Fix env duplication on merged-usr systems (#200)

On modern (merged-usr) Linux systems, `/bin`, `/sbin` and `/usr/sbin`
are symlinks to `/usr/bin`.  There is no point in reporting the same
Python installation four times, so canonicalize search paths before
searching.

Before:

	Breakdown for finding Environments:
	-----------------------------------
	GlobalVirtualEnvs    : 17.412201ms
	Locators             : 225.284494ms
	Path                 : 433.162905ms
	Workspaces           : 2.161556ms

	Environments (41):
	------------------
	GlobalPaths          : 14
	LinuxGlobal          : 16
	VirtualEnvWrapper    : 11

After:

	Breakdown for finding Environments:
	-----------------------------------
	GlobalVirtualEnvs    : 16.595382ms
	Locators             : 223.759511ms
	Path                 : 313.276036ms
	Workspaces           : 1.418024ms

	Environments (21):
	------------------
	GlobalPaths          : 2
	LinuxGlobal          : 8
	VirtualEnvWrapper    : 11

---------

Co-authored-by: Karthik Nadig <[email protected]>

Author: elprans

crates/pet-env-var-path/src/lib.rs

@@ -2,6 +2,8 @@
 // Licensed under the MIT License.
 
 use pet_core::os_environment::Environment;
+use std::collections::HashSet;
+use std::fs;
 use std::path::PathBuf;
 
 pub fn get_search_paths_from_env_variables(environment: &dyn Environment) -> Vec<PathBuf> {
@@ -16,10 +18,12 @@ pub fn get_search_paths_from_env_variables(environment: &dyn Environment) -> Vec
 
         environment
             .get_know_global_search_locations()
-            .clone()
+            .into_iter()
+            .map(|p| fs::canonicalize(&p).unwrap_or(p))
+            .collect::<HashSet<PathBuf>>()
             .into_iter()
             .filter(|p| !p.starts_with(apps_path.clone()))
-            .collect::<Vec<PathBuf>>()
+            .collect()
     } else {
         Vec::new()
     }

crates/pet-linux-global-python/src/lib.rs

@@ -2,7 +2,7 @@
 // Licensed under the MIT License.
 
 use std::{
-    collections::HashMap,
+    collections::{HashMap, HashSet},
     fs,
     path::{Path, PathBuf},
     sync::{Arc, Mutex},
@@ -38,10 +38,17 @@ impl LinuxGlobalPython {
             return;
         }
         // Look through the /bin, /usr/bin, /usr/local/bin directories
+        let bin_dirs: HashSet<_> = [
+            Path::new("/bin"),
+            Path::new("/usr/bin"),
+            Path::new("/usr/local/bin"),
+        ]
+        .map(|p| fs::canonicalize(p).unwrap_or(p.to_path_buf()))
+        .into();
         thread::scope(|s| {
-            for bin in ["/bin", "/usr/bin", "/usr/local/bin"] {
+            for bin in bin_dirs {
                 s.spawn(move || {
-                    find_and_report_global_pythons_in(bin, reporter, &self.reported_executables);
+                    find_and_report_global_pythons_in(&bin, reporter, &self.reported_executables);
                 });
             }
         });
@@ -103,11 +110,11 @@ impl Locator for LinuxGlobalPython {
 }
 
 fn find_and_report_global_pythons_in(
-    bin: &str,
+    bin: &Path,
     reporter: Option<&dyn Reporter>,
     reported_executables: &Arc<Mutex<HashMap<PathBuf, PythonEnvironment>>>,
 ) {
-    let python_executables = find_executables(Path::new(bin));
+    let python_executables = find_executables(bin);
 
     for exe in python_executables.clone().iter() {
         if reported_executables.lock().unwrap().contains_key(exe) {

crates/pet/tests/ci_jupyter_container.rs

@@ -71,30 +71,16 @@ fn verify_python_in_jupyter_contaner() {
         }),
         ..Default::default()
     };
-    let codespace_python = PythonEnvironment {
-        kind: Some(PythonEnvironmentKind::GlobalPaths),
-        executable: Some(PathBuf::from("/home/codespace/.python/current/bin/python")),
-        prefix: Some(PathBuf::from("/usr/local/python/3.10.13")),
-        version: Some("3.10.13.final.0".to_string()),
-        arch: Some(Architecture::X64),
-        symlinks: Some(vec![
-            PathBuf::from("/home/codespace/.python/current/bin/python"),
-            PathBuf::from("/home/codespace/.python/current/bin/python3"),
-            PathBuf::from("/home/codespace/.python/current/bin/python3.10"),
-        ]),
-        manager: None,
-        ..Default::default()
-    };
     let current_python = PythonEnvironment {
         kind: Some(PythonEnvironmentKind::GlobalPaths),
-        executable: Some(PathBuf::from("/usr/local/python/current/bin/python")),
+        executable: Some(PathBuf::from("/usr/local/python/3.10.13/bin/python")),
         prefix: Some(PathBuf::from("/usr/local/python/3.10.13")),
         version: Some("3.10.13.final.0".to_string()),
         arch: Some(Architecture::X64),
         symlinks: Some(vec![
-            PathBuf::from("/usr/local/python/current/bin/python"),
-            PathBuf::from("/usr/local/python/current/bin/python3"),
-            PathBuf::from("/usr/local/python/current/bin/python3.10"),
+            PathBuf::from("/usr/local/python/3.10.13/bin/python"),
+            PathBuf::from("/usr/local/python/3.10.13/bin/python3"),
+            PathBuf::from("/usr/local/python/3.10.13/bin/python3.10"),
         ]),
         manager: None,
         ..Default::default()
@@ -112,29 +98,8 @@ fn verify_python_in_jupyter_contaner() {
         manager: None,
         ..Default::default()
     };
-    let bin_python = PythonEnvironment {
-        kind: Some(PythonEnvironmentKind::LinuxGlobal),
-        executable: Some(PathBuf::from("/bin/python3")),
-        prefix: Some(PathBuf::from("/usr")),
-        version: Some("3.8.10.final.0".to_string()),
-        arch: Some(Architecture::X64),
-        symlinks: Some(vec![
-            PathBuf::from("/bin/python3"),
-            PathBuf::from("/bin/python3.8"),
-        ]),
-        manager: None,
-        ..Default::default()
-    };
 
-    for env in [
-        conda,
-        codespace_python,
-        current_python,
-        usr_bin_python,
-        bin_python,
-    ]
-    .iter()
-    {
+    for env in [conda, current_python, usr_bin_python].iter() {
         let python_env = environments
             .iter()
             .find(|e| e.executable == env.executable)