properly join two files based on 2 columns in common
I have two files I'm trying to join/merge based on columns 1 and 2. They look something like this, with file1 (58210 lines) being much shorter than file2 (815530 lines) and I'd like to find the intersection of these two files based on fields 1 and 2 as an index:
file1:
2L 25753 33158
2L 28813 33158
2L 31003 33158
2L 31077 33161
2L 31279 33161
3L 32124 45339
3L 33256 45339
...
file2:
2L 20242 0.5 0.307692307692308
2L 22141 0.32258064516129 0.692307692307692
2L 24439 0.413793103448276 0.625
2L 24710 0.371428571428571 0.631578947368421
2L 25753 0.967741935483871 0.869565217391304
2L 28813 0.181818181818182 0.692307692307692
2L 31003 0.36 0.666666666666667
2L 31077 0.611111111111111 0.931034482758621
2L 31279 0.75 1
3L 32124 0.558823529411765 0.857142857142857
3L 33256 0.769230769230769 0.90625
...
I've been using the following couple of commands but end up with different numbers of lines:
awk 'FNR==NR{a[$1$2]=$3;next} {if($1$2 in a) print}' file1 file2 | wc -l
awk 'FNR==NR{a[$1$2]=$3;next} {if($1$2 in a) print}' file2 file1 | wc -l
I'm not sure why this happens, and I've tried sorting prior to comparison, just in case I have duplicate lines (based on columns 1 and 2) in either of the files, but it doesn't seem to help. (Any insights on why this is so are also appreciated)
How can I just merge the files so that just the lines of file2 that have the corresponding columns 1 and 2 in file1 get printed, with column 3 of file1 added on, to look something like this:
2L 25753 0.967741935483871 0.869565217391304 33158
2L 28813 0.181818181818182 0.692307692307692 33158
2L 31003 0.36 0.666666666666667 33158
2L 31077 0.611111111111111 0.931034482758621 33161
2L 31279 0.75 1 33161
3L 32124 0.558823529411765 0.857142857142857 45339
3L 33256 0.769230769230769 0.90625 45339
awk 'NR==FNR{a[$1,$2]=$3;next} ($1,$2) in a{print $0, a[$1,$2]}' file1 file2
Look:
$ cat file1
2L 5753 33158
2L 8813 33158
2L 7885 33159
2L 1279 33159
2L 5095 33158
$
$ cat file2
2L 8813 0.6 1.2
2L 5762 0.4 0.5
2L 1279 0.5 0.9
$
$ awk 'NR==FNR{a[$1,$2]=$3;next} ($1,$2) in a{print $0, a[$1,$2]}' file1 file2
2L 8813 0.6 1.2 33158
2L 1279 0.5 0.9 33159
$
If that's not what you want, please clarify and perhaps post some more representative sample input/output.
Commented version of the above code to provide requested explanation:
awk ' # START SCRIPT
# IF the number of records read so far across all files is equal
# to the number of records read so far in the current file, a
# condition which can only be true for the first file read, THEN
NR==FNR {
# populate array "a" such that the value indexed by the first
# 2 fields from this record in file1 is the value of the third
# field from the first file.
a[$1,$2]=$3
# Move on to the next record so we don't do any processing intended
# for records from the second file. This is like an "else" for the
# NR==FNR condition.
next
} # END THEN
# We only reach this part of the code if the above condition is false,
# i.e. if the current record is from file2, not from file1.
# IF the array index constructed from the first 2 fields of the current
# record exist in array a, as would occur if these same values existed
# in file1, THEN
($1,$2) in a {
# print the current record from file2 followed by the value from file1
# that occurred at field 3 of the record that had the same values for
# field 1 and field 2 in file1 as the current record from file2.
print $0, a[$1,$2]
} # END THEN
' file1 file2 # END SCRIPT
Hope that helps.
If you like to join the files line by line then use this command:
join -o 1.2,1.3,2.4,2.5,1.4 <(cat -n file1) <(cat -n file2)
As you updated the question:
join -o 1.1,2.2,2.3,1.2 <(sed 's/[[:space:]]\+/@/' file1|sort) \
<(sed 's/[[:space:]]\+/@/' file2|sort)|sed 's/@/\t/'
First replace the first delimiter in each row with some non-space character and sort both input files. Then use join to make the actual join. Filter out its output to replace the non-space char with space.
This is the output from the files as in question:
xyz]$ join -o 1.1,2.2,2.3,1.2 <(sed 's/[[:space:]]\+/@/' file1|sort) \
<(sed 's/[[:space:]]\+/@/' file2|sort)|sed 's/@/\t/'
2L 25753 0.967741935483871 0.869565217391304 33158
2L 28813 0.181818181818182 0.692307692307692 33158
2L 31003 0.36 0.666666666666667 33158
2L 31077 0.611111111111111 0.931034482758621 33161
2L 31279 0.75 1 33161
3L 32124 0.558823529411765 0.857142857142857 45339
3L 33256 0.769230769230769 0.90625 45339
You can use the join command but you need to create a single join field in each data table. Assuming that you do have values other that 2L in column 1, then this code should work regardless of the sorted or unsorted nature of the two input files:
tmp=${TMPDIR:-/tmp}/tmp.$$
trap "rm -f $tmp.?; exit 1" 0 1 2 3 13 15
awk '{print $1 ":" $2, $0}' file1 | sort > $tmp.1
awk '{print $1 ":" $2, $0}' file2 | sort > $tmp.2
join -o 2.2,2.3,2.4,2.5,1.4 $tmp.1 $tmp.2
rm -f $tmp.?
trap 0
If you have bash and 'process substitution', or if you know that the data is already sorted appropriately, you can simplify the processing.
I'm not entirely sure why your code wasn't working, but I'd probably be using a[$1,$2] for the subscripts; it will give you less trouble if some of your column 1 values are pure numeric and can therefore be confused when you concatenate columns 1 and 2. That's why the 'key creation' awk scripts used a colon between the fields.
With revised data files as shown:
file1
2L 5753 33158
2L 8813 33158
2L 7885 33158
2L 7885 33159
2L 1279 33158
2L 5095 33158
2L 3256 33158
2L 5372 33158
2L 7088 33161
2L 5762 33161
file2
2L 5095 0.666666666666667 1
2L 5372 0.5 0.925925925925926
2L 5762 0.434782608695652 0.580645161290323
2L 5904 0.571428571428571 0.869565217391304
2L 5974 0.434782608695652 0.694444444444444
2L 6353 0.785714285714286 0.84
2L 7088 0.590909090909091 0.733333333333333
2L 7885 0.714285714285714 0.864864864864865
2L 7902 0.642857142857143 0.810810810810811
2L 8263 0.833333333333333 0.787878787878788
(Unchanged from the question.)
Output
2L 5095 0.666666666666667 1 33158
2L 5372 0.5 0.925925925925926 33158
2L 5762 0.434782608695652 0.580645161290323 33161
2L 7088 0.590909090909091 0.733333333333333 33161
2L 7885 0.714285714285714 0.864864864864865 33158
2L 7885 0.714285714285714 0.864864864864865 33159