Create helix-stdx crate for stdlib extensions

helix-stdx is meant to carry extensions to the stdlib or low-level
dependencies that are useful in all other crates. This commit starts
with all of the path functions from helix-core and the CWD tracking that
lived in helix-loader.

The CWD tracking in helix-loader was previously unable to call the
canonicalization functions in helix-core. Switching to our custom
canonicalization code should make no noticeable difference though
since `std::env::current_dir` returns a canonicalized path with
symlinks resolved (at least on unix).

5 files changed, 378 insertions, 0 deletions

diff --git a/helix-stdx/Cargo.toml b/helix-stdx/Cargo.toml
new file mode 100644
index 00000000..216a3b40
--- /dev/null
+++ b/helix-stdx/Cargo.toml
@@ -0,0 +1,19 @@
+[package]
+name = "helix-stdx"
+description = "Standard library extensions"
+include = ["src/**/*", "README.md"]
+version.workspace = true
+authors.workspace = true
+edition.workspace = true
+license.workspace = true
+rust-version.workspace = true
+categories.workspace = true
+repository.workspace = true
+homepage.workspace = true
+
+[dependencies]
+dunce = "1.0"
+etcetera = "0.8"
+
+[dev-dependencies]
+tempfile = "3.9"
diff --git a/helix-stdx/src/env.rs b/helix-stdx/src/env.rs
new file mode 100644
index 00000000..864ba828
--- /dev/null
+++ b/helix-stdx/src/env.rs
@@ -0,0 +1,48 @@
+use std::{
+    path::{Path, PathBuf},
+    sync::RwLock,
+};
+
+static CWD: RwLock<Option<PathBuf>> = RwLock::new(None);
+
+// Get the current working directory.
+// This information is managed internally as the call to std::env::current_dir
+// might fail if the cwd has been deleted.
+pub fn current_working_dir() -> PathBuf {
+    if let Some(path) = &*CWD.read().unwrap() {
+        return path.clone();
+    }
+
+    let path = std::env::current_dir()
+        .map(crate::path::normalize)
+        .expect("Couldn't determine current working directory");
+    let mut cwd = CWD.write().unwrap();
+    *cwd = Some(path.clone());
+
+    path
+}
+
+pub fn set_current_working_dir(path: impl AsRef<Path>) -> std::io::Result<()> {
+    let path = crate::path::canonicalize(path);
+    std::env::set_current_dir(&path)?;
+    let mut cwd = CWD.write().unwrap();
+    *cwd = Some(path);
+    Ok(())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{current_working_dir, set_current_working_dir};
+
+    #[test]
+    fn current_dir_is_set() {
+        let new_path = dunce::canonicalize(std::env::temp_dir()).unwrap();
+        let cwd = current_working_dir();
+        assert_ne!(cwd, new_path);
+
+        set_current_working_dir(&new_path).expect("Couldn't set new path");
+
+        let cwd = current_working_dir();
+        assert_eq!(cwd, new_path);
+    }
+}
diff --git a/helix-stdx/src/lib.rs b/helix-stdx/src/lib.rs
new file mode 100644
index 00000000..ae3c3a98
--- /dev/null
+++ b/helix-stdx/src/lib.rs
@@ -0,0 +1,2 @@
+pub mod env;
+pub mod path;
diff --git a/helix-stdx/src/path.rs b/helix-stdx/src/path.rs
new file mode 100644
index 00000000..5746657c
--- /dev/null
+++ b/helix-stdx/src/path.rs
@@ -0,0 +1,185 @@
+pub use etcetera::home_dir;
+
+use std::path::{Component, Path, PathBuf};
+
+use crate::env::current_working_dir;
+
+/// Replaces users home directory from `path` with tilde `~` if the directory
+/// is available, otherwise returns the path unchanged.
+pub fn fold_home_dir(path: &Path) -> PathBuf {
+    if let Ok(home) = home_dir() {
+        if let Ok(stripped) = path.strip_prefix(&home) {
+            return PathBuf::from("~").join(stripped);
+        }
+    }
+
+    path.to_path_buf()
+}
+
+/// Expands tilde `~` into users home directory if available, otherwise returns the path
+/// unchanged. The tilde will only be expanded when present as the first component of the path
+/// and only slash follows it.
+pub fn expand_tilde(path: impl AsRef<Path>) -> PathBuf {
+    let path = path.as_ref();
+    let mut components = path.components().peekable();
+    if let Some(Component::Normal(c)) = components.peek() {
+        if c == &"~" {
+            if let Ok(home) = home_dir() {
+                // it's ok to unwrap, the path starts with `~`
+                return home.join(path.strip_prefix("~").unwrap());
+            }
+        }
+    }
+
+    path.to_path_buf()
+}
+
+/// Normalize a path without resolving symlinks.
+// Strategy: start from the first component and move up. Cannonicalize previous path,
+// join component, cannonicalize new path, strip prefix and join to the final result.
+pub fn normalize(path: impl AsRef<Path>) -> PathBuf {
+    let mut components = path.as_ref().components().peekable();
+    let mut ret = if let Some(c @ Component::Prefix(..)) = components.peek().cloned() {
+        components.next();
+        PathBuf::from(c.as_os_str())
+    } else {
+        PathBuf::new()
+    };
+
+    for component in components {
+        match component {
+            Component::Prefix(..) => unreachable!(),
+            Component::RootDir => {
+                ret.push(component.as_os_str());
+            }
+            Component::CurDir => {}
+            #[cfg(not(windows))]
+            Component::ParentDir => {
+                ret.pop();
+            }
+            #[cfg(windows)]
+            Component::ParentDir => {
+                if let Some(head) = ret.components().next_back() {
+                    match head {
+                        Component::Prefix(_) | Component::RootDir => {}
+                        Component::CurDir => unreachable!(),
+                        // If we left previous component as ".." it means we met a symlink before and we can't pop path.
+                        Component::ParentDir => {
+                            ret.push("..");
+                        }
+                        Component::Normal(_) => {
+                            if ret.is_symlink() {
+                                ret.push("..");
+                            } else {
+                                ret.pop();
+                            }
+                        }
+                    }
+                }
+            }
+            #[cfg(not(windows))]
+            Component::Normal(c) => {
+                ret.push(c);
+            }
+            #[cfg(windows)]
+            Component::Normal(c) => 'normal: {
+                use std::fs::canonicalize;
+
+                let new_path = ret.join(c);
+                if new_path.is_symlink() {
+                    ret = new_path;
+                    break 'normal;
+                }
+                let (can_new, can_old) = (canonicalize(&new_path), canonicalize(&ret));
+                match (can_new, can_old) {
+                    (Ok(can_new), Ok(can_old)) => {
+                        let striped = can_new.strip_prefix(can_old);
+                        ret.push(striped.unwrap_or_else(|_| c.as_ref()));
+                    }
+                    _ => ret.push(c),
+                }
+            }
+        }
+    }
+    dunce::simplified(&ret).to_path_buf()
+}
+
+/// Returns the canonical, absolute form of a path with all intermediate components normalized.
+///
+/// This function is used instead of [`std::fs::canonicalize`] because we don't want to verify
+/// here if the path exists, just normalize it's components.
+pub fn canonicalize(path: impl AsRef<Path>) -> PathBuf {
+    let path = expand_tilde(path);
+    let path = if path.is_relative() {
+        current_working_dir().join(path)
+    } else {
+        path
+    };
+
+    normalize(path)
+}
+
+pub fn get_relative_path(path: impl AsRef<Path>) -> PathBuf {
+    let path = PathBuf::from(path.as_ref());
+    let path = if path.is_absolute() {
+        let cwdir = normalize(current_working_dir());
+        normalize(&path)
+            .strip_prefix(cwdir)
+            .map(PathBuf::from)
+            .unwrap_or(path)
+    } else {
+        path
+    };
+    fold_home_dir(&path)
+}
+
+/// Returns a truncated filepath where the basepart of the path is reduced to the first
+/// char of the folder and the whole filename appended.
+///
+/// Also strip the current working directory from the beginning of the path.
+/// Note that this function does not check if the truncated path is unambiguous.
+///
+/// ```
+///    use helix_stdx::path::get_truncated_path;
+///    use std::path::Path;
+///
+///    assert_eq!(
+///         get_truncated_path("/home/cnorris/documents/jokes.txt").as_path(),
+///         Path::new("/h/c/d/jokes.txt")
+///     );
+///     assert_eq!(
+///         get_truncated_path("jokes.txt").as_path(),
+///         Path::new("jokes.txt")
+///     );
+///     assert_eq!(
+///         get_truncated_path("/jokes.txt").as_path(),
+///         Path::new("/jokes.txt")
+///     );
+///     assert_eq!(
+///         get_truncated_path("/h/c/d/jokes.txt").as_path(),
+///         Path::new("/h/c/d/jokes.txt")
+///     );
+///     assert_eq!(get_truncated_path("").as_path(), Path::new(""));
+/// ```
+///
+pub fn get_truncated_path(path: impl AsRef<Path>) -> PathBuf {
+    let cwd = current_working_dir();
+    let path = path
+        .as_ref()
+        .strip_prefix(cwd)
+        .unwrap_or_else(|_| path.as_ref());
+    let file = path.file_name().unwrap_or_default();
+    let base = path.parent().unwrap_or_else(|| Path::new(""));
+    let mut ret = PathBuf::new();
+    for d in base {
+        ret.push(
+            d.to_string_lossy()
+                .chars()
+                .next()
+                .unwrap_or_default()
+                .to_string(),
+        );
+    }
+    ret.push(file);
+    ret
+}
diff --git a/helix-stdx/tests/path.rs b/helix-stdx/tests/path.rs
new file mode 100644
index 00000000..cc3c15cb
--- /dev/null
+++ b/helix-stdx/tests/path.rs
@@ -0,0 +1,124 @@
+#![cfg(windows)]
+
+use std::{
+    env::set_current_dir,
+    error::Error,
+    path::{Component, Path, PathBuf},
+};
+
+use helix_stdx::path;
+use tempfile::Builder;
+
+// Paths on Windows are almost always case-insensitive.
+// Normalization should return the original path.
+// E.g. mkdir `CaSe`, normalize(`case`) = `CaSe`.
+#[test]
+fn test_case_folding_windows() -> Result<(), Box<dyn Error>> {
+    // tmp/root/case
+    let tmp_prefix = std::env::temp_dir();
+    set_current_dir(&tmp_prefix)?;
+
+    let root = Builder::new().prefix("root-").tempdir()?;
+    let case = Builder::new().prefix("CaSe-").tempdir_in(&root)?;
+
+    let root_without_prefix = root.path().strip_prefix(&tmp_prefix)?;
+
+    let lowercase_case = format!(
+        "case-{}",
+        case.path()
+            .file_name()
+            .unwrap()
+            .to_string_lossy()
+            .split_at(5)
+            .1
+    );
+    let test_path = root_without_prefix.join(lowercase_case);
+    assert_eq!(
+        path::normalize(&test_path),
+        case.path().strip_prefix(&tmp_prefix)?
+    );
+
+    Ok(())
+}
+
+#[test]
+fn test_normalize_path() -> Result<(), Box<dyn Error>> {
+    /*
+    tmp/root/
+    ├── link -> dir1/orig_file
+    ├── dir1/
+    │   └── orig_file
+    └── dir2/
+        └── dir_link -> ../dir1/
+    */
+
+    let tmp_prefix = std::env::temp_dir();
+    set_current_dir(&tmp_prefix)?;
+
+    // Create a tree structure as shown above
+    let root = Builder::new().prefix("root-").tempdir()?;
+    let dir1 = Builder::new().prefix("dir1-").tempdir_in(&root)?;
+    let orig_file = Builder::new().prefix("orig_file-").tempfile_in(&dir1)?;
+    let dir2 = Builder::new().prefix("dir2-").tempdir_in(&root)?;
+
+    // Create path and delete existing file
+    let dir_link = Builder::new()
+        .prefix("dir_link-")
+        .tempfile_in(&dir2)?
+        .path()
+        .to_owned();
+    let link = Builder::new()
+        .prefix("link-")
+        .tempfile_in(&root)?
+        .path()
+        .to_owned();
+
+    use std::os::windows;
+    windows::fs::symlink_dir(&dir1, &dir_link)?;
+    windows::fs::symlink_file(&orig_file, &link)?;
+
+    // root/link
+    let path = link.strip_prefix(&tmp_prefix)?;
+    assert_eq!(
+        path::normalize(path),
+        path,
+        "input {:?} and symlink last component shouldn't be resolved",
+        path
+    );
+
+    // root/dir2/dir_link/orig_file/../..
+    let path = dir_link
+        .strip_prefix(&tmp_prefix)
+        .unwrap()
+        .join(orig_file.path().file_name().unwrap())
+        .join(Component::ParentDir)
+        .join(Component::ParentDir);
+    let expected = dir_link
+        .strip_prefix(&tmp_prefix)
+        .unwrap()
+        .join(Component::ParentDir);
+    assert_eq!(
+        path::normalize(&path),
+        expected,
+        "input {:?} and \"..\" should not erase the simlink that goes ahead",
+        &path
+    );
+
+    // root/link/.././../dir2/../
+    let path = link
+        .strip_prefix(&tmp_prefix)
+        .unwrap()
+        .join(Component::ParentDir)
+        .join(Component::CurDir)
+        .join(Component::ParentDir)
+        .join(dir2.path().file_name().unwrap())
+        .join(Component::ParentDir);
+    let expected = link
+        .strip_prefix(&tmp_prefix)
+        .unwrap()
+        .join(Component::ParentDir)
+        .join(Component::ParentDir);
+    assert_eq!(path::normalize(&path), expected, "input {:?}", &path);
+
+    Ok(())
+}