Refactor env_vars_to_delete

crates/nu-cli/src/env/environment.rs

@@ -40,8 +40,12 @@ impl Env for Box<dyn Env> {
 #[derive(Debug, Default)]
 struct DirectorySpecificEnvironment {
     pub whitelisted_directories: Vec<PathBuf>,
-    pub added_env_vars: HashMap<PathBuf, Vec<String>>, //Directory -> Env key. If an environment var has been added from a .nu in a directory, we track it here so we can remove it when the user leaves the directory.
+
-    pub overwritten_env_values: HashMap<PathBuf, Vec<(String, String)>>, //Directory -> (env_key, value). If a .nu overwrites some existing environment variables, they are added here so that they can be restored later.
+    //Directory -> Env key. If an environment var has been added from a .nu in a directory, we track it here so we can remove it when the user leaves the directory.
+    pub added_env_vars: HashMap<PathBuf, Vec<String>>,
+
+    //Directory -> (env_key, value). If a .nu overwrites some existing environment variables, they are added here so that they can be restored later.
+    pub overwritten_env_values: HashMap<PathBuf, Vec<(String, String)>>,
 }
 
 impl DirectorySpecificEnvironment {
@@ -88,31 +92,27 @@ impl DirectorySpecificEnvironment {
         Ok(vars_to_add)
     }
 
-    //If the user has left directories which added env vars through .nurc, we clear those vars
+    //If the user has left directories which added env vars through .nu, we clear those vars
-    //For each directory d in nurc_env_vars:
-    //if current_dir does not have d as a parent (possibly recursive), the vars set by d should be removed
     pub fn env_vars_to_delete(&mut self) -> std::io::Result<Vec<String>> {
         let current_dir = std::env::current_dir()?;
 
-        let mut new_nurc_env_vars = HashMap::new();
+        let vars_to_delete = self.added_env_vars.iter().fold(
-        for (d, v) in self.added_env_vars.iter() {
+            Vec::new(),
-            let mut working_dir = Some(current_dir.as_path());
+            |mut vars_to_delete, (directory, env_vars)| {
-            while working_dir.is_some() {
+                let mut working_dir = Some(current_dir.as_path());
-                if working_dir.unwrap() == d {
-                    new_nurc_env_vars.insert(d.clone(), v.clone());
-                    break;
-                } else {
-                    working_dir = working_dir.unwrap().parent();
-                }
-            }
-        }
 
-        let mut vars_to_delete = vec![];
+                while let Some(wdir) = working_dir {
-        for (path, vals) in self.added_env_vars.iter() {
+                    if &wdir == directory {
-            if !new_nurc_env_vars.contains_key(path) {
+                        return vars_to_delete;
-                vars_to_delete.extend(vals.clone());
+                    } else {
-            }
+                        working_dir = working_dir.unwrap().parent();
-        }
+                    }
+                }
+                //only delete vars from directories we are not in
+                vars_to_delete.extend(env_vars.clone());
+                vars_to_delete
+            },
+        );
 
         Ok(vars_to_delete)
     }

post.org

@@ -12,9 +12,11 @@ In order for a .nu-file to be read, the directory it is in must be listed in the
 ```
 nu_env_dirs = ["/home/sam", "/home/sam/github", "/home/sam/github/test"]
 ```
-
+This was implemented for the sake of security. I do not believe that it is appropiate for nushell to pick up random .nu-files unless 
-#  The way it works now is that whenever you run the function `maintain_nurc_environment_vars`, the current directory is checked for a `.nu` file and if it exists the variables in it are added. This works.
+the user has explicitly said that it's alright. An adversary could hide a .nu-file in an otherwise unsuspicous folder that you download, 
-# Of course, when you leave a directory the variables should be unset. I track this by having a map between directory and a vector of environment variables. If the user is not in the directory or one of its subdirectories, its environment variables are removed... And then for some reason they are re-added again?
+and now you suddenly have nushell picking up whatever environment variables they set.
+This meant that the code necessarily become more involved than just looking for a .nu-file in the current directory and applying its variables,
+but this extra code also allowed for some other nice features and behavior which is listed below.
 
 Behavior:
 - If you are in a subdirectory to a directory with a .nu-file, the vars in that .nu-file are applied.
@@ -24,8 +26,9 @@ Behavior:
   - TODO: What happens if you overwrite twice?
 
 Questions:
-- ´add_env´ does not overwrite variables. Need ´add_env_force´?
+`remove_env()` accesses the environment directly to remove variables.
-- `ctx.get_env()` in `cli.rs` lacks access to the config, which is required. Is it ok to do it through the sync call instead?
+Just using what I think was expected, envs.entries.remove(key), does not work.
+If this is a problem, how can I make it work better with existing code?
 
 TODO: take care of situation where a directory overwrites an existing .nu conf.