tl;dr

Similar to how we make variables available in lateral subqueries via over ... with <var>=<expr>, it seems users might benefit from being able to do from eval (...) with <var>=<expr>.

Details

At the time this issue is being opened, super is at commit 87ab6b7.

Consider the following query, which is a simplification of something like the example query currently in the super repo README.

$ super -version
Version: v1.18.0-160-g87ab6b72

$ echo '{user:"mccanne"} {user:"philrz"}' |
   super -c 'from eval(f"https://api.github.com/users/{user}") | yield {user:user,created_at:time(created_at)}' -
{user:error("missing"),created_at:2012-08-07T18:52:36Z}
{user:error("missing"),created_at:2013-11-13T23:32:04Z}

As a naive user, I assumed that the user values upstream of my from could still be available post-from, but instead I've got only the new dataflow resulting from the from eval(...).

To get around this, I could use an approach like the one in the super repo README of using fork to split off the original dataflow and then later using join to somehow combine it back with the data that came out of the from eval(...). However, fork and join are more advanced language features, so that got me to thinking that users might appreciate a more direct way to handle this.

Following existing patterns in the language, the variables in over ... with ... came to mind, e.g., we might offer something like:

$ echo '{user:"mccanne"} {user:"philrz"}' |
   super -c 'from eval(f"https://api.github.com/users/{user}") with orig_user=user | yield {user:orig_user,created_at:time(created_at)}' -
{user:"mccanne",created_at:2012-08-07T18:52:36Z}
{user:"philrz",created_at:2013-11-13T23:32:04Z}