Merge pull request #26 from pranavanba/use-internal-archive-for-data-…

…sync RMHDR-252 Update external parquet pipeline to use the internal parquet archive
Sage-Bionetworks · Apr 17, 2024 · 4f0e58d · 4f0e58d
2 parents 1574e66 + f461d3c
commit 4f0e58d
Show file tree

Hide file tree

Showing 7 changed files with 42 additions and 22 deletions.
diff --git a/config.yml b/config.yml
@@ -25,6 +25,7 @@ prod:
   STAGING_TO_ARCHIVE_DOWNLOAD_LOCATION: ./temp_staging_to_archive/
   ARCHIVE_TO_CURRENT_DOWNLOAD_LOCATION: ./temp_archive_to_current/
   POST_WITHDRAW_LOCATION: ./temp_post_withdraw
+  DICTIONARIES_FOLDER: syn52316269
 
 staging:
   inherits: prod

diff --git a/deidentification.R → scripts/deidentification/deidentification.R b/deidentification.R → scripts/deidentification/deidentification.R
@@ -10,6 +10,11 @@ unlink('./dictionaries/', recursive = T, force = T)
 # Get dictionaries --------------------------------------------------------
 system('synapse get -r syn52316269 --downloadLocation ./dictionaries/ --manifest suppress')
 
+list.files("./dictionaries", full.names = T) %>% lapply(function(x) {
+  y <- x %>% stringr::str_remove_all("[0-9]")
+  file.rename(from = x, to = y)
+})
+
 junk <- lapply(list.files("./dictionaries/", full.names = T), function(f) {
   lines <- readLines(f)
 
@@ -103,7 +108,8 @@ for (i in seq_along(deidentified_results$deidentified_datasets)) {
 
   arrow::write_dataset(dataset = deidentified_results$deidentified_datasets[[i]], 
                        path = file.path(PARQUET_FINAL_LOCATION, names(deidentified_results$deidentified_datasets)[[i]]), 
-                       max_rows_per_file = 1000000,
+                       max_open_files = 2048,
+                       max_rows_per_file = 5000000,
                        partitioning = c('cohort'), 
                        existing_data_behavior = 'delete_matching',
                        basename_template = paste0("part-0000{i}.", as.character("parquet")))
@@ -122,15 +128,15 @@ for (i in seq_along(deidentified_results$values_to_review)) {
 
 # Index each file in Synapse
 latest_commit <- gh::gh("/repos/:owner/:repo/commits/main", owner = "Sage-Bionetworks", repo = "recover-parquet-external")
-latest_commit_file_url <- latest_commit$files[[1]]$blob_url
+latest_commit_file_url <- paste0(latest_commit$html_url %>% stringr::str_replace("commit", "blob"), "/deidentification.R")
 
 for (i in seq_along(list.files('./dictionaries/new_to_review/'))) {
-  synStore(File(path = list.files('./dictionaries/new_to_review', full.names = T)[i], 
+  synStore(File(path = list.files('./dictionaries/new_to_review', full.names = T)[i],
                 parent = DEID_VALS_TO_REVIEW),
           activityName = "Indexing",
           activityDescription = "Indexing files containing new PII values to review for deidentification step",
-          used = c((synGetChildren('syn52316269') %>% as.list())[[i]]$id, 
-                   synFindEntityId(names(deidentified_results$deidentified_datasets)[i], 
+          used = c((synGetChildren(DICTIONARIES_FOLDER) %>% as.list())[[i]]$id,
+                   synFindEntityId(names(deidentified_results$deidentified_datasets)[i],
                                    parent = PARQUET_FOLDER_INTERNAL)),
           executed = latest_commit_file_url
           )

diff --git a/filtering.R → scripts/filtering/filtering.R b/filtering.R → scripts/filtering/filtering.R
@@ -22,7 +22,8 @@ dob2age <- function(dataset, column, input = AWS_PARQUET_DOWNLOAD_LOCATION, part
       dplyr::collect() %>% 
       dplyr::mutate(age = ifelse(age>89, "90+", age)) %>% 
       arrow::write_dataset(path = input_path, 
-                           max_rows_per_file = 100000, 
+                           max_open_files = 2048,
+                           max_rows_per_file = 1000000, 
                            partitioning = partitions, 
                            existing_data_behavior = 'delete_matching')
   }
@@ -51,7 +52,8 @@ drop_cols_datasets <- function(dataset, columns=c(), input = AWS_PARQUET_DOWNLOA
     arrow::open_dataset(sources = input_path) %>% 
       dplyr::select(!dplyr::any_of(columns)) %>% 
       arrow::write_dataset(path = final_path, 
-                           max_rows_per_file = 1000000,
+                           max_open_files = 2048,
+                           max_rows_per_file = 5000000,
                            partitioning = partitions, 
                            existing_data_behavior = 'delete_matching',
                            basename_template = paste0("part-0000{i}.", as.character("parquet")))
@@ -78,7 +80,10 @@ cols_to_drop <- lapply(datasets_to_filter, function(x) {
 
 tmp <- 
   lapply(seq_along(datasets_to_filter), function(i) {
-    cat(i, "Dropping", cols_to_drop[[i]], "from", datasets_to_filter[[i]], "\n\n")
+    cat(i)
+    cat(": Dropping from [", datasets_to_filter[[i]], "]", "\n")
+    cat(cols_to_drop[[i]], sep = ", ")
+    cat("\n\n")
     drop_cols_datasets(dataset = datasets_to_filter[[i]], 
                        columns = cols_to_drop[[i]], 
                        input = AWS_PARQUET_DOWNLOAD_LOCATION, 

diff --git a/archive-to-current.R → scripts/main/archive-to-current.R b/archive-to-current.R → scripts/main/archive-to-current.R
diff --git a/staging_to_archive.R → scripts/main/staging_to_archive.R b/staging_to_archive.R → scripts/main/staging_to_archive.R
diff --git a/sts_synindex_external.R → scripts/main/sts_synindex_external.R b/sts_synindex_external.R → scripts/main/sts_synindex_external.R
@@ -1,8 +1,6 @@
 library(synapser)
 library(arrow)
 library(dplyr)
-library(synapserutils)
-library(rjson)
 
 
 # Functions ---------------------------------------------------------------
@@ -115,19 +113,28 @@ Sys.setenv('AWS_ACCESS_KEY_ID'=token$accessKeyId,
            'AWS_SECRET_ACCESS_KEY'=token$secretAccessKey,
            'AWS_SESSION_TOKEN'=token$sessionToken)
 
+s3 <- arrow::S3FileSystem$create(access_key = token$accessKeyId,
+                                 secret_key = token$secretAccessKey,
+                                 session_token = token$sessionToken,
+                                 region="us-east-1")
+
+archive_uri <- paste0(token$bucket, '/', token$baseKey, '/archive/')
+archive_list <- s3$GetFileInfo(arrow::FileSelector$create(archive_uri, recursive=F))
+latest_archive <- archive_list[[length(archive_list)]]$path
+
 
 # Sync bucket to local dir ------------------------------------------------
 unlink(AWS_PARQUET_DOWNLOAD_LOCATION, recursive = T, force = T)
-sync_cmd <- glue::glue('aws s3 sync {base_s3_uri} {AWS_PARQUET_DOWNLOAD_LOCATION} --exclude "*owner.txt*" --exclude "*archive*"')
+sync_cmd <- glue::glue('aws s3 sync s3://{latest_archive}/ {AWS_PARQUET_DOWNLOAD_LOCATION} --exclude "*owner.txt*" --exclude "*archive*"')
 system(sync_cmd)
 
 
 # Remove withdrawn participants -------------------------------------------
-source("~/recover-parquet-external/remove_withdrawn_participants.R")
+source("~/recover-parquet-external/scripts/withdrawal/remove_withdrawn_participants.R")
 
 
 # Filter parquet datasets -------------------------------------------------
-source('~/recover-parquet-external/filtering.R')
+source('~/recover-parquet-external/scripts/filtering/filtering.R')
 
 # Copy unfiltered parquet datasets to new location with filtered parquet datasets
 unlink(PARQUET_FINAL_LOCATION, recursive = T, force = T)
@@ -141,11 +148,11 @@ unlink(PARQUET_FILTERED_LOCATION, recursive = T, force = T)
 
 
 # De-identify parquet datasets --------------------------------------------
-source('~/recover-parquet-external/deidentification.R')
+source('~/recover-parquet-external/scripts/deidentification/deidentification.R')
 
 
 # Sync final parquets to bucket -------------------------------------------
-date <- lubridate::today()
+date <- latest_archive %>% stringr::str_extract("[0-9]{4}_[0-9]{2}_[0-9]{2}")
 sync_cmd <- glue::glue('aws s3 --profile service-catalog sync {PARQUET_FINAL_LOCATION} {base_s3_uri_archive}{date}/ --exclude "*owner.txt*" --exclude "*archive*"')
 system(sync_cmd)
 
@@ -180,7 +187,6 @@ synapse_manifest <-
   dplyr::mutate(file_key = gsub("cohort_", "cohort=", file_key),
                 s3_file_key = gsub("cohort_", "cohort=", s3_file_key))
 
-
 # List all files currently indexed in Synapse
 synapse_fileview <- 
   synapser::synTableQuery(paste0('SELECT * FROM ', SYNAPSE_FILEVIEW_ID))$filepath %>%
@@ -204,7 +210,12 @@ if (nrow(synapse_fileview)>0) {
 
 # Index each file in Synapse
 latest_commit <- gh::gh("/repos/:owner/:repo/commits/main", owner = "Sage-Bionetworks", repo = "recover-parquet-external")
-latest_commit_tree_url <- latest_commit$html_url %>% stringr::str_replace("commit", "tree")
+latest_commit_this_file <- paste0(latest_commit$html_url %>% stringr::str_replace("commit", "blob"), "/sts_synindex_external.R")
+
+act <- synapser::Activity(name = "Indexing",
+                          description = "Indexing external parquet datasets",
+                          used = paste0("s3://", latest_archive), 
+                          executed = latest_commit_this_file)
 
 if(nrow(synapse_manifest_to_upload) > 0){
   for(file_number in seq_len(nrow(synapse_manifest_to_upload))){
@@ -225,11 +236,7 @@ if(nrow(synapse_manifest_to_upload) > 0){
               parentId = tmp$parent,
               name = new_fileName)
 
-    f <- synStore(f, 
-                  activityName = "Indexing", 
-                  activityDescription = "Indexing external parquet datasets",
-                  used = PARQUET_FOLDER_INTERNAL, 
-                  executed = latest_commit_tree_url)
+    f <- synStore(f,  activity = act)
 
   }
 }
diff --git a/remove_withdrawn_participants.R → ...ithdrawal/remove_withdrawn_participants.R b/remove_withdrawn_participants.R → ...ithdrawal/remove_withdrawn_participants.R
@@ -89,6 +89,7 @@ if (length(participants_to_withdraw) > 0) {
     d %>% 
       arrow::write_dataset(
         path = file.path(POST_WITHDRAW_LOCATION, basename(x)),
+        max_open_files = 2048,
         max_rows_per_file = 5000000,
         partitioning = "cohort",
         existing_data_behavior = 'delete_matching',