {
      
      # If matrix doesn't exist, create it
      if (!exists("transcripts_females_matrix")) {
        transcripts_females_matrix <- transcripts_treatment_fpkm_list[[index]]
        
        # If matrix does exist, append column to matrix
      } else
          transcripts_females_matrix <- cbind(transcripts_treatment_fpkm_list[[index]])
      
      # If matrix doesn't exist, create it
      if (!exists("genes_females_matrix")) {
        genes_females_matrix <- genes_treatment_fpkm_list[[index]]
        
        # If matrix does exist, append column to matrix
      } else
          genes_females_matrix <- cbind(genes_treatment_fpkm_list[[index]])

---

And, I reconfigured it to this beauty (for very similar data set/calculations), which is only ~60 lines of code and far more legible/manageable:

for (comparison in comparisons) {
  
  # Parse comparisons to use subsequent strings for
  # column labels and file naming
  
  ## Get the left side of the comparison
  first_comparison <- sapply(
    strsplit(comparison, "_"),
    head,1
    )
  
  ## Get the right side of the comparison
  second_comparison <- sapply(
    strsplit(comparison,"_"), 
    tail,1
    )
  
  # Set first treatment column name
  first_treatment_col <- paste(first_comparison,
                         "mean.methylation.CoV",
                         sep = ".")
  
  # Set second treatment column name
  second_treatment_col <- paste(second_comparison,
                         "mean.methylation.CoV",
                         sep = ".")
  
  ##### For all samples #####
  if (comparison == "all") {
    
    # Set treatment column name
    treatment_col <- paste("all", "mean.methylation.CoV", sep = ".")
    
    # Calculate coefficients of variation using rowcvs(), convert to data frame and add to list
    # Requires list() to add data frame to list and keep data frame class
    methylation.transposed.rownames.list[[comparison]][treatment_col] <-
      as.data.frame(
        rowcvs(
          methylation.transposed.matrix,
          unbiased = TRUE
          )
        )

  } else {
    
    # Calculate coefficients of variation using rowcvs(), convert to data frame and add to list
    # Requires list() to add data frame to list and keep data frame class
    # Uses get() to convert vector name to string
    methylation.transposed.rownames.list[[comparison]][first_treatment_col] <-
      as.data.frame(
        methylation.transposed.matrix %>%
          subset(., TRUE, get(first_comparison)) %>% 
        rowcvs(., unbiased = TRUE)
        )
    
    # Calculate coefficients of variation using rowcvs(), convert to data frame and add to list
    # Requires list() to add data frame to list and keep data frame class
    # Uses get() to convert vector name to string
    methylation.transposed.rownames.list[[comparison]][second_treatment_col] <-
      as.data.frame(
        methylation.transposed.matrix %>%
          subset(., TRUE, get(second_comparison)) %>%
        rowcvs(., unbiased = TRUE)
      )
    }
  }

---

Also, the code refactoring I've done will now also allow me to easily write all the data frames in the list (using a one-liner with one of the apply() functions) to different files, by using the names assigned to each data frame in the list.

The key to reconfiguring the code primarily thinking about ways to re-use information, while trying to keep the data manipulations as uniform as possible. This way, I didn't have to re-write variations of the same code to handle different scenarios.