alex/src/server/backups.rs

use flate2::write::GzEncoder;
use flate2::Compression;
use std::fs::File;
use std::io;
use std::path::{Path, PathBuf};
use chrono::{Utc, Local};
use std::collections::HashSet;

#[link(name = "c")]
extern "C" {
    fn geteuid() -> u32;
    fn getegid() -> u32;
}

static FILENAME_FORMAT: &str = "%Y-%m-%d_%H-%M-%S.tar.gz";

pub struct BackupManager {
    backup_dir: PathBuf,
    config_dir: PathBuf,
    world_dir: PathBuf,
    max_backups: u64,
    /// Start time of the last successful backup
    last_start_time: Option<chrono::DateTime<chrono::Utc>>,
    /// Files contained in the last successful backup
    last_files: HashSet<(PathBuf, PathBuf)>
}

impl BackupManager {
    pub fn open(
        backup_dir: PathBuf,
        config_dir: PathBuf,
        world_dir: PathBuf,
        max_backups: u64,
    ) -> Self {
        BackupManager {
            backup_dir,
            config_dir,
            world_dir,
            max_backups,
            last_start_time: None,
            last_files: HashSet::new()
        }
    }

    fn files_to_backup(&mut self) -> io::Result<HashSet<(PathBuf, PathBuf)>> {
        let mut dirs = vec![
            (PathBuf::from("worlds"), self.world_dir.clone()),
            (PathBuf::from("config"), self.config_dir.clone()),
        ];
        let mut files: HashSet<(PathBuf, PathBuf)> = HashSet::new();

        while let Some((path_in_tar, path)) = dirs.pop() {
            for res in path.read_dir()? {
                let entry = res?;

                if entry.file_name() == "cache" {
                    continue;
                }

                let new_path_in_tar = path_in_tar.join(entry.file_name());

                // All dirs get expanded recursively, while all files get returned as output
                // NOTE: does this remove empty directories from backups? Is this a problem?
                if entry.file_type()?.is_dir() {
                    dirs.push((new_path_in_tar, entry.path()));
                } else {
                    // Only add files that have been updated since the last backup (incremental backup)
                    if let Some(last_start_time) = self.last_start_time {
                        let last_modified = entry.path().metadata()?.modified();

                        if let Ok(last_modified) = last_modified {
                            let t: chrono::DateTime<Utc> = last_modified.into();
                            let t = t.with_timezone(&Local);

                            if t < last_start_time {
                                continue
                            }
                        }
                    }

                    files.insert((new_path_in_tar, entry.path()));
                }
            }
        }

        Ok(files)
    }

    pub fn create_archive(&mut self) -> io::Result<()> {
        let start_time = chrono::offset::Utc::now();

        let filename = format!("{}", start_time.format(FILENAME_FORMAT));
        let path = self.backup_dir.join(filename);
        let tar_gz = File::create(path)?;
        let enc = GzEncoder::new(tar_gz, Compression::default());
        let mut ar = tar::Builder::new(enc);

        let files = self.files_to_backup()?;

        for (path_in_tar, path) in &files {
            ar.append_path_with_name(path, path_in_tar)?;
        }

        let deleted_files = self.last_files.difference(&files);

        println!("{} {}", files.len(), self.last_files.len());

        for (path_in_tar, path) in deleted_files {
            println!("{path_in_tar:?}: {path:?}");
        }

        // TODO re-add this info file in some way
        // We add a file to the backup describing for what version it was made
        // let info = format!("{} {}", self.type_, self.version);
        // let info_bytes = info.as_bytes();

        // let mut header = tar::Header::new_gnu();
        // header.set_size(info_bytes.len().try_into().unwrap());
        // header.set_mode(0o100644);
        // unsafe {
        //     header.set_gid(getegid().into());
        //     header.set_uid(geteuid().into());
        // }

        // tar.append_data(&mut header, "info.txt", info_bytes)?;
        
        // After a successful backup, we store the original metadata
        self.last_start_time = Some(start_time);
        self.last_files = files;

        Ok(())
    }

    /// Remove the oldest backups
    pub fn remove_old_backups(&mut self) -> std::io::Result<()> {
        // The naming format used allows us to sort the backups by name and still get a sorting by
        // creation time
        let mut backups = self
            .backup_dir
            .read_dir()?
            .filter_map(|res| res.map(|e| e.path()).ok())
            .collect::<Vec<PathBuf>>();
        backups.sort();

        let max_backups: usize = self.max_backups.try_into().unwrap();

        if backups.len() > max_backups {
            let excess_backups = backups.len() - max_backups;

            for backup in &backups[0..excess_backups] {
                std::fs::remove_file(backup)?;
            }
        }

        Ok(())
    }
}