Using awk to count number of row group - awk

Question

I have a data set: (file.txt)
X Y
1 a
2 b
3 c
10 d
11 e
12 f
15 g
20 h
25 i
30 j
35 k
40 l
41 m
42 n
43 o
46 p
I want to add two columns which are Up10 and Down10,
Up10: From (X) to (X-10) count of row.
Down10 : From (X) to (X+10)
count of row
For example:
X Y Up10 Down10
35 k 3 5
For Up10; 35-10 X=35 X=30 X=25 Total = 3 row
For Down10; 35+10 X=35 X=40 X=41 X=42 X=42 Total = 5 row
Desired Output:
X Y Up10 Down10
1 a 1 5
2 b 2 5
3 c 3 4
10 d 4 5
11 e 5 4
12 f 5 3
15 g 4 3
20 h 5 3
25 i 3 3
30 j 3 3
35 k 3 5
40 l 3 5
41 m 3 4
42 n 4 3
43 o 5 2
46 p 5 1
This is the Pierre François' solution: Thanks again #Pierre François
awk '
BEGIN{OFS="\t"; print "X\tY\tUp10\tDown10"}
(NR == FNR) && (FNR > 1){a[$1] = $1 + 0}
(NR > FNR) && (FNR > 1){
up = 0; upl = $1 - 10
down = 0; downl = $1 + 10
for (i in a) { i += 0 # tricky: convert i to integer
if ((i >= upl) && (i <= $1)) {up++}
if ((i >= $1) && (i <= downl)) {down++}
}
print $1, $2, up, down;
}
' file.txt file.txt > file-2.txt
But when i use this command for 13GB data, it takes too long.
I have used this way for 13GB data again:
awk 'BEGIN{ FS=OFS="\t" }
NR==FNR{a[NR]=$1;next} {x=y=FNR;while(--x in a&&$1-10<a[x]){} while(++y in a&&$1+10>a[y]){} print $0,FNR-x,y-FNR}
' file.txt file.txt > file-2.txt
When file-2.txt reaches 1.1GB it is frozen. I am waiting several hours, but i can not see finish of command and final output file.
Note: I am working on Gogole cloud. Machine type
e2-highmem-8 (8 vCPUs, 64 GB memory)

Answer 1

A single pass awk that keeps the sliding window of 10 last records and uses that to count the ups and downs. For symmetricy's sake there should be deletes in the END but I guess a few extra array elements in memory isn't gonna make a difference:
$ awk '
BEGIN {
FS=OFS="\t"
}
NR==1 {
print $1,$2,"Up10","Down10"
}
NR>1 {
a[NR]=$1
b[NR]=$2
for(i=NR-9;i<=NR;i++) {
if(a[i]>=a[NR]-10&&i>=2)
up[NR]++
if(a[i]<=a[NR-9]+10&&i>=2)
down[NR-9]++
}
}
NR>10 {
print a[NR-9],b[NR-9],up[NR-9],down[NR-9]
delete a[NR-9]
delete b[NR-9]
delete up[NR-9]
delete down[NR-9]
}
END {
for(nr=NR+1;nr<=NR+9;nr++) {
for(i=nr-9;i<=nr;i++)
if(a[i]<=a[nr-9]+10&&i>=2&&i<=NR)
down[nr-9]++
print a[nr-9],b[nr-9],up[nr-9],down[nr-9]
}
}' file
Output:
X Y Up10 Down10
1 a 1 5
2 b 2 5
...
35 k 3 5
...
43 o 5 2
46 p 5 1

Answer 2

Another single pass approach with a sliding window
awk '
NR == 1 { next } # skip the header
NR == 2 { min = max = cur = 1; X[cur] = $1; Y[cur] = $2; next }
{ X[++max] = $1; Y[max] = $2
if (X[cur] >= $1 - 10) next
for (; X[cur] + 10 < X[max]; ++cur) {
for (; X[min] < X[cur] - 10; ++min) {
delete X[min]
delete Y[min]
}
print X[cur], Y[cur], cur - min + 1, max - cur
}
}
END {
for (; cur <= max; ++cur) {
for (; X[min] < X[cur] - 10; ++min);
for (i = max; i > cur && X[cur] + 10 < X[i]; --i);
print X[cur], Y[cur], cur - min + 1, i - cur + 1
}
}
' file
The script assumes the X column is ordered numerically.