match Expression and switch Statement

match expression

match subject { pattern => value, ... } returns the value of the first arm whose pattern matches, like Rust's match. Because it is an expression, it fits naturally into assignments, print and f-strings.

zawk '{ print match $2 { "info" => 1, "error" | "fatal" => 3, 3 => 3, /^W/ => 2, _ => 0 } }'

{
    level = match $3 {
        /^E/ => "error",
        "warn" | "W" => "warn",
        /(\d+)ms/ as m => m[1] > 1000 ? "slow" : "ok",
        _ => "info"
    }
    count[level]++
}

Rules:

• A pattern is a string, number or regex literal, or several joined with |. The first matching arm wins.
• Numbers compare numerically with a numeric-looking subject (a field, getline input...), as in gawk's switch: 3 => ... matches 3.0.
• /re/ as m => value binds the match to the array m: m[0] is the whole match, m[1] the first group. m is visible only in value.
• _ matches anything and must be the last arm. Without _, the value is "" when nothing matches.
• The subject is evaluated once. Arms are separated by commas or newlines.
• match(s, re) is still the builtin function; match (x) { starts a match expression.

Examples

HTTP status classes:

zawk '{ c[match $9 { /^2/ => "2xx", /^3/ => "3xx", /^4/ => "4xx", /^5/ => "5xx", _ => "other" }]++ }
      END { for (k, v in c) print k, v }' access.log

Extract a number with a capture group:

$ echo "GET /api took 1532ms" | zawk '{ print match $0 { /(\d+)ms/ as m => m[1] / 1000 " s", _ => "-" } }'
1.532 s

switch statement

zawk supports gawk's switch statement:

switch ($1) {
case 1:
case "one":
    print "one"
    break
case /^[0-9]+$/:
    print "a number"
    break
default:
    print "something else"
}

• The value is evaluated once and compared with each case label in order. A label is a numeric constant (1, -2.5, 0x10), a string constant or a regex constant.
• Numeric and string labels are compared with ==, regex labels with ~.
• Control jumps to the first matching case, or to default, and then falls through until a break. continue inside a switch continues the enclosing loop.
• Duplicate labels (1, 1.0 and "1") and a second default are syntax errors, as in gawk.

Which one to use?

	match	switch
Kind	expression (returns a value)	statement
Fall-through	never	until break
Regex groups	/re/ as m	no
Several patterns per arm	"a" | "b"	stacked case labels
Portable to gawk	no	yes

Use match to compute a value, and switch when each branch runs several statements or when the script must also run with gawk.