Depends what you mean by environment

eval in clojure doesn't have access to the local lexical environment eval was called in

So no let bound names

q symbol is in a namespace, that its used in (:use ...) , i dont use eval, load-string, the weird is that it worked in past. hm

Oh, then it is likely because the value of the *ns* (the current namespace for compilation/evaluation) is not what you expect it to be

It is best to consider *ns* to be sort of undefined, if you haven't explicitly set it to be anything, it could be anything

hm i will try once more with leinengen as i did it in past, and see if it will works, thank you for helping me.

When eval encounters a non-locally bound symbol, it looks at the namespace bound to *ns* to find it

But *ns* is generally only used like that during compilation / evaluation

When code is actually running the value could be anything and it often isn't ever even looked at

This can trip people up when moving from running code on the repl, where the repl maintains a binding of *ns*

To running code outside of the repl

i cant say that i understanded what you mean, but with leinengen without aot, same code is ok with load-string .

"code" #ordered/map (["aggregate" "1"] ["pipeline" []] ["cursor" #ordered/map nil] ["maxTimeMS" 1200000])
"load-code" #ordered/map (["aggregate" "1"] ["pipeline" []] ["cursor" #ordered/map nil] ["maxTimeMS" 1200000])

it seems that its related with aot maybe or with the https://github.com/clojurephant/clojurephant gradle tool that i used

seems that its the aot, lein with aot didn't work also.

now it works fine, q was a macro so instead of

(load-string "(q :squerylive.1 {:command true})")

(load-string "(mynamespace/q :squerylive.1 {:command true})")

anyway thank you, sometime i should understand how load-string works : )

Aot and not lein will be launching the code differently

Without aot *ns* however lein launches the code ends up being the namespace you expect

But with aot it is likely clojure.core or possibly user

whatever lein's behavior is without aot may be different again with the clojure.main launcher

So you need to set *ns* to whatever you want it to be using binding around the call to load-string

thank you : ) i will read those you said and i will try load-string again, for now problem is solved, even if i dont real understand the cause

I suppose as a simple workaround, you could check all symbols in the code that are within the current namespace & see if they are captured. Otherwise, error.

To me, it's really weird to see an affordance for refactoring being built into a language. Usually it's a part of the IDE that's familiar with the language. E.g. in IntelliJ IDEA, I can select a piece of Java code and extract it into a function - everything will be done for me, automatically. All names that come from the outside of of that piece of code will become arguments (unless they're members of the class). Cursive doesn't seem to have that capability, but just manually extracting something will let you know which symbols need to be defined. It takes a few seconds to add all those to the arg vector in a newly created function.

You might also take a look at the way the Unison programming language https://www.unison-lang.org/docs/the-big-idea/ - an even more general capability that opens up powerful tools for refactoring, manipulating, and otherwise working with code as an immutable data structure.