Julia jupyter interpreter is up!

Though it's not as simple as other interpreters...
(The kernel has to be manually started and stopped)

README.md

@@ -64,7 +64,7 @@ An example in C, look in the command area:
 
 ![](ressources/visual_assets/demo_c.gif)
 
-##### The result can be displayed in multiple (even at the same time) ways: 
+##### The result can be displayed in multiple (even at the same time) ways:
 
 [Classic](ressources/display_classic.md)|  [Virtual Text](ressources/display_virtualtext.md)
 :------------------------------------------:|:------------------:
@@ -95,7 +95,7 @@ Basically, it allows you to run a part of your code.
 Do either of:
 
 - Position the cursor on a line `:SnipRun`
-- Select some visual range, `:'<,'>SnipRun` 
+- Select some visual range, `:'<,'>SnipRun`
 - Combine a motion with the operator
 
   (preferably through a shortcut!)
@@ -111,7 +111,7 @@ Sniprun will then:
 - **Add boilerplate** when it exists. In C, it surrounds your snip with "int main() {", "}". (disclaimer: oversimplifed)
 - **Build** (write to a script file, or compile) the code
 - **Execute** the code
-- Return stdout, or stderr using the 
+- Return stdout, or stderr using the
 
 
 ![](ressources/visual_assets/760091.png)
@@ -275,7 +275,7 @@ require'sniprun'.setup({
     -- "LongTempFloatingWindow",  -- "same as above, but only long results. To use with VirtualText__
     -- "Terminal"                 -- "display results in a vertical split
     },
-    
+
   -- customize highlight groups (setting this overrides colorscheme)
  snipruncolors = {
     SniprunVirtualTextOk   =  {bg="#66eeff",fg="#000000",ctermbg="Cyan",cterfg="Black"},
@@ -387,15 +387,15 @@ println!("-> {}", alphabet);
 | ------------ | ------------- | --- | ---------- | ---------------- |
 | Ada          | Line          |     | Java       | Bloc             |
 | Bash/Shell   | Bloc + REPL\* |     | JavaScript | Bloc             |
-| C            | Import        |     | Julia      | Bloc             |
+| C            | Import        |     | Julia      | Bloc + REPL\*\*  |
 | C++          | Import        |     | Lisp       | Untested         |
 | Clojure      | Untested      |     | Lua        | Bloc             |
 | COBOL        | Untested      |     | Lua-nvim   | Bloc             |
-| Coffeescript | Bloc          |     | Markdown (GFM)   | Bloc + REPL \***         |
+| Coffeescript | Bloc          |     | Markdown (GFM)   | Bloc + REPL \*\*\*   |
 | C#           | Untested      |     | Perl6      | Line             |
 | D            | Bloc          |     | Perl       | Line             |
 | Elixir       | Untested      |     | PHP        | Untested         |
-| Elm          | Untested      |     | Python3    | Import +REPL\*\* |
+| Elm          | Untested      |     | Python3    | Import +REPL*\*   |
 | Erlang       | Untested      |     | R          | Bloc + REPL \*\* |
 | F#           | Untested      |     | Ruby       | Bloc             |
 | Go           | Bloc          |     | Rust       | Bloc             |
@@ -425,7 +425,7 @@ Due to its nature, Sniprun may have trouble with programs that :
 
 ## Troubleshooting
 
-begin by updating the plugin and running `:SnipReset` and then `:checkhealth sniprun` 
+begin by updating the plugin and running `:SnipReset` and then `:checkhealth sniprun`
 
 - **Silent fail**: the sniprun binary may be incompatible with your distro/OS/arch. Use `bash ./install.sh 1` as post-install to compile locally.
 - Terminal and Floating Window display mode do not work: Linked to [this](https://github.com/michaelb/sniprun/issues/70) issue, no fix found yet.

doc/Julia_jupyter.md

@@ -0,0 +1,12 @@
+The setup for the julia_jupyter interpreter is quite convoluted:
+
+Indeed, the Julia jupyter kernel MUST be started before Sniprun can run (and there is even a consequent delay since the kernel is so slow to start).
+
+You should start a julia jupyter kernel with the following command: 
+` jupyter-kernel --kernel=julia-1.5 --KernelManager.connection_file=$HOME/.cache/sniprun/julia_jupyter/kernel_sniprun.json`
+
+(adapt to your XDG_CACHE location if you're on Mac)
+
+You manage kernel startup AND shutodwn manually. Why? There is a terrible data race if sniprun does it. Python_jupyter works, julia doesn't. That's about it.
+
+If you want to use another kernel location, post a feature request at github.com/michaelb/sniprun and i'll see what I can do.

src/interpreters/Julia_jupyter.rs

@@ -123,28 +123,10 @@ impl Interpreter for Julia_jupyter {
 impl ReplLikeInterpreter for Julia_jupyter {
     fn fetch_code_repl(&mut self) -> Result<(), SniprunError> {
         self.fetch_code()?;
-        let saved_code = self.read_previous_code();
-        if saved_code.is_empty() {
-            //initialize kernel. Relying on self.read_previous_code to
-            //know when to start a new kernel is important as
-            //this will be cleared by the SnipReplMemoryClean command
-            let _res = std::fs::remove_file(&self.kernel_file);
-            let _res = Command::new("jupyter-kernel")
-                .arg("--kernel=julia-1.5")
-                .arg(String::from("--KernelManager.connection_file=") + &self.kernel_file)
-                .spawn();
-            info!("Initialized kernel at {}", self.kernel_file);
-        } else {
-            // kernel already running
-            info!(
-                "Using already loaded jupyter kernel at {}",
-                self.kernel_file
-            );
+        if !std::path::Path::new(&self.kernel_file).exists() {
+            info!("no kernel file found");
+            return Err(SniprunError::RuntimeError("No running kernel found".to_string()));
         }
-
-        // save kernel
-        self.save_code(saved_code);
-
         Ok(())
     }
     fn add_boilerplate_repl(&mut self) -> Result<(), SniprunError> {
@@ -186,15 +168,6 @@ impl ReplLikeInterpreter for Julia_jupyter {
     }
 
     fn execute_repl(&mut self) -> Result<String, SniprunError> {
-        if !std::path::Path::new(&self.kernel_file).exists() {
-            info!("no kernel file found");
-            return Err(SniprunError::RuntimeError("No kernel found or provided".to_string()));
-        }
-        info!(
-            "json kernel file exists yet? {}",
-            std::path::Path::new(&self.kernel_file).exists()
-        );
-
         info!("starting executing repl: bash {}",&self.launcher_path);
         let output = Command::new("bash")
             .arg(&self.launcher_path)
@@ -234,25 +207,3 @@ impl ReplLikeInterpreter for Julia_jupyter {
         }
     }
 }
-
-#[cfg(test)]
-mod test_julia_jupyter {
-    use super::*;
-    use crate::*;
-
-    #[test]
-    fn run_all() {
-        simple_print();
-    }
-
-    fn simple_print() {
-        let mut data = DataHolder::new();
-        data.current_bloc = String::from("println(\"a\")");
-        let mut interpreter = Julia_jupyter::new(data);
-        let res = interpreter.run_at_level(SupportLevel::Bloc);
-
-        // should panic if not an Ok()
-        let string_result = res.unwrap();
-        assert!(string_result.contains(&"a"));
-    }
-}