awk string inequality seemingly not working

Question

I have a file where I want to find all lines where column three and four differ. My file looks like:

chr1:109506687  [T/G]   BOT     TOP
chr1:109506690  [T/G]   BOT     TOP
...

The code I use to find these lines is

awk '$3 != $4 {print $0}' Cardio-Metabo_Chip_11395247_A.txt | shuf -n 10

Problem is that using this command I get results like

rs3218791       [A/C]   TOP     TOP

Where column three and four are the same.

When I use the conditional for equality, namely == I get no output, which tells me that awk never considers the two columns $3 and $4 equal, despite them often being so.

Ps. using :set list in vim, my file looks like:

chr1:109506687^I[T/G]^IBOT^ITOP$
chr1:109506690^I[T/G]^IBOT^ITOP$
....

My awk version is GNU Awk 3.1.8, but I can't imagine that having to do much with anything. This should have been right in 1.0

What might be wrong?

Answer 1

Though I can't reproduce your issue (see below), I think you're evaluating those values numerically rather than as strings (all nonempty strings —even "0"— numerically evaluate to 1). Try this:

awk '$3 != $4 "" {print $0}' test

That concatenates $4 with an empty string and should therefore force your desired string comparison.

---

I failed to reproduce your problem with mawk 1.2 and gawk 4.0.1:

$ cat test
chr1:109506687  [T/G]   BOT     TOP
chr1:109506690  [T/G]   BOT     TOP
rs3218791       [A/C]   TOP     TOP
$ mawk '$3 != $4 {print $0}' test
chr1:109506687  [T/G]   BOT     TOP
chr1:109506690  [T/G]   BOT     TOP
$ gawk '$3 != $4 {print $0}' test
chr1:109506687  [T/G]   BOT     TOP
chr1:109506690  [T/G]   BOT     TOP

The shuf pipe shouldn't have anything to do with it, nor should tabs vs spaces. (Though to be safe, I tried all combinations in my test.)

Fun tip: {print $0} is implied if there's only one clause with no action. Therefore, awk '$3 != $4' is the same as awk '$3 != $4 {print $0}' ... though be sure you're not making code harder for your peers to read.