New AWK Syntax

zawk keeps the AWK language: patterns and actions, BEGIN/END, fields, associative arrays, printf, getline, user-defined functions and gawk's switch. On top of it, zawk adds a small set of modern constructs that make everyday scripts shorter and safer. Every addition is designed so that existing AWK programs keep their meaning (the few exceptions are listed in Differences from gawk).

At a glance

Syntax	Example	Like
f-string	f"user {$1} at line {NR:05d}"	Python f-strings
Raw string	r"C:\Program Files", $0 ~ r"\d+\.\d+"	Rust raw strings
Default value ??	"Hello " name ?? "Guest"	C#/JS ??
Default assignment ??=	OFS ??= ","	JS ??=
try ... else	body = try read_all(url) else ""	Kotlin/Rust ?
!in	$1 !in seen { seen[$1]; print }	Kotlin !in
match expression	match $2 { "error" | "fatal" => 3, /^W/ => 2, _ => 0 }	Rust match
switch statement	switch ($1) { case /^a/: ...; break }	gawk
local variables	{ local total = 0; ... }	JS let
Key-value loop	for (k, v in count) print k, v	Python items()
Range loop	for (i in 1..=10)	Rust ranges
Negative field index	$(-1), $(-2)	Python xs[-1]
assert	assert(NF == 5, f"line {NR}: bad record")	many languages
Field list shorthand	zawk '$1,$3' file	cut
Structured I/O	zawk -i csv -H '{ print $FI["name"] }'	xsv, DuckDB

A taste

#!/usr/bin/env -S zawk -E
# @desc Summarize API latency per endpoint from a JSON Lines log

{
    local ms = $FI["latency_ms"] ?? 0
    level = match $FI["status"] {
        /^2/ => "ok",
        /^4/ => "client",
        /^5/ => "server",
        _ => "unknown"
    }
    lat[$FI["path"]] += ms
    hits[$FI["path"]]++
    errors[level]++
}

END {
    for (path, total in lat)
        print f"{path:<24} {hits[path]:>6} {total / hits[path]:>8.1f} ms"
    for (k, v in errors) print k, v > "/dev/stderr"
}

zawk -i jsonl -f latency.awk api.jsonl

Keywords

local, try and match ... { are new keywords: they can no longer be used as variable or function names. f"..." and r"..." are always string literals, so write f "x" (with a space) to concatenate a variable named f with a string. zawk lint reports these cases (W103).