Explain awk code

What this awk sample shows

fields

FS = "," splits every line at its commas, so $2 is the region and $4 is the amount. awk turns $4 into a number when it adds it: 12.50 adds 12.5, and text that doesn't start with a number adds 0.

patterns and actions

Each pattern { action } runs on every line its pattern matches, in order from the top. NR == 1 { next } matches the first line, and next moves straight on to the line after, so the header never reaches the totals. $4 == "" catches a sale with no amount, and the last block has no pattern, so it runs on every line still left.

associative arrays

total[$2] += $4 uses the region as the key. The first time a region turns up its entry is empty, which counts as 0, so nothing has to be set up first. rows[$2]++ counts each region's sales the same way.

BEGIN and END

BEGIN runs before the first line is read, which is why FS is set there, and END runs after the last one. for (region in total) visits the keys in no set order, so each line goes through sort -k2 -rn, which sorts on the second column as a number, largest first.

Good to know

• Run on six sales with one missing amount, the script prints west 32.50 from 2 sales, east 28.00 from 2 and north 12.50 from 1, then 1 rows had no amount.
• close("sort -k2 -rn") waits for sort to finish. Without it, 1 rows had no amount came out above the totals in this run, because sort can't print anything until its input is closed, and awk only closes it on exit.
• Given two files, NR keeps counting across both, so only the first header is skipped: the second file's header adds a region called region with 0.00. FNR == 1 starts again for each file and skips both headers.
• The sample was run with the awk that comes with macOS, awk version 20200816, and uses only standard awk. Sorting inside awk with asort or PROCINFO["sorted_in"] is gawk-only, which is why the sample hands its lines to sort.

Watch out: awk splits at every comma, even one inside quotes. A product written "poster, large" becomes two fields, so $4 is the text large" with a space in front, which counts as 0: that sale's 30 is lost and west shows 12.50 from 2 sales, with no error. gawk 5.3 and later have a --csv option that reads quoted fields, while this macOS awk prints unknown option --csv ignored.