Update scripts/paraquet_conversion.R

scripts/paraquet_conversion.R

@@ -10,7 +10,8 @@ library(arrow)
 library(data.table)
 
 # URL of the CSV file
-url <- "https://huggingface.co/datasets/cjerzak/LinkOrgs/resolve/main/PosMatches_mat.csv"
+#url <- "https://huggingface.co/datasets/cjerzak/LinkOrgs/resolve/main/PosMatches_mat.csv"
+url <- "https://huggingface.co/datasets/cjerzak/LinkOrgs/resolve/main/PosMatches_mat_hold.csv"
 
 # Temporary file path to save the downloaded CSV
 temp_file <- tempfile(fileext = ".csv")
@@ -18,23 +19,21 @@ temp_file <- tempfile(fileext = ".csv")
 # Download the CSV file
 download.file(url, temp_file, mode = "wb")
 
-# Read the CSV file. Assuming it may not have headers, we'll add them.
-# If it does have headers, set header = TRUE and skip the colnames line.
-data <- fread(temp_file, header = FALSE)
-
-# Add column names if needed (assuming 256 columns based on common structure)
-if (ncol(data) == 256) {
-  setnames(data, paste("D", 1:256, sep = ""))
-} else {
-  # Adjust based on actual number of columns
-  setnames(data, paste("V", 1:ncol(data), sep = ""))
-}
+# Read the CSV file
+data <- fread(temp_file, header = TRUE)
 
 # Write the data to Parquet format, which is efficient and preferred for large datasets on Hugging Face
-write_parquet(data, "~/Downloads/PosMatches_mat.parquet")
+# write_parquet(data, "~/Downloads/PosMatches_mat.parquet")
+
+# Write a sample
+# samp_ <- sample(1:nrow(data),1000)
+data <- as.data.frame(data)
+data <- data[,colnames(data) %in% c("name1","name2","matchProb") ]
+write_parquet(data, "~/Downloads/PosMatches_mat_sample.parquet")
+data.table::fwrite(data, "~/Downloads/PosMatches_mat_sample.csv")
 
 # Optional: Clean up temporary file
 unlink(temp_file)
 
 # Message
-cat("Data has been downloaded, processed, and saved as PosMatches_mat.parquet\n")
+cat("Data has been downloaded, processed, and saved as PosMatches_mat_sample.parquet\n")