option to create a variable with output

[ag4ve]

Bash printf has a -v option to create a variable with the output. Just to show the impact to performance this has:
swilson@localhost ~]$ time for i in {0..1000}; do unset var; var="$(echo "foo")"; done

real 0m1.212s
user 0m0.458s
sys 0m0.843s
[swilson@localhost ~]$ time for i in {0..1000}; do unset var; printf -v var "foo"; done

real 0m0.008s
user 0m0.008s
sys 0m0.000s

I'm hoping it's possible (printf is also a builtin - not sure if this allows more features than a normal command has access to) to allow a command to declare a variable and be able to execute much faster in a script vs needing to run in a subshell?

[wtlangford]

Bash's printf is a builtin, which means it has some extra powers. The only ways I know to do this from another program are to either output bash code and eval it (scary!) Or to do something like `VAR=$(jq ...)` (less scary, but only sets one variable at a time.)

…

On Sun, Dec 3, 2017, 19:48 shawn ***@***.***> wrote: Bash printf has a -v option to create a variable with the output. Just to show the impact to performance this has: ***@***.*** ~]$ time for i in {0..1000}; do unset var; var="$(echo "foo")"; done real 0m1.212s user 0m0.458s sys 0m0.843s ***@***.*** ~]$ time for i in {0..1000}; do unset var; printf -v var "foo"; done real 0m0.008s user 0m0.008s sys 0m0.000s I'm hoping it's possible (printf is also a builtin - not sure if this allows more features than a normal command has access to) to allow a command to declare a variable and be able to execute much faster in a script vs needing to run in a subshell? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1546>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADQ4V8BaX9ejP1ss9nsnjsopsTQtNvrlks5s80FLgaJpZM4Qz-lN> .

[pkoppstein]

For multiple values, consider mapfile, e.g.

$ mapfile -t aname <<< $(jq -c '.[]'  <<< '[[1,2],{"a":3}]')
$ echo ${#aname[@]}
2

(This is known to work with bash 4.2)

See also How can a stream of JSON texts produced by jq be converted into a bash array of corresponding values? at the jq FAQ

[wtlangford]

That is a really neat builtin I've never seen before!

…

On Mon, Dec 4, 2017, 09:32 pkoppstein ***@***.***> wrote: For multiple values, consider mapfile, e.g. $ mapfile -t aname <<< $(jq -c '.[]' <<< '[[1,2],{"a":3}]') $ echo ${#aname[@]} 2 See also How can a stream of JSON texts produced by jq be converted into a bash array of corresponding values? at the jq FAQ <https://github.com/stedolan/jq/wiki/FAQ> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1546 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADQ4V210vTTZBSKqTzTZZMTp57MnxHQRks5s9AJrgaJpZM4Qz-lN> .

[ag4ve]

Yeah, that's the same as a while read loop (unless you use it for preprocessing) as that seems to be a common question (when I was googling to see if my question was answered) it may be worth adding to jq's docs. However, it doesn't get around the speed hit of calling a subshell. it's also worth noting that mapfile is the same as readarray and is only in bash 4.2 (iirc - not default bash on OSX or RHEL 6.5 anyway). My use case is:

declare JQ_STR
jq_etl () {
JQ_STR="$(jq "$@" <<< "$JQ_STR")"
}

So the expectation is that that etl function is called multiple times for different functionality (I'm already calling it 3x times after a day - it'll be called much more).

[nicowilliams]

Since jq is not a shell-builtin, it can't behave like printf. The canonical way to extract data from JSON into the shell using jq is to use the @sh format and eval an assignment.

[ag4ve]

@sh ?

(y'all can tag and close this if you want - or i'll close it in a few days - thanks)