# EOF within quoted string

So I was trying to parse this gff file from MGI, with mouse gene annotations. And, well, I’m an idiot. But in a way that is potentially instructive.

The documentation for the file is a docx file (not really a recommended format for such metadata), but it seems rather simple, really: tab delimited, with 9 columns, the ninth column being a bunch of pasted attributes that needs to be further parsed, but we’ll skip over that detail.

I’d want to use fread() from the data.table package, but it turns out that the file has a bunch of lines with “###” interspersed within the data, and I couldn’t see a way to skip over those in fread(), so I fell back to the usual base R function, read.table().

Let’s first download and unzip the file.

# download the file
file <- "MGI.20170803.gff3.gz"
url <- paste0(site, "/", file)

# unzip to a temporary file
file <- sub(".gz\$", "", file)
tmpfile <- tempfile()
remove_tmpfile <- FALSE
if(!file.exists(file)) { # need to unzip
system(paste0("gunzip -c ", file, ".gz > ", tmpfile))
remove_tmpfile <- TRUE
file <- tmpfile
}

Okay, now to read it into R with read.table().

tab <- read.table(file, sep="\t", header=FALSE, comment.char="#",
na.strings=".", stringsAsFactors=FALSE)

This gives a warning message:

Warning message:
In scan(file = file, what = what, sep = sep, quote = quote, dec = dec,  :
EOF within quoted string

Hmm. What does that mean? Oh, no matter, let’s move on…

Wait, there are no genes on chromosomes 5, 8, 15, 18, Y, or MT. How could that be? Something must be wrong with the file. Let’s look at another file at that site, MGI.20160103.gff3.gz. That one’s missing chromosomes 8 and 13.

So I ask Dan Gatti: “Hey, those files are corrupted. Who should I talk to about them?”

And he’s like, “That’d be a disaster, but they look fine to me [parsed with read.delim()].”

So I tried using read.delim() and sure enough, no warning, genes on all chromosomes, and about twice as many records. Oops.

## read.delim() vs read.table()

So what’s the difference between read.delim() and read.table()? Well, read.delim() calls read.table() with a particular set of default values for the arguments:

> read.delim
function (file, header = TRUE, sep = "\\t", quote = "\\"", dec = ".",
fill = TRUE, comment.char = "", ...)
read.table(file = file, header = header, sep = sep, quote = quote,
dec = dec, fill = fill, comment.char = comment.char, ...)

The key argument here is quote, in that read.table() uses quote="'\"" (that is, looking for either single- or double-quotes) while read.delim() uses quote="\"" (just looking for double-quotes).

There are no double-quotes in the file, but that ninth column includes some single-quotes, and so my use of read.table() was mucking everything up. And presumably there was an odd number of them, so the end-of-file (EOF) character was inside one of those quoted strings.

To read the file properly, I should have used quote="\"". Even better, I could use quote="", and for that matter also fill=FALSE (since every line is supposed to contain all nine columns).

tab <- read.table(file, sep="\t", header=FALSE, comment.char="#",
na.strings=".", stringsAsFactors=FALSE,
quote="", fill=FALSE)

## Lessons

There are several lessons here.

• I shouldn’t have ignored the “EOF within quoted string” warning.

• I should have compared the number of lines I read in with the number of lines in the input file. If I’d done so, I’d have seen that I had just about half as many lines as I should’ve, and so I’d clearly messed something up.

• When I run into a problem like this, it’s more likely that there’s a problem with my code than that there’s a problem with the file.

• With a file of this sort, I should have used quote="" and fill=FALSE in my call to read.table(). I’m not expecting any quoted fields, and I’m expecting that every line will have exactly nine columns.

• It’s good to have a friend like Dan.