# sprintf

Unlike `printf` which sends its output to `stdout`, this function takes a variable name to store its output.

```
 1 #!/bin/sh
 2 case "$BASH_VERSION" in
 3 3.1*|4.*)
 4     sprintf() # $var_to_set $format [$arguments ...]
 5     {
 6         local __var_to_set="$1" __tmp
 7         shift 1 # var_to_set
 8         printf -v __tmp "$@"
 9         eval $__var_to_set=\"\${__tmp%\$NL}\"
10     }
11     ;;
12 *)
13     # NB: On FreeBSD, sh(1) runs this faster than bash(1) runs the above
14     sprintf() # $var_to_set $format [$arguments ...]
15     {
16         local __var_to_set="$1"
17         shift 1 # var_to_set
18         eval $__var_to_set=\$\( printf -- \"\$@\" \)
19     }
20 esac
21 sprintf "$@"
22 eval echo \"$1=[\$$1]\"
```

If we detect that we are running under `bash` version 3.1 or higher, the function is defined to use `printf -v varname ...` because it is faster (`bash` has not optimized for the case of `varname=$( printf ... )`). On FreeBSD, `/bin/sh` has been optimized for the case of `varname=$( printf ... )` and it runs faster than bash's `printf -v varname ...` syntax.

Usually `varname=$( ... )` results in a sub-shell, impacting performance. On FreeBSD you can rely on the fact that calling `printf` in a sub-shell does **not** impact performance as it does in `bash`. Shells that neither provide `printf -v varname ...` like `bash` nor optimize for `printf` in a sub-shell like FreeBSD may be slower.