SQL concatenate many tsv files into single table in a database, while keeping track of file source (MonetDBLite)

Question

I am using the MonetDBLite R package to create a MonetDB. I can create database tables just fine using the instructions from here, with the following code:

library(DBI)
library(MonetDBLite)

# Write tsv file of mtcars
write.table(mtcars, "mtcars.tsv", row.names=FALSE, sep= "\t")

# Initialize MonetDB
dbdir <- "/Users/admin/my_directory"
con <- dbConnect(MonetDBLite::MonetDBLite(), dbdir)

# Write table
dbWriteTable(con, "test4", "mtcars.tsv", delim="\t")

and the following query gives

> dbGetQuery(con, "SELECT * FROM test4 LIMIT 3")
mpg cyl  disp  hp drat    wt  qsec vs am gear carb
1  21.0   6 160.0 110 3.90 2.620 16.46  0  1    4    4
2  21.0   6 160.0 110 3.90 2.875 17.02  0  1    4    4
3  22.8   4 108.0  93 3.85 2.320 18.61  1  1    4    1

So far so good. But, say I have another file mtcars2 with different mpg values:

mtcars2 <- mtcars
mtcars2$mpg <- mtcars2$mpg + 5
write.table(mtcars2, "mtcars2.tsv", row.names= FALSE, sep = "\t")

I can load it to another table:

dbWriteTable(con, "test5", "mtcars2.tsv", delim = "\t")
> dbGetQuery(con, "SELECT * FROM test5 LIMIT 3")
mpg cyl disp  hp drat    wt  qsec vs am gear carb
1 26.0   6  160 110 3.90 2.620 16.46  0  1    4    4
2 26.0   6  160 110 3.90 2.875 17.02  0  1    4    4
3 27.8   4  108  93 3.85 2.320 18.61  1  1    4    1

Also fine. But my problem is this: I want to later on look up the mpg for all cars with 6 cyl, and know which dataset it came from (mtcars or mtcars2). From what I understand of SQL indexing (which is not a lot and basically what I've read here), I should have all my data in one table to have the most efficient searches. I tried loading the first tsv file, then added another column using ALTER TABLE test4 ADD dataset TEXT and UPDATE test4 SET dataset = dataset1 sql commands-

dbSendQuery(con, "UPDATE test4 SET dataset = dataset1")
dbSendQuery(con, "UPDATE test4 SET dataset = 1")
> dbGetQuery(con, "SELECT * FROM test4 LIMIT 3")
mpg cyl disp  hp drat    wt  qsec vs am gear carb dataset
1 21.0   6  160 110 3.90 2.620 16.46  0  1    4    4       1
2 21.0   6  160 110 3.90 2.875 17.02  0  1    4    4       1
3 22.8   4  108  93 3.85 2.320 18.61  1  1    4    1       1

but then when I tried to append mtcars2 to the table, it had a different number of columns (as I should have expected, duh). What's the best way to concatenate data from many tsv files with identical columns into a single table, while keeping track of the data's source?

EDIT- as you might have guessed, the real data is not mtcars- it's flat tsv files millions of lines long, meaning I want to avoid reading the whole file into memory and manipulating with R.

Answer 1

Here is what I would do, given a set of files with the same structure and file names desired in the final table, which is otherwise a combination of the data from all files:

# say we have those files
write.table(mtcars, "mtcars1.tsv", row.names=FALSE, sep= "\t")
write.table(mtcars, "mtcars2.tsv", row.names=FALSE, sep= "\t")

# write them individually, and add a column that contains the file name
dbWriteTable(con, "mtcars1", "mtcars1.tsv", delim="\t")
dbSendQuery(con, "ALTER TABLE mtcars1 ADD COLUMN file STRING DEFAULT 'mtcars1.tsv';")
dbWriteTable(con, "mtcars2", "mtcars2.tsv", delim="\t")
dbSendQuery(con, "ALTER TABLE mtcars2 ADD COLUMN file STRING DEFAULT 'mtcars2.tsv';")

# now combine into a new table
dbSendQuery(con, "CREATE TABLE mtcars_mat AS SELECT * FROM mtcars1 UNION ALL SELECT * FROM mtcars2")

# or a view if you don't need to modify the data in the mtcars table (faster)
dbSendQuery(con, "CREATE view mtcars AS SELECT * FROM mtcars1 UNION ALL SELECT * FROM mtcars2")



# and here is the same as a loop with a filename glob and some added robustness (handy if you have 1000 files)
files <- Sys.glob("/some/path/mtcars*.tsv")
tables <- dbQuoteIdentifier(con, tools::file_path_sans_ext(basename(files)))
dbBegin(con)
for (i in 1:length(files)) {
  dbWriteTable(con, tables[i], files[i], delim="\t", transaction=FALSE)
  dbSendQuery(con, paste0("ALTER TABLE ", tables[i], " ADD COLUMN file STRING DEFAULT ",dbQuoteString(con, files[i]),";"))
}
dbSendQuery(con, paste0("CREATE TABLE somefinalresult AS ", paste0("SELECT * FROM ",tables, collapse=" UNION ALL ")))
# remove the parts again, optional
dbSendQuery(con, paste0("DROP TABLE ", tables, ";", collapse=" "))
dbCommit(con)

Answer 2

Following xQbert's suggestion, I solved using SQL commands only (necessary and faster than bash commands, considering my data is 10s of files, each millions of lines long).

library(DBI)
library(MonetDBLite)

# Write tsv file of mtcars
write.table(mtcars, "mtcars.tsv", row.names=FALSE, sep= "\t")

# Write tsv of second mtcars
mtcars2 <- mtcars
mtcars2$mpg <- mtcars2$mpg + 5
write.table(mtcars2, "mtcars2.tsv", row.names= FALSE, sep = "\t")

# Initialize MonetDB
dbdir <- "/Users/admin/"
con <- dbConnect(MonetDBLite::MonetDBLite(), dbdir)

# Write table
dbWriteTable(con, "test4", "mtcars.tsv", delim="\t")

# Add data source information
dbSendQuery(con, "ALTER TABLE test4 ADD source TEXT")
dbSendQuery(con, "UPDATE test4 SET source = 'dataset1'")

# Write second dataset to a temporary table
dbWriteTable(con, "temptable", "mtcars2.tsv", delim="\t")

# Add data source information
dbSendQuery(con, "ALTER TABLE temptable ADD source TEXT")
dbSendQuery(con, "UPDATE temptable SET source = 'dataset2'")

# Insert temp table into main table
dbSendQuery(con, "INSERT INTO test4 SELECT * FROM temptable")

# Drop temp table
dbSendQuery(con, "DROP TABLE temptable")

# Checking the data, truncated for clarity
> dbGetQuery(con, "SELECT * FROM test4")
mpg cyl  disp  hp drat    wt  qsec vs am gear carb   source
1  21.0   6 160.0 110 3.90 2.620 16.46  0  1    4    4 dataset1
2  21.0   6 160.0 110 3.90 2.875 17.02  0  1    4    4 dataset1
3  22.8   4 108.0  93 3.85 2.320 18.61  1  1    4    1 dataset1
...
33 26.0   6 160.0 110 3.90 2.620 16.46  0  1    4    4 dataset2
34 26.0   6 160.0 110 3.90 2.875 17.02  0  1    4    4 dataset2
35 27.8   4 108.0  93 3.85 2.320 18.61  1  1    4    1 dataset2
...
64 26.4   4 121.0 109 4.11 2.780 18.60  1  1    4    2 dataset2

Sorry if I didn't make it clear enough in the question that my data is much larger than mtcars- if you have medium sized data, the data.tables package is probably a better solution than a database.

Answer 3

You should be able to do what you want with executing dbWriteTable() after reading the file create a new variable in the data.frame. Something like:

  library(DBI)
  library(MonetDBLite)
  library(data.table)

  # Write tsv file of mtcars
  tmp <- tempfile()
  write.table(mtcars, tmp, row.names=FALSE, sep= "\t")

  # Initialize MonetDB
  dbdir <- "~/Desktop/temp"
  con <- dbConnect(MonetDBLite::MonetDBLite(), dbdir)

  test4df <- fread(tmp)
  test4df$dataset <- 1
  dbWriteTable(con, "test4", test4df)

  dbReadTable(con, "test4")

  test5df <- fread(tmp)
  test5df$mpg <- test5df$mpg + 5
  test5df$dataset <- 2
  dbWriteTable(con, "test4", test5df, append = TRUE)

  dbReadTable(con, "test4")

Edit (Suggestion on the way without opening the file)

If you want to the work without opening a file in one time, you can do something like this to modify the file and attach another field. As I wrote this will work with an OS with bash.

infile <- tmp
outfile <- tempfile()

# open connections
incon <- file(description = infile, open = "r")
outcon <- file(description = outfile, open = "w")

# count the number of lines (this will work only with Mac/Linux)
com <- paste("wc -l ", infile, " | awk '{ print $1 }'", sep="")
n <- system(command=com, intern=TRUE)

# work with the first line
txt <- scan(file = incon, what = character(), nlines=1, quiet=TRUE)
txt <- c(txt, "dataset")
cat(paste(txt, collapse = "\t"), "\n", file = outcon, sep = "")

# work with the rest of the file
for(i in 2:n) {
  txt <- scan(file = incon, what = character(), nlines=1, quiet=TRUE)
  txt <- c(txt, "1")
  cat(paste(txt, collapse = "\t"), "\n", file = outcon, sep = "")
}
close(incon);close(outcon)
dbWriteTable(con, "test4", outfile, delim = "\t")
# do the similar for other files