Merge pull request #3259 from jfinkels/df-loops

df: separate functions for two main modes of df

src/uu/df/src/df.rs

@@ -230,39 +230,49 @@ fn filter_mount_list(vmi: Vec<MountInfo>, opt: &Options) -> Vec<MountInfo> {
     result
 }
 
-/// Assign 1 `MountInfo` entry to each path
-/// `lofs` entries are skipped and dummy mount points are skipped
-/// Only the longest matching prefix for that path is considered
-/// `lofs` is for Solaris style loopback filesystem and is present in Solaris and FreeBSD.
-/// It works similar to symlinks
-fn get_point_list(vmi: &[MountInfo], paths: &[String]) -> Vec<MountInfo> {
+/// Get all currently mounted filesystems.
+///
+/// `opt` excludes certain filesystems from consideration; see
+/// [`Options`] for more information.
+fn get_all_filesystems(opt: &Options) -> Vec<Filesystem> {
+    // The list of all mounted filesystems.
+    //
+    // Filesystems excluded by the command-line options are
+    // not considered.
+    let mounts: Vec<MountInfo> = filter_mount_list(read_fs_list(), opt);
+
+    // Convert each `MountInfo` into a `Filesystem`, which contains
+    // both the mount information and usage information.
+    mounts.into_iter().filter_map(Filesystem::new).collect()
+}
+
+/// For each path, get the filesystem that contains that path.
+fn get_named_filesystems<P>(paths: &[P]) -> Vec<Filesystem>
+where
+    P: AsRef<Path>,
+{
+    // The list of all mounted filesystems.
+    //
+    // Filesystems marked as `dummy` or of type "lofs" are not
+    // considered. The "lofs" filesystem is a loopback
+    // filesystem present on Solaris and FreeBSD systems. It
+    // is similar to a symbolic link.
+    let mounts: Vec<MountInfo> = read_fs_list()
+        .into_iter()
+        .filter(|mi| mi.fs_type != "lofs" && !mi.dummy)
+        .collect();
+
+    // Convert each path into a `Filesystem`, which contains
+    // both the mount information and usage information.
     paths
         .iter()
-        .map(|p| {
-            vmi.iter()
-                .filter(|mi| mi.fs_type.ne("lofs"))
-                .filter(|mi| !mi.dummy)
-                .filter(|mi| p.starts_with(&mi.mount_dir))
-                .max_by_key(|mi| mi.mount_dir.len())
-                .unwrap()
-                .clone()
-        })
-        .collect::<Vec<MountInfo>>()
+        .filter_map(|p| Filesystem::from_path(&mounts, p))
+        .collect()
 }
 
 #[uucore::main]
 pub fn uumain(args: impl uucore::Args) -> UResult<()> {
     let matches = uu_app().get_matches_from(args);
-
-    // Canonicalize the input_paths and then convert to string
-    let paths = matches
-        .values_of(OPT_PATHS)
-        .unwrap_or_default()
-        .map(Path::new)
-        .filter_map(|v| v.canonicalize().ok())
-        .filter_map(|v| v.into_os_string().into_string().ok())
-        .collect::<Vec<_>>();
-
     #[cfg(windows)]
     {
         if matches.is_present(OPT_INODES) {
@@ -273,27 +283,31 @@ pub fn uumain(args: impl uucore::Args) -> UResult<()> {
 
     let opt = Options::from(&matches).map_err(|e| USimpleError::new(1, format!("{}", e)))?;
 
-    let mounts = read_fs_list();
-
-    let op_mount_points: Vec<MountInfo> = if paths.is_empty() {
-        // Get all entries
-        filter_mount_list(mounts, &opt)
-    } else {
-        // Get Point for each input_path
-        get_point_list(&mounts, &paths)
+    // Get the list of filesystems to display in the output table.
+    let filesystems: Vec<Filesystem> = match matches.values_of(OPT_PATHS) {
+        None => get_all_filesystems(&opt),
+        Some(paths) => {
+            let paths: Vec<&str> = paths.collect();
+            get_named_filesystems(&paths)
+        }
     };
-    let data: Vec<Row> = op_mount_points
-        .into_iter()
-        .filter_map(Filesystem::new)
-        .filter(|fs| fs.usage.blocks != 0 || opt.show_all_fs || opt.show_listed_fs)
-        .map(Into::into)
-        .collect();
+
+    // The running total of filesystem sizes and usage.
+    //
+    // This accumulator is computed in case we need to display the
+    // total counts in the last row of the table.
+    let mut total = Row::new("total");
 
     println!("{}", Header::new(&opt));
-    let mut total = Row::new("total");
-    for row in data {
-        println!("{}", DisplayRow::new(&row, &opt));
-        total += row;
+    for filesystem in filesystems {
+        // If the filesystem is not empty, or if the options require
+        // showing all filesystems, then print the data as a row in
+        // the output table.
+        if opt.show_all_fs || opt.show_listed_fs || filesystem.usage.blocks > 0 {
+            let row = Row::from(filesystem);
+            println!("{}", DisplayRow::new(&row, &opt));
+            total += row;
+        }
     }
     if opt.show_total {
         println!("{}", DisplayRow::new(&total, &opt));

src/uu/df/src/filesystem.rs

@@ -7,7 +7,7 @@
 //! A [`Filesystem`] struct represents a device containing a
 //! filesystem mounted at a particular directory. It also includes
 //! information on amount of space available and amount of space used.
-#[cfg(windows)]
+// spell-checker:ignore canonicalized
 use std::path::Path;
 
 #[cfg(unix)]
@@ -30,6 +30,40 @@ pub(crate) struct Filesystem {
     pub usage: FsUsage,
 }
 
+/// Find the mount info that best matches a given filesystem path.
+///
+/// This function returns the element of `mounts` on which `path` is
+/// mounted. If there are no matches, this function returns
+/// [`None`]. If there are two or more matches, then the single
+/// [`MountInfo`] with the longest mount directory is returned.
+///
+/// If `canonicalize` is `true`, then the `path` is canonicalized
+/// before checking whether it matches any mount directories.
+///
+/// # See also
+///
+/// * [`Path::canonicalize`]
+/// * [`MountInfo::mount_dir`]
+fn mount_info_from_path<P>(
+    mounts: &[MountInfo],
+    path: P,
+    // This is really only used for testing purposes.
+    canonicalize: bool,
+) -> Option<&MountInfo>
+where
+    P: AsRef<Path>,
+{
+    // TODO Refactor this function with `Stater::find_mount_point()`
+    // in the `stat` crate.
+    let path = if canonicalize {
+        path.as_ref().canonicalize().ok()?
+    } else {
+        path.as_ref().to_path_buf()
+    };
+    let matches = mounts.iter().filter(|mi| path.starts_with(&mi.mount_dir));
+    matches.max_by_key(|mi| mi.mount_dir.len())
+}
+
 impl Filesystem {
     // TODO: resolve uuid in `mount_info.dev_name` if exists
     pub(crate) fn new(mount_info: MountInfo) -> Option<Self> {
@@ -52,4 +86,106 @@ impl Filesystem {
         let usage = FsUsage::new(Path::new(&_stat_path));
         Some(Self { mount_info, usage })
     }
+
+    /// Find and create the filesystem that best matches a given path.
+    ///
+    /// This function returns a new `Filesystem` derived from the
+    /// element of `mounts` on which `path` is mounted. If there are
+    /// no matches, this function returns [`None`]. If there are two
+    /// or more matches, then the single [`Filesystem`] with the
+    /// longest mount directory is returned.
+    ///
+    /// The `path` is canonicalized before checking whether it matches
+    /// any mount directories.
+    ///
+    /// # See also
+    ///
+    /// * [`Path::canonicalize`]
+    /// * [`MountInfo::mount_dir`]
+    ///
+    pub(crate) fn from_path<P>(mounts: &[MountInfo], path: P) -> Option<Self>
+    where
+        P: AsRef<Path>,
+    {
+        let canonicalize = true;
+        let mount_info = mount_info_from_path(mounts, path, canonicalize)?;
+        // TODO Make it so that we do not need to clone the `mount_info`.
+        let mount_info = (*mount_info).clone();
+        Self::new(mount_info)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    mod mount_info_from_path {
+
+        use uucore::fsext::MountInfo;
+
+        use crate::filesystem::mount_info_from_path;
+
+        // Create a fake `MountInfo` with the given directory name.
+        fn mount_info(mount_dir: &str) -> MountInfo {
+            MountInfo {
+                dev_id: Default::default(),
+                dev_name: Default::default(),
+                fs_type: Default::default(),
+                mount_dir: String::from(mount_dir),
+                mount_option: Default::default(),
+                mount_root: Default::default(),
+                remote: Default::default(),
+                dummy: Default::default(),
+            }
+        }
+
+        // Check whether two `MountInfo` instances are equal.
+        fn mount_info_eq(m1: &MountInfo, m2: &MountInfo) -> bool {
+            m1.dev_id == m2.dev_id
+                && m1.dev_name == m2.dev_name
+                && m1.fs_type == m2.fs_type
+                && m1.mount_dir == m2.mount_dir
+                && m1.mount_option == m2.mount_option
+                && m1.mount_root == m2.mount_root
+                && m1.remote == m2.remote
+                && m1.dummy == m2.dummy
+        }
+
+        #[test]
+        fn test_empty_mounts() {
+            assert!(mount_info_from_path(&[], "/", false).is_none());
+        }
+
+        #[test]
+        fn test_exact_match() {
+            let mounts = [mount_info("/foo")];
+            let actual = mount_info_from_path(&mounts, "/foo", false).unwrap();
+            assert!(mount_info_eq(actual, &mounts[0]));
+        }
+
+        #[test]
+        fn test_prefix_match() {
+            let mounts = [mount_info("/foo")];
+            let actual = mount_info_from_path(&mounts, "/foo/bar", false).unwrap();
+            assert!(mount_info_eq(actual, &mounts[0]));
+        }
+
+        #[test]
+        fn test_multiple_matches() {
+            let mounts = [mount_info("/foo"), mount_info("/foo/bar")];
+            let actual = mount_info_from_path(&mounts, "/foo/bar", false).unwrap();
+            assert!(mount_info_eq(actual, &mounts[1]));
+        }
+
+        #[test]
+        fn test_no_match() {
+            let mounts = [mount_info("/foo")];
+            assert!(mount_info_from_path(&mounts, "/bar", false).is_none());
+        }
+
+        #[test]
+        fn test_partial_match() {
+            let mounts = [mount_info("/foo/bar")];
+            assert!(mount_info_from_path(&mounts, "/foo/baz", false).is_none());
+        }
+    }
 }