Cfrtjryk

In Awk, when I store a regular expression in a variable, can I use the variable wherever a regular expression is expected?

The AWK Programming Language by Aho says

Note that the string-matching pattern

/Asia/ 

is a shorthand for

$O ~ /Asia/

I have a text file:

$ cat f1

line 1; li

ne

2

line 3

lin

e 4

Why do the following two ways work

$ awk -v pat='in' '{if (match($0, pat)) print $0; } ' f1

line 1; li

line 3

lin

$ awk -v pat='in' ' $0 ~ pat {print $0} ' f1

line 1; li

line 3

lin

while the following doesn't

$ awk -v pat='in' ' pat {print $0} ' f1

line 1; li

ne

2

line 3

lin

e 4

?

Thanks.

Only /foo/ alone is short for $0 ~ /foo/.

In ... ~ /.../ or match(/.../, ...)..., it's only some form of quoting operator for regexps, while in other contexts, it's more an operator that resolves to a number (0 or 1).

That double meaning can be a bit confusing. There are a lot of those double meanings / ambiguities in awk.

/foo/ expands to 1 or 0 depending on whether $0 matches the foo regexp or not but "1" ~ /foo/ is not "1" ~ "1" when $0 happens to match foo, here /foo/ is no longer short for ($0 ~ /foo/). In the case of"1" ~ (/foo/)or"1" ~ +/foo/`, you'll see the behaviour varies between implementations though.

var is only var.

var as a condition means true if the variable is numeric or a numeric string and resolves to a number other than zero or if it's a string and resolves to a non-empty string.

variables declared with -v var=value are of those that may considered numeric strings if they look like numbers and strings otherwise.

awk -v var=in 'var {print "x"}'

prints x for every record because in doesn't look like a number and is not the empty string.

awk -v var=0 'var {print "x"}'

Would not print x, while:

awk 'BEGIN{var = "0"}; var {print "x"}'

would print x for every record as var was explicitly declared as a string variable. So even though it looks like a number, it's not considered as such.

That's another one of those double meanings. A variable may be considered as numerical or string depending on context. See also > that depending on context is taken as a comparison operator or a redirection operator (which again several ambiguous situations where the behaviour varies between implementations).

Note that you can also do things like:

awk '{print /foo/ + /bar/}'

Which is the same as:

awk '{print ($0 ~ /foo/) + ($0 ~ /bar/)}'

But if using concatenation instead of +

awk '{print /foo/ /bar/}'

that doesn't work as there's again an ambiguity between the /RE/ operator and the / division operator. When in doubt, use parens:

awk '{print (/foo/) (/bar/)}'

By the way, you should avoid using -v to store regexps or anything that may contain backslashes, as ANSI escape sequences are expanded in them. Instead, you should use environment variables:

RE='.txt$' awk '$0 ~ ENVIRON["RE"] {...}'

for instance.