awk: print sum of array values and remaining record

beca123456 · 05-06-2016, 06:43 PM

Hi,

input:

Code:

comment_1|A_(1);B_(2)|comment_2|comment_3
comment_4|A_(3)|comment_5|comment_6
comment_7|A_(4);B_(5);C_(6);C_(7)|comment_8|comment_9

output:

Code:

comment_1|3|A_(1);B_(2)|comment_2|comment_3
comment_4|3|A_(3)|comment_5|comment_6
comment_7|22|A_(4);B_(5);C_(6);C_(7)|comment_8|comment_9

Details:
The input file has a constant number of fields (NR=4).
Using awk, I am trying to place (as the second field of the output) the sum of the figure(s) between brackets in field 2 of the input.

So far, I have only been able to simplify the second field of the input and print the entire corresponding record using this code,

Code:

gawk 'BEGIN{FS=OFS="|"}{sum1=$2; print gensub(/(^[^_]+_\()|(\))|([^;]+_\()/,"","g",sum1) FS $0}'

ending up with this .temp output:

Code:

1;2|comment_1|A_(1);B_(2)|comment_2|comment_3
3|comment_4|A_(3)|comment_5|comment_6
4;5;6;7|comment_7|A_(4);B_(5);C_(6);C_(7)|comment_8|comment_9

From this output, I cannot manage to get the sum of the figures in $1. I know how to sum values from the same column but not the same record/array.

allend · 05-06-2016, 08:16 PM

You want to extract the numbers from the $2 field of your input. A regular expression to match numbers is '[:digit:]+'. This can be used in the patsplit function.

Code:

awk -F"|" '{t=0; n=patsplit($2,a,/[[:digit:]]+/); for (i in a) t+=a[i]; print n,t}' input

beca123456 · 05-07-2016, 01:47 PM

Thanks allend !
I have never heard about patsplit before, it is great !

Although, it seems to work only with gawk ( you used awk in your command).

HMW · 05-07-2016, 02:05 PM

If you don't speak awk fluently, you could use Python.

Code:

#!/usr/bin/env python3

toInsert = 0
outFile = open("beca_new.txt", "w")

# Open the file
with open("beca.txt") as inFile:
	for line in inFile: # Read line by line
		listLine = line.split('|') # Split into list using | as delim
		for char in listLine[1]: # Add up the digits in index 1
			if char.isdigit():
				toInsert += int(char)
		listLine.insert(1, toInsert) # Insert into our list at index 1
		toInsert = 0
		listLine[1] = str(listLine[1])
		print("|".join(listLine), end="") # Print or...
		outFile.write("|".join(listLine)) # write to file

outFile.close()

exit(0)

Yes, a lot more characters, I know. But it gets the job done:

Code:

./beca.py 
comment_1|3|A_(1);B_(2)|comment_2|comment_3
comment_4|3|A_(3)|comment_5|comment_6
comment_7|22|A_(4);B_(5);C_(6);C_(7)|comment_8|comment_9

I have to admit though, that the awk oneliner was impressive in its, err, incomprehensibleness (yes, that's a word - I just now invented it).

Best regards,
HMW

syg00 · 05-07-2016, 05:50 PM

Quote:

Originally Posted by beca123456

Although, it seems to work only with gawk ( you used awk in your command).

It is documented as a gawk extension - you should check the doco yourself once offered a solution.

allend · 05-07-2016, 07:56 PM

'awk' is symlinked to 'gawk' on my setup. I believe patsplit appeared circa 2010.

"incomprehensibleness" has an air of incomprehensibility about it.

syg00 · 05-07-2016, 08:14 PM

I hope this doesn't add to the incomprehensibolity, but you could replicate the patsplit with "split" using the regex allend posted earlier and use the sep array to sum.