nick
/
hb
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

tweaked CLI options, fixed -p behaviour

main
nick 2024-03-14 09:53:12 -04:00
parent 0af9b732ab
commit 518310b00f
3 changed files with 115 additions and 104 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

72
src/args.rs
View File

@ -12,12 +12,21 @@ use std::iter::once_with;
#[derive(Parser, Debug, Clone)]
pub struct Args {
#[arg(
short, long, default_value_t = false,
short, long,
help = "keep going if an error occurs",
long_help = "keep going if an error occurs (ex. unreadable subdirectories in a readable directory)"
long_help = "keep going if an error occurs, silencing them in the process",
default_value_t = false,
)]
persistant: bool,
#[arg(
short, long,
help = "suppress error messages",
long_help = "suppress error messages, but still quit if an error occurs",
default_value_t = false,
)]
quiet: bool,
#[arg(
short, long,
help = "minimize output",
@ -79,7 +88,7 @@ pub struct Args {
#[arg(
short='x', long = "exclude",
help = "include in search, but exclude from printing",
long_help = "include in search, but exclude from printing. accepts glob syntax",
long_help = "include in search, but exclude from printing. accepts glob syntax. separate rules by comma",
default_values_t = once_with(|| Pattern::new(".*").unwrap()),
value_parser = parse_glob,
value_delimiter = ',',
@ -90,6 +99,7 @@ pub struct Args {
#[arg(
short='X', long,
help = "exclude from search and printing",
long_help = "exclude from search and printin. accepts glob syntax. separate rules by comma",
value_parser = parse_glob,
value_delimiter = ',',
action = ArgAction::Append,
@ -114,25 +124,27 @@ pub struct Args {
path: Vec<String>,
}
impl Args {
/// utility method to chuck default values on the end
/// utility method to chuck default values on the end.
/// it feels like I should be able to do this with
/// clever `clap` macros but I don't know how
pub fn post_process(mut self) -> Self {
if self.base_two {
self.unit = Unit::Kibi;
} else if self.si {
self.unit = Unit::Kilo;
}
if self.show_hidden {
self.exclude_print = Vec::new();
}
if self.path.is_empty() {
self.path = vec![ ".".to_owned() ];
}
pub fn parse_and_process() -> Self {
let mut this = Self::parse();
self
if this.base_two {
this.unit = Unit::Kibi;
} else if this.si {
this.unit = Unit::Kilo;
}
if this.show_hidden {
this.exclude_print = Vec::new();
}
if this.path.is_empty() {
this.path = vec![ ".".to_owned() ];
}
this
}
pub fn should_exclude(&self, path: &Path, file: &Metadata) -> bool {
@ -149,16 +161,9 @@ impl Args {
any_pattern_matches_any_component(&self.exclude_search, path)
}
pub fn should_print(&self, path: &Path) -> bool {
! any_pattern_matches_any_component(&self.exclude_print, path)
// TODO: this exists because when a file matches an exclude pattern
// is it still returned, just with no size or children, so in order
// to not accidentally print things that we said we were excluding,
// we also have to check that it's not excluded by search.
// `self.exclude_print.extend(&self.exclude_search)` is wasteful,
// but until I find a better way this is what it's gotta be`
&& ! any_pattern_matches_any_component(&self.exclude_search, path)
}
pub const fn persistant(&self) -> bool {
@ -168,15 +173,15 @@ impl Args {
pub const fn minimal(&self) -> bool {
self.minimal
}
pub const fn tree(&self) -> bool {
self.tree
}
pub const fn total(&self) -> bool {
self.total
}
pub const fn unit(&self) -> Unit {
self.unit
}
@ -184,6 +189,10 @@ impl Args {
pub fn iter(&self) -> Iter<'_, String> {
self.path.iter()
}
pub fn quiet(&self) -> bool {
self.quiet
}
}
fn validate_path(s: &str) -> Result<String, String> {
@ -210,7 +219,8 @@ fn any_pattern_matches_any_component(patterns: &[Pattern], path: &Path) -> bool
let Some(s) = cmp.to_str() else {
// this is a code smell
// I don't believe it, but I can't think
// of anything worthwhile to do
// of anything worthwhile to do when
// you can't get a usable &str
continue
};
if pat.matches(s) {

135
src/directory.rs
View File

@ -30,12 +30,11 @@ impl Directory {
let path = path.as_ref();
// NOTE: I go back and forth on canonicalize()ing all the time.
// I feel like it changes every commit. The performance loss seems
// to be negligible, even when I do crazy things like `hb -p /`, which
// is the most I can currently do.
let name = match (path.canonicalize(), args.persistant()) {
(Ok(path), _) => path,
(Err(_), true) => return Ok(None),
(Err(e), false) => return Err(e.into()),
// to be negligible, even when I do crazy things like `hb -p /`
let name = match path.canonicalize() {
Ok(path) => path,
Err(_) if args.persistant() => return Ok(None),
Err(e) => return Err(e.into()),
}
.file_name()
.map_or_else(|| OsString::from("/"), ToOwned::to_owned)
@ -43,24 +42,17 @@ impl Directory {
// symlink_metadata() is the same as metadata() but it doesn't
// traverse symlinks, so that we can exclude them if necessary
let meta = match (path.symlink_metadata(), args.persistant()) {
(Ok(md), _) => md,
(Err(_), true) => return Ok(None),
(Err(e), false) => return Err(e.into()),
let meta = match path.symlink_metadata() {
Ok(md) => md,
Err(_) if args.persistant() => return Ok(None),
Err(e) => return Err(e.into()),
};
if args.should_exclude(path, &meta) {
// Ok(None) is only meant to arise from an error
// while persistant. When that happens, a no-op
// entry is substituted, which is precisely
// we want to happen when we hit an excluded file.
// NOTE: return Ok(None) is *not* equivalent, because
// that's only produced when an error occurs but
// the program is running in persistant mode. I used
// to return that, but that causes incredibly
// bizarre and wrong behaviour.
return Ok(Some( Self { name, size: 0, children: Vec::new() } ))
// finding a file to exclude is behaviourally
// equivalent to hitting an error in persistant
// mode: just continue
return Ok(None)
}
let dir = match read_dir(path) {
@ -85,41 +77,47 @@ impl Directory {
// this is a compicated iterator pattern. I'll do my best to explain.
// 1. the end result is that we `reduce()` the iterator to a single
// (u64, Vec<Directory>) tuple to return. this is done by...
let (size, children) =
// 2. taking the iterator over the directory and parallelising it...
dir.par_bridge()
// 3, this is the recursive step: try to create new Directory
// objects from each item in the iterator
.map(|entry| Self::new(entry?.path(), args))
// 4. the fold (this is try_fold because we're iterating over Result.).
// each fold adds a directory as a child and increases the total size
.try_fold(
|| (0, Vec::new()),
|(mut size, mut children), dir| -> Result<(u64, Vec<Self>)> {
let Some(dir) = Result::from(dir)?
else {
// some intermediate operation failed, but we
// are persistant, so just skip
return Result::Ok((0, Vec::new()))
};
size += dir.size;
let (size, children) = match
// 2. taking the iterator over the directory and parallelising it...
dir.par_bridge()
// 3, this is the recursive step: try to create new Directory
// objects from each item in the iterator
.map(|entry| Self::new(entry?.path(), args))
// 4. the fold (this is try_fold because we're iterating over Result.).
// each fold adds a directory as a child and increases the total size
.try_fold(
|| (0, Vec::new()),
|(mut size, mut children), dir| -> Result<(u64, Vec<Self>)> {
let dir = match (dir, args.persistant()) {
(Ok(Some(d)), _) => d,
(Ok(None), _) | (Err(_), true) => return Result::Ok((size, children)),
(Err(e), false) => return Err(e),
};
size += dir.size;
if args.should_print(dir.path()) {
// since size was increased, this just prevents
// the directory from appearing in printing
children.push(dir);
// have to specify anyhow::Result::Ok otherwise it complains
// that it can't infer the E in Result<T, E>
Result::Ok((size, children))
}
)
// 5. the final step is to reduce, which is as simple as concatenating
// every vector and summing up their sizes.
.try_reduce(
|| (0, Vec::new()),
|(asize, mut avec), (bsize, bvec)| {
avec.extend(bvec);
Result::Ok((asize + bsize, avec))
}
)?;
// ^ note the Try, because of course any of these operations could
// fail
// have to specify anyhow::Result::Ok otherwise it complains
// that it can't infer the E in Result<T, E>
Result::Ok((size, children))
}
)
// 5. the final step is to reduce, which is as simple as concatenating
// every vector and summing up their sizes.
.try_reduce(
|| (0, Vec::new()),
|(asize, mut avec), (bsize, bvec)| {
avec.extend(bvec);
Result::Ok((asize + bsize, avec))
}
) {
// remember that this is a match statement?
Ok(tuple) => tuple,
Err(_) if args.persistant() => return Ok(None),
Err(e) => return Err(e),
};
// final notes:
// 1. I am unsure if it is better to do a bunch of partial sums
@ -142,11 +140,11 @@ impl Directory {
))
}
pub fn tree(self, args: &Args) -> String {
pub fn tree(self, unit: Unit) -> String {
// since self.size is definitionally the greatest value, the tab length
// is just the length of self.len, plus two for a tab width
let tab_size = self.size.to_string().len() + 2;
self.vectorise(args)
let tab_size = unit.convert(self.size).len() + 2;
self.vectorise(unit)
.iter()
.map(|e| e.stringify_tabbed(tab_size))
.reduce(|s1, s2| s1 + "\n" + &s2)
@ -155,26 +153,32 @@ impl Directory {
/// TODO: make not recursive, take &self if possible,
/// and maybe write directly to stdout to not use so much mem
fn vectorise(mut self, args: &Args) -> Vec<TreeEntry> {
fn vectorise(self, unit: Unit) -> Vec<TreeEntry> {
let mut result = Vec::new();
result.push(TreeEntry::new(
self.name.display().to_string(), self.size, args.unit()
self.name.display().to_string(), self.size, unit
));
let mut new_entry_part = TreePart::First;
let mut continue_part = TreePart::Wait;
self.children.retain(|dir| args.should_print(dir.path()));
let len = self.children.len();
// this is the display algorithm. it's built on the variables
// `new_entry_part` and `continue_part`. for most times, when
// we introduce a new item (which happens every iteration of
// the loop), it is `first` tree part and we can pad with the
// `wait` part. the last element of each one should however
// be introduced with a `last` part, and padding should with
// `blank`
for (idx, child) in self.children.into_iter().enumerate() {
if idx+1 == len {
new_entry_part = TreePart::Last;
continue_part = TreePart::Blank;
}
let subtree = child.vectorise(args);
let subtree = child.vectorise(unit);
for mut item in subtree {
if item.parts.is_empty() {
@ -203,14 +207,13 @@ impl TreeEntry {
parts: Vec::new(), path, size, unit
}
}
fn stringify_tabbed(&self, tab_size: usize) -> String {
let mut result = format!("{:<tab_size$}", self.unit.convert(self.size));
for part in self.parts.iter().rev() {
result += part.display();
}
// dont add the space to empty entries
result += " ";
result += &self.path;
@ -220,9 +223,13 @@ impl TreeEntry {
#[derive(PartialEq, Eq, Debug, Clone, Copy)]
enum TreePart {
/// `├──`
First,
/// `│ `
Wait,
/// `└──`
Last,
/// (blank)
Blank
}
impl TreePart {

12
src/main.rs
View File

@ -7,13 +7,10 @@ mod unit;
use args::Args;
use directory::Directory;
use clap::Parser;
use std::process::ExitCode;
fn main() -> ExitCode {
let args = Args::parse().post_process();
// dbg!(&args);
let args = Args::parse_and_process();
let mut total = 0;
for path in args.iter() {
@ -23,10 +20,7 @@ fn main() -> ExitCode {
// so we don't need a match guard
Ok(None) => continue,
Err(e) => {
if args.persistant() {
continue;
}
if !args.minimal() {
if !args.minimal() && !args.quiet() {
eprintln!("hb: {e}");
}
return ExitCode::FAILURE;
@ -41,7 +35,7 @@ fn main() -> ExitCode {
}
if args.tree() {
println!("{}", dir_structure.tree(&args));
println!("{}", dir_structure.tree(args.unit()));
} else {
println!(
"{}: {}",