use std::collections::{BTreeSet, HashMap, HashSet}; use std::fs::{DirEntry, File, Metadata}; use std::io::Write; use std::io::*; use std::mem; use std::panic; use std::path::{Path, PathBuf}; use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering}; use std::sync::Arc; use std::time::SystemTime; use bk_tree::BKTree; use crossbeam_channel::Receiver; use futures::channel::mpsc::UnboundedSender; use humansize::format_size; use humansize::BINARY; use image::GenericImageView; use image_hasher::{FilterType, HashAlg, HasherConfig}; use rayon::prelude::*; use serde::{Deserialize, Serialize}; #[cfg(feature = "heif")] use crate::common::get_dynamic_image_from_heic; use crate::common::{ check_folder_children, create_crash_message, get_dynamic_image_from_raw_image, get_number_of_threads, open_cache_folder, prepare_thread_handler_common, send_info_and_wait_for_ending_all_threads, Common, HEIC_EXTENSIONS, IMAGE_RS_SIMILAR_IMAGES_EXTENSIONS, RAW_IMAGE_EXTENSIONS, }; use crate::common_dir_traversal::{common_get_entry_data_metadata, common_read_dir, get_lowercase_name, get_modified_time, CheckingMethod, ProgressData}; use crate::common_directory::Directories; use crate::common_extensions::Extensions; use crate::common_items::ExcludedItems; use crate::common_messages::Messages; use crate::common_traits::{DebugPrint, PrintResults, SaveResults}; use crate::flc; type ImHash = Vec; pub const SIMILAR_VALUES: [[u32; 6]; 4] = [ [1, 2, 5, 7, 14, 20], // 8 [2, 5, 15, 30, 40, 40], // 16 [4, 10, 20, 40, 40, 40], // 32 [6, 20, 40, 40, 40, 40], // 64 ]; #[derive(Clone, Debug, Serialize, Deserialize)] pub struct FileEntry { pub path: PathBuf, pub size: u64, pub dimensions: String, pub modified_date: u64, pub hash: ImHash, pub similarity: u32, } /// Used by CLI tool when we cannot use directly values #[derive(Clone, Debug, Copy)] pub enum SimilarityPreset { Original, VeryHigh, High, Medium, Small, VerySmall, Minimal, None, } /// Distance metric to use with the BK-tree. struct Hamming; impl bk_tree::Metric for Hamming { fn distance(&self, a: &ImHash, b: &ImHash) -> u32 { hamming::distance_fast(a, b).unwrap() as u32 } fn threshold_distance(&self, a: &ImHash, b: &ImHash, _threshold: u32) -> Option { Some(self.distance(a, b)) } } /// Struct to store most basics info about all folder pub struct SimilarImages { information: Info, text_messages: Messages, directories: Directories, allowed_extensions: Extensions, excluded_items: ExcludedItems, bktree: BKTree, similar_vectors: Vec>, similar_referenced_vectors: Vec<(FileEntry, Vec)>, recursive_search: bool, minimal_file_size: u64, maximal_file_size: u64, image_hashes: HashMap>, // Hashmap with image hashes and Vector with names of files stopped_search: bool, similarity: u32, images_to_check: HashMap, hash_size: u8, hash_alg: HashAlg, image_filter: FilterType, use_cache: bool, delete_outdated_cache: bool, exclude_images_with_same_size: bool, use_reference_folders: bool, save_also_as_json: bool, } /// Info struck with helpful information's about results #[derive(Default)] pub struct Info { pub number_of_duplicates: usize, pub number_of_groups: u64, } impl Info { #[must_use] pub fn new() -> Self { Default::default() } } /// Method implementation for `EmptyFolder` impl SimilarImages { /// New function providing basics values #[must_use] pub fn new() -> Self { Self { information: Default::default(), text_messages: Messages::new(), directories: Directories::new(), excluded_items: Default::default(), allowed_extensions: Extensions::new(), bktree: BKTree::new(Hamming), similar_vectors: vec![], similar_referenced_vectors: Default::default(), recursive_search: true, minimal_file_size: 1024 * 16, // 16 KB should be enough to exclude too small images from search maximal_file_size: u64::MAX, image_hashes: Default::default(), stopped_search: false, similarity: 0, images_to_check: Default::default(), hash_size: 8, hash_alg: HashAlg::Gradient, image_filter: FilterType::Lanczos3, use_cache: true, delete_outdated_cache: true, exclude_images_with_same_size: false, use_reference_folders: false, save_also_as_json: false, } } pub fn set_hash_size(&mut self, hash_size: u8) { self.hash_size = match hash_size { 8 | 16 | 32 | 64 => hash_size, e => { panic!("Invalid value of hash size {e}"); } } } pub fn set_delete_outdated_cache(&mut self, delete_outdated_cache: bool) { self.delete_outdated_cache = delete_outdated_cache; } pub fn set_exclude_images_with_same_size(&mut self, exclude_images_with_same_size: bool) { self.exclude_images_with_same_size = exclude_images_with_same_size; } pub fn set_hash_alg(&mut self, hash_alg: HashAlg) { self.hash_alg = hash_alg; } pub fn set_image_filter(&mut self, image_filter: FilterType) { self.image_filter = image_filter; } pub fn set_save_also_as_json(&mut self, save_also_as_json: bool) { self.save_also_as_json = save_also_as_json; } #[must_use] pub fn get_stopped_search(&self) -> bool { self.stopped_search } #[must_use] pub const fn get_text_messages(&self) -> &Messages { &self.text_messages } #[must_use] pub const fn get_similar_images(&self) -> &Vec> { &self.similar_vectors } #[must_use] pub fn get_similar_images_referenced(&self) -> &Vec<(FileEntry, Vec)> { &self.similar_referenced_vectors } #[must_use] pub fn get_use_reference(&self) -> bool { self.use_reference_folders } #[must_use] pub const fn get_information(&self) -> &Info { &self.information } pub fn set_use_cache(&mut self, use_cache: bool) { self.use_cache = use_cache; } pub fn set_recursive_search(&mut self, recursive_search: bool) { self.recursive_search = recursive_search; } #[cfg(target_family = "unix")] pub fn set_exclude_other_filesystems(&mut self, exclude_other_filesystems: bool) { self.directories.set_exclude_other_filesystems(exclude_other_filesystems); } #[cfg(not(target_family = "unix"))] pub fn set_exclude_other_filesystems(&mut self, _exclude_other_filesystems: bool) {} pub fn set_allowed_extensions(&mut self, allowed_extensions: String) { self.allowed_extensions.set_allowed_extensions(allowed_extensions, &mut self.text_messages); } pub fn set_minimal_file_size(&mut self, minimal_file_size: u64) { self.minimal_file_size = match minimal_file_size { 0 => 1, t => t, }; } pub fn set_maximal_file_size(&mut self, maximal_file_size: u64) { self.maximal_file_size = match maximal_file_size { 0 => 1, t => t, }; } pub fn set_similarity(&mut self, similarity: u32) { self.similarity = similarity; } /// Public function used by CLI to search for empty folders pub fn find_similar_images(&mut self, stop_receiver: Option<&Receiver<()>>, progress_sender: Option<&UnboundedSender>) { self.directories.optimize_directories(true, &mut self.text_messages); self.use_reference_folders = !self.directories.reference_directories.is_empty(); if !self.check_for_similar_images(stop_receiver, progress_sender) { self.stopped_search = true; return; } if !self.hash_images(stop_receiver, progress_sender) { self.stopped_search = true; return; } if !self.find_similar_hashes(stop_receiver, progress_sender) { self.stopped_search = true; return; } // if self.delete_folders { // self.delete_empty_folders(); // } self.debug_print(); } // pub fn set_delete_folder(&mut self, delete_folder: bool) { // self.delete_folders = delete_folder; // } /// Function to check if folder are empty. /// Parameter `initial_checking` for second check before deleting to be sure that checked folder is still empty fn check_for_similar_images(&mut self, stop_receiver: Option<&Receiver<()>>, progress_sender: Option<&UnboundedSender>) -> bool { let start_time: SystemTime = SystemTime::now(); let mut folders_to_check: Vec = Vec::with_capacity(1024 * 2); // This should be small enough too not see to big difference and big enough to store most of paths without needing to resize vector if !self.allowed_extensions.using_custom_extensions() { self.allowed_extensions.extend_allowed_extensions(IMAGE_RS_SIMILAR_IMAGES_EXTENSIONS); self.allowed_extensions.extend_allowed_extensions(RAW_IMAGE_EXTENSIONS); #[cfg(feature = "heif")] self.allowed_extensions.extend_allowed_extensions(HEIC_EXTENSIONS); } else { self.allowed_extensions .validate_allowed_extensions(&[IMAGE_RS_SIMILAR_IMAGES_EXTENSIONS, RAW_IMAGE_EXTENSIONS, HEIC_EXTENSIONS].concat()); if !self.allowed_extensions.using_custom_extensions() { return true; } } // Add root folders for finding for id in &self.directories.included_directories { folders_to_check.push(id.clone()); } let progress_thread_run = Arc::new(AtomicBool::new(true)); let atomic_counter = Arc::new(AtomicUsize::new(0)); let progress_thread_handle = prepare_thread_handler_common(progress_sender, &progress_thread_run, &atomic_counter, 0, 2, 0, CheckingMethod::None); while !folders_to_check.is_empty() { if stop_receiver.is_some() && stop_receiver.unwrap().try_recv().is_ok() { send_info_and_wait_for_ending_all_threads(&progress_thread_run, progress_thread_handle); return false; } let segments: Vec<_> = folders_to_check .par_iter() .map(|current_folder| { let mut dir_result = vec![]; let mut warnings = vec![]; let mut fe_result = vec![]; let Some(read_dir) = common_read_dir(current_folder, &mut warnings) else { return (dir_result, warnings, fe_result); }; for entry in read_dir { let Some((entry_data, metadata)) = common_get_entry_data_metadata(&entry, &mut warnings, current_folder) else { continue; }; if metadata.is_dir() { check_folder_children( &mut dir_result, &mut warnings, current_folder, entry_data, self.recursive_search, &self.directories, &self.excluded_items, ); } else if metadata.is_file() { atomic_counter.fetch_add(1, Ordering::Relaxed); self.add_file_entry(&metadata, current_folder, entry_data, &mut fe_result, &mut warnings); } } (dir_result, warnings, fe_result) }) .collect(); // Advance the frontier folders_to_check.clear(); // Process collected data for (segment, warnings, fe_result) in segments { folders_to_check.extend(segment); self.text_messages.warnings.extend(warnings); for (name, fe) in fe_result { self.images_to_check.insert(name, fe); } } } send_info_and_wait_for_ending_all_threads(&progress_thread_run, progress_thread_handle); Common::print_time(start_time, SystemTime::now(), "check_for_similar_images"); true } fn add_file_entry(&self, metadata: &Metadata, current_folder: &Path, entry_data: &DirEntry, fe_result: &mut Vec<(String, FileEntry)>, warnings: &mut Vec) { let Some(file_name_lowercase) = get_lowercase_name(entry_data, warnings) else { return; }; if !self.allowed_extensions.matches_filename(&file_name_lowercase) { return; } // Checking files if (self.minimal_file_size..=self.maximal_file_size).contains(&metadata.len()) { let current_file_name = current_folder.join(entry_data.file_name()); if self.excluded_items.is_excluded(¤t_file_name) { return; } let fe: FileEntry = FileEntry { path: current_file_name.clone(), size: metadata.len(), dimensions: String::new(), modified_date: get_modified_time(metadata, warnings, ¤t_file_name, false), hash: Vec::new(), similarity: 0, }; fe_result.push((current_file_name.to_string_lossy().to_string(), fe)); } } fn hash_images_load_cache(&mut self) -> (HashMap, HashMap, HashMap) { let loaded_hash_map; let mut records_already_cached: HashMap = Default::default(); let mut non_cached_files_to_check: HashMap = Default::default(); if self.use_cache { loaded_hash_map = match load_hashes_from_file(&mut self.text_messages, self.delete_outdated_cache, self.hash_size, self.hash_alg, self.image_filter) { Some(t) => t, None => Default::default(), }; for (name, file_entry) in &self.images_to_check { #[allow(clippy::if_same_then_else)] if !loaded_hash_map.contains_key(name) { // If loaded data doesn't contains current image info non_cached_files_to_check.insert(name.clone(), file_entry.clone()); } else if file_entry.size != loaded_hash_map.get(name).unwrap().size || file_entry.modified_date != loaded_hash_map.get(name).unwrap().modified_date { // When size or modification date of image changed, then it is clear that is different image non_cached_files_to_check.insert(name.clone(), file_entry.clone()); } else { // Checking may be omitted when already there is entry with same size and modification date records_already_cached.insert(name.clone(), loaded_hash_map.get(name).unwrap().clone()); } } } else { loaded_hash_map = Default::default(); mem::swap(&mut self.images_to_check, &mut non_cached_files_to_check); } (loaded_hash_map, records_already_cached, non_cached_files_to_check) } // Cache algorithm: // - Load data from file // - Remove from data to search, already loaded entries from cache(size and modified datamust match) // - Check hash of files which doesn't have saved entry // - Join already read hashes with hashes which were read from file // - Join all hashes and save it to file fn hash_images(&mut self, stop_receiver: Option<&Receiver<()>>, progress_sender: Option<&UnboundedSender>) -> bool { let hash_map_modification = SystemTime::now(); let (loaded_hash_map, records_already_cached, non_cached_files_to_check) = self.hash_images_load_cache(); Common::print_time(hash_map_modification, SystemTime::now(), "sort_images - reading data from cache and preparing them"); let hash_map_modification = SystemTime::now(); let check_was_stopped = AtomicBool::new(false); // Used for breaking from GUI and ending check thread let progress_thread_run = Arc::new(AtomicBool::new(true)); let atomic_counter = Arc::new(AtomicUsize::new(0)); let progress_thread_handle = prepare_thread_handler_common( progress_sender, &progress_thread_run, &atomic_counter, 1, 2, non_cached_files_to_check.len(), CheckingMethod::None, ); let mut vec_file_entry: Vec<(FileEntry, ImHash)> = non_cached_files_to_check .into_par_iter() .map(|(_s, file_entry)| { atomic_counter.fetch_add(1, Ordering::Relaxed); if stop_receiver.is_some() && stop_receiver.unwrap().try_recv().is_ok() { check_was_stopped.store(true, Ordering::Relaxed); return None; } Some(Some(self.collect_image_file_entry(file_entry))) }) .while_some() .filter(Option::is_some) .map(Option::unwrap) .collect::>(); send_info_and_wait_for_ending_all_threads(&progress_thread_run, progress_thread_handle); Common::print_time(hash_map_modification, SystemTime::now(), "sort_images - reading data from files in parallel"); let hash_map_modification = SystemTime::now(); // Just connect loaded results with already calculated hashes for (_name, file_entry) in records_already_cached { vec_file_entry.push((file_entry.clone(), file_entry.hash)); } // All valid entries are used to create bktree used to check for hash similarity for (file_entry, buf) in &vec_file_entry { // Only use to comparing, non broken hashes(all 0 or 255 hashes means that algorithm fails to decode them because e.g. contains a log of alpha channel) if !(buf.is_empty() || buf.iter().all(|e| *e == 0) || buf.iter().all(|e| *e == 255)) { self.image_hashes.entry(buf.clone()).or_insert_with(Vec::::new).push(file_entry.clone()); } } if self.use_cache { // Must save all results to file, old loaded from file with all currently counted results let mut all_results: HashMap = loaded_hash_map; for (file_entry, _hash) in vec_file_entry { all_results.insert(file_entry.path.to_string_lossy().to_string(), file_entry); } save_hashes_to_file( &all_results, &mut self.text_messages, self.save_also_as_json, self.hash_size, self.hash_alg, self.image_filter, ); } // Break if stop was clicked after saving to cache if check_was_stopped.load(Ordering::Relaxed) { return false; } Common::print_time(hash_map_modification, SystemTime::now(), "sort_images - saving data to files"); true } fn collect_image_file_entry(&self, mut file_entry: FileEntry) -> (FileEntry, ImHash) { let file_name_lowercase = file_entry.path.to_string_lossy().to_lowercase(); let image; #[allow(clippy::never_loop)] // Required to implement nice if/else 'krztyna: loop { if RAW_IMAGE_EXTENSIONS.iter().any(|e| file_name_lowercase.ends_with(e)) { image = match get_dynamic_image_from_raw_image(&file_entry.path) { Some(t) => t, None => return (file_entry, Vec::new()), }; break 'krztyna; } #[cfg(feature = "heif")] if HEIC_EXTENSIONS.iter().any(|e| file_name_lowercase.ends_with(e)) { image = match get_dynamic_image_from_heic(&file_entry.path.to_string_lossy()) { Ok(t) => t, Err(_) => { return (file_entry, Vec::new()); } }; break 'krztyna; } // Normal image extension, when any other fail, not using if/else let result = panic::catch_unwind(|| { match image::open(file_entry.path.clone()) { Ok(t) => Ok(t), // Err(_inspected) => return Some(None), // Something is wrong with image, // For broken images empty hash is used, because without it will try to resecan files each time when it is called(missing cache file is responsible for it) // This may cause problems(very rarely), when e.g. file was not available due lack of permissions, but it is available now Err(_inspected) => Err(()), } }); // If image crashed during opening, we just skip checking its hash and go on if let Ok(image_result) = result { if let Ok(image2) = image_result { image = image2; } else { return (file_entry, Vec::new()); } } else { let message = create_crash_message("Image-rs", &file_entry.path.to_string_lossy(), "https://github.com/image-rs/image/issues"); println!("{message}"); return (file_entry, Vec::new()); } break 'krztyna; } let dimensions = image.dimensions(); file_entry.dimensions = format!("{}x{}", dimensions.0, dimensions.1); let hasher_config = HasherConfig::new() .hash_size(self.hash_size as u32, self.hash_size as u32) .hash_alg(self.hash_alg) .resize_filter(self.image_filter); let hasher = hasher_config.to_hasher(); let hash = hasher.hash_image(&image); let buf: ImHash = hash.as_bytes().to_vec(); file_entry.hash = buf.clone(); (file_entry, buf) } fn compare_hashes( &self, hashes_to_check: &[ImHash], atomic_counter: &Arc, stop_receiver: Option<&Receiver<()>>, check_was_stopped: &AtomicBool, tolerance: u32, hashes_with_multiple_images: &HashSet, all_hashed_images: &HashMap>, ) -> Option<(HashMap, HashMap)> { let mut hashes_parents: HashMap = Default::default(); // Hashes used as parent (hash, children_number_of_hash) let mut hashes_similarity: HashMap = Default::default(); // Hashes used as child, (parent_hash, similarity) // Sprawdź czy hash nie jest użyty jako master gdzie indziej // Jeśli tak to przejdź do sprawdzania kolejnego elementu // Zweryfikuj czy sprawdzany element ma rodzica // Jeśli ma to sprawdź czy similarity nowego rodzica jest mniejsze niż starego // // Jeśli tak to zmniejsz ilość dzieci starego rodzica, dodaj ilość dzieci w nowym rodzicu i podmień rekord hashes_similarity // // Jeśli nie to dodaj nowy rekord w hashes_similarity jak i hashes_parents z liczbą dzieci równą 1 for (index, hash_to_check) in hashes_to_check.iter().enumerate() { // Don't check for user stop too often // Also don't add too often data to atomic variable const CYCLES_COUNTER: usize = 0b11_1111; if ((index & CYCLES_COUNTER) == CYCLES_COUNTER) && index != 0 { atomic_counter.fetch_add(CYCLES_COUNTER, Ordering::Relaxed); if stop_receiver.is_some() && stop_receiver.unwrap().try_recv().is_ok() { check_was_stopped.store(true, Ordering::Relaxed); return None; } } hashes_parents.insert((*hash_to_check).clone(), 0); let mut found_items = self .bktree .find(hash_to_check, tolerance) .filter(|(similarity, _hash)| if self.use_reference_folders { true } else { *similarity != 0 }) .collect::>(); found_items.sort_unstable_by_key(|f| f.0); for (similarity, compared_hash) in found_items { image_to_check( &mut hashes_parents, &mut hashes_similarity, hashes_with_multiple_images, hash_to_check, compared_hash, similarity, ); } } debug_check_for_duplicated_things(self.use_reference_folders, &hashes_parents, &hashes_similarity, all_hashed_images, "BEFORE"); Some((hashes_parents, hashes_similarity)) } fn chunk_hashes(&mut self, all_hashed_images: &HashMap>, all_hashes: &Vec) -> (Vec>, HashSet) { let mut hashes_with_multiple_images: HashSet = Default::default(); // Fast way to check if hash have multiple images let mut files_from_referenced_folders: HashMap> = HashMap::new(); let mut normal_files: HashMap> = HashMap::new(); let number_of_processors = get_number_of_threads(); let chunk_size; let mut initial_hashes: Vec = Vec::new(); let mut additional_chunk_to_check: Vec = Default::default(); let mut chunks: Vec>; if self.use_reference_folders { let reference_directories = self.directories.reference_directories.clone(); all_hashed_images.clone().into_iter().for_each(|(hash, vec_file_entry)| { for file_entry in vec_file_entry { if reference_directories.iter().any(|e| file_entry.path.starts_with(e)) { files_from_referenced_folders.entry(hash.clone()).or_insert_with(Vec::new).push(file_entry); } else { normal_files.entry(hash.clone()).or_insert_with(Vec::new).push(file_entry); } } }); for (hash, vec_files) in normal_files { if vec_files.len() >= 2 { hashes_with_multiple_images.insert(hash.clone()); } self.bktree.add(hash); } for (hash, vec_files) in files_from_referenced_folders { if vec_files.len() >= 2 { hashes_with_multiple_images.insert(hash.clone()); } initial_hashes.push(hash); } chunk_size = initial_hashes.len() / number_of_processors; chunks = if chunk_size > 0 { initial_hashes.chunks(chunk_size).map(<[std::vec::Vec]>::to_vec).collect::>() } else { vec![initial_hashes] }; } else { for (hash, vec_files) in all_hashed_images { if vec_files.len() >= 2 { additional_chunk_to_check.push(hash.clone()); hashes_with_multiple_images.insert(hash.clone()); } else { self.bktree.add(hash.clone()); } } chunk_size = all_hashes.len() / number_of_processors; chunks = if chunk_size > 0 { all_hashes.chunks(chunk_size).map(<[Vec]>::to_vec).collect::>() } else { vec![all_hashes.clone()] }; chunks.push(additional_chunk_to_check); } (chunks, hashes_with_multiple_images) } fn collect_hash_compare_result( &self, hashes_parents: HashMap, hashes_with_multiple_images: &HashSet, all_hashed_images: &HashMap>, collected_similar_images: &mut HashMap>, hashes_similarity: HashMap, ) { if self.use_reference_folders { // This is same step as without reference folders, but also checks if children are inside/outside reference directories, because may happen, that one file is inside reference folder and other outside // Collecting results to vector for (parent_hash, child_number) in hashes_parents { // If hash contains other hasher OR multiple images are available for checked hash if child_number > 0 || hashes_with_multiple_images.contains(&parent_hash) { let vec_fe = all_hashed_images .get(&parent_hash) .unwrap() .iter() .filter(|e| is_in_reference_folder(&self.directories.reference_directories, &e.path)) .cloned() .collect(); collected_similar_images.insert(parent_hash.clone(), vec_fe); } } for (child_hash, (parent_hash, similarity)) in hashes_similarity { let mut vec_fe: Vec<_> = all_hashed_images .get(&child_hash) .unwrap() .iter() .filter(|e| !is_in_reference_folder(&self.directories.reference_directories, &e.path)) .cloned() .collect(); for mut fe in &mut vec_fe { fe.similarity = similarity; } collected_similar_images.get_mut(&parent_hash).unwrap().append(&mut vec_fe); } } else { // Collecting results to vector for (parent_hash, child_number) in hashes_parents { // If hash contains other hasher OR multiple images are available for checked hash if child_number > 0 || hashes_with_multiple_images.contains(&parent_hash) { let vec_fe = all_hashed_images.get(&parent_hash).unwrap().clone(); collected_similar_images.insert(parent_hash.clone(), vec_fe); } } for (child_hash, (parent_hash, similarity)) in hashes_similarity { let mut vec_fe = all_hashed_images.get(&child_hash).unwrap().clone(); for mut fe in &mut vec_fe { fe.similarity = similarity; } collected_similar_images.get_mut(&parent_hash).unwrap().append(&mut vec_fe); } } } fn check_for_duplicate_hashes( &self, parts: Vec<(HashMap, HashMap)>, hashes_with_multiple_images: &HashSet, all_hashed_images: &HashMap>, collected_similar_images: &mut HashMap>, ) { let mut hashes_parents: HashMap = Default::default(); let mut hashes_similarity: HashMap = Default::default(); let mut iter = parts.into_iter(); // At start fill arrays with first item // Normal algorithm would do exactly same thing, but slower, one record after one if let Some((first_hashes_parents, first_hashes_similarity)) = iter.next() { hashes_parents = first_hashes_parents; hashes_similarity = first_hashes_similarity; } for (partial_hashes_with_parents, partial_hashes_with_similarity) in iter { for (parent_hash, _child_number) in partial_hashes_with_parents { if !hashes_parents.contains_key(&parent_hash) && !hashes_similarity.contains_key(&parent_hash) { hashes_parents.insert(parent_hash, 0); } } for (hash_to_check, (compared_hash, similarity)) in partial_hashes_with_similarity { image_to_check( &mut hashes_parents, &mut hashes_similarity, hashes_with_multiple_images, &hash_to_check, &compared_hash, similarity, ); } } debug_check_for_duplicated_things(self.use_reference_folders, &hashes_parents, &hashes_similarity, all_hashed_images, "LATTER"); // Just simple check if all original hashes with multiple entries are available in end results let original_hashes_at_start = hashes_with_multiple_images.len(); let original_hashes_in_end_results = hashes_parents .iter() .filter(|(parent_hash, _child_number)| hashes_with_multiple_images.contains(*parent_hash)) .count(); if !self.use_reference_folders { assert_eq!(original_hashes_at_start, original_hashes_in_end_results); } self.collect_hash_compare_result(hashes_parents, hashes_with_multiple_images, all_hashed_images, collected_similar_images, hashes_similarity); } fn find_similar_hashes(&mut self, stop_receiver: Option<&Receiver<()>>, progress_sender: Option<&UnboundedSender>) -> bool { if self.image_hashes.is_empty() { return true; } let hash_map_modification = SystemTime::now(); let tolerance = self.similarity; // Results let mut collected_similar_images: HashMap> = Default::default(); let mut all_hashed_images = Default::default(); mem::swap(&mut all_hashed_images, &mut self.image_hashes); let all_hashes: Vec<_> = all_hashed_images.clone().into_keys().collect(); // Checking entries with tolerance 0 is really easy and fast, because only entries with same hashes needs to be checked if tolerance == 0 { for (hash, vec_file_entry) in all_hashed_images.clone() { if vec_file_entry.len() >= 2 { collected_similar_images.insert(hash, vec_file_entry); } } } else { let check_was_stopped = AtomicBool::new(false); // Used for breaking from GUI and ending check thread let progress_thread_run = Arc::new(AtomicBool::new(true)); let atomic_counter = Arc::new(AtomicUsize::new(0)); let progress_thread_handle = prepare_thread_handler_common(progress_sender, &progress_thread_run, &atomic_counter, 2, 2, all_hashes.len(), CheckingMethod::None); // Don't use hashes with multiple images in bktree, because they will always be master of group and cannot be find by other hashes let (chunks, hashes_with_multiple_images) = self.chunk_hashes(&all_hashed_images, &all_hashes); let parts: Vec<_> = chunks .into_par_iter() .map(|hashes_to_check| { self.compare_hashes( &hashes_to_check, &atomic_counter, stop_receiver, &check_was_stopped, tolerance, &hashes_with_multiple_images, &all_hashed_images, ) }) .while_some() .collect(); send_info_and_wait_for_ending_all_threads(&progress_thread_run, progress_thread_handle); if check_was_stopped.load(Ordering::Relaxed) { return false; } self.check_for_duplicate_hashes(parts, &hashes_with_multiple_images, &all_hashed_images, &mut collected_similar_images); } self.verify_duplicated_items(&collected_similar_images); self.similar_vectors = collected_similar_images.into_values().collect(); self.exclude_items_with_same_size(); self.check_for_reference_folders(); Common::print_time(hash_map_modification, SystemTime::now(), "sort_images - selecting data from HashMap"); if self.use_reference_folders { for (_fe, vector) in &self.similar_referenced_vectors { self.information.number_of_duplicates += vector.len(); self.information.number_of_groups += 1; } } else { for vector in &self.similar_vectors { self.information.number_of_duplicates += vector.len() - 1; self.information.number_of_groups += 1; } } // Clean unused data self.image_hashes = Default::default(); self.images_to_check = Default::default(); self.bktree = BKTree::new(Hamming); true } fn exclude_items_with_same_size(&mut self) { if self.exclude_images_with_same_size { let mut new_vector = Default::default(); mem::swap(&mut self.similar_vectors, &mut new_vector); for vec_file_entry in new_vector { let mut bt_sizes: BTreeSet = Default::default(); let mut vec_values = Vec::new(); for file_entry in vec_file_entry { if !bt_sizes.contains(&file_entry.size) { bt_sizes.insert(file_entry.size); vec_values.push(file_entry); } } if vec_values.len() > 1 { self.similar_vectors.push(vec_values); } } } } fn check_for_reference_folders(&mut self) { if self.use_reference_folders { let mut similar_vector = Default::default(); mem::swap(&mut self.similar_vectors, &mut similar_vector); let reference_directories = self.directories.reference_directories.clone(); self.similar_referenced_vectors = similar_vector .into_iter() .filter_map(|vec_file_entry| { let mut files_from_referenced_folders = Vec::new(); let mut normal_files = Vec::new(); for file_entry in vec_file_entry { if reference_directories.iter().any(|e| file_entry.path.starts_with(e)) { files_from_referenced_folders.push(file_entry); } else { normal_files.push(file_entry); } } if files_from_referenced_folders.is_empty() || normal_files.is_empty() { None } else { Some((files_from_referenced_folders.pop().unwrap(), normal_files)) } }) .collect::)>>(); } } #[allow(dead_code)] #[allow(unreachable_code)] #[allow(unused_variables)] pub fn verify_duplicated_items(&self, collected_similar_images: &HashMap>) { #[cfg(not(debug_assertions))] return; // Validating if group contains duplicated results let mut result_hashset: HashSet = Default::default(); let mut found = false; for vec_file_entry in collected_similar_images.values() { if vec_file_entry.is_empty() { println!("Empty group"); found = true; continue; } if vec_file_entry.len() == 1 { println!("Single Element {vec_file_entry:?}"); found = true; continue; } for file_entry in vec_file_entry { let st = file_entry.path.to_string_lossy().to_string(); if result_hashset.contains(&st) { found = true; println!("Duplicated Element {st}"); } else { result_hashset.insert(st); } } } assert!(!found, "Found Invalid entries, verify errors before"); // TODO crashes with empty result with reference folder, verify why } /// Set included dir which needs to be relative, exists etc. pub fn set_included_directory(&mut self, included_directory: Vec) { self.directories.set_included_directory(included_directory, &mut self.text_messages); } pub fn set_reference_directory(&mut self, reference_directory: Vec) { self.directories.set_reference_directory(reference_directory); } pub fn set_excluded_directory(&mut self, excluded_directory: Vec) { self.directories.set_excluded_directory(excluded_directory, &mut self.text_messages); } pub fn set_excluded_items(&mut self, excluded_items: Vec) { self.excluded_items.set_excluded_items(excluded_items, &mut self.text_messages); } } fn image_to_check<'a>( hashes_parents: &mut HashMap, hashes_similarity: &mut HashMap, hashes_with_multiple_images: &HashSet, hash_to_check: &'a ImHash, compared_hash: &'a ImHash, similarity: u32, ) { if let Some(children_number) = hashes_parents.get(compared_hash) { if *children_number > 0 || hashes_with_multiple_images.contains(compared_hash) { return; } } // If there is already record, with smaller sensitivity, then replace it let mut need_to_add = false; let mut need_to_check = false; // TODO consider to replace variables from above with closures // If current checked hash, have parent, first we must check if similarity between them is lower than checked item if let Some((current_parent_hash, current_similarity_with_parent)) = hashes_similarity.get(hash_to_check) { if *current_similarity_with_parent > similarity { need_to_check = true; *hashes_parents.get_mut(current_parent_hash).unwrap() -= 1; hashes_similarity.remove(hash_to_check).unwrap(); } } else { need_to_check = true; } if need_to_check { if let Some((other_parent_hash, other_similarity)) = hashes_similarity.get(compared_hash) { if *other_similarity > similarity { need_to_add = true; *hashes_parents.get_mut(other_parent_hash).unwrap() -= 1; } } // But when there is no record, just add it else { need_to_add = true; } } if need_to_add { hashes_similarity.insert(compared_hash.clone(), (hash_to_check.clone(), similarity)); if let Some(number_of_children) = hashes_parents.get_mut(hash_to_check) { *number_of_children += 1; } else { hashes_parents.insert(hash_to_check.clone(), 1); } } } fn is_in_reference_folder(reference_directories: &[PathBuf], path: &Path) -> bool { reference_directories.iter().any(|e| path.starts_with(e)) } impl Default for SimilarImages { fn default() -> Self { Self::new() } } impl DebugPrint for SimilarImages { #[allow(dead_code)] #[allow(unreachable_code)] fn debug_print(&self) { #[cfg(not(debug_assertions))] { return; } println!("---------------DEBUG PRINT---------------"); println!("Included directories - {:?}", self.directories.included_directories); println!("-----------------------------------------"); } } impl SaveResults for SimilarImages { fn save_results_to_file(&mut self, file_name: &str) -> bool { let start_time: SystemTime = SystemTime::now(); let file_name: String = match file_name { "" => "results.txt".to_string(), k => k.to_string(), }; let file_handler = match File::create(&file_name) { Ok(t) => t, Err(e) => { self.text_messages.errors.push(format!("Failed to create file {file_name}, reason {e}")); return false; } }; let mut writer = BufWriter::new(file_handler); if let Err(e) = writeln!( writer, "Results of searching {:?} with excluded directories {:?} and excluded items {:?}", self.directories.included_directories, self.directories.excluded_directories, self.excluded_items.items ) { self.text_messages.errors.push(format!("Failed to save results to file {file_name}, reason {e}")); return false; } if !self.similar_vectors.is_empty() { write!(writer, "{} images which have similar friends\n\n", self.similar_vectors.len()).unwrap(); for struct_similar in &self.similar_vectors { writeln!(writer, "Found {} images which have similar friends", self.similar_vectors.len()).unwrap(); for file_entry in struct_similar { writeln!( writer, "{} - {} - {} - {}", file_entry.path.display(), file_entry.dimensions, format_size(file_entry.size, BINARY), get_string_from_similarity(&file_entry.similarity, self.hash_size) ) .unwrap(); } writeln!(writer).unwrap(); } } else { write!(writer, "Not found any similar images.").unwrap(); } Common::print_time(start_time, SystemTime::now(), "save_results_to_file"); true } } impl PrintResults for SimilarImages { fn print_results(&self) { if !self.similar_vectors.is_empty() { println!("Found {} images which have similar friends", self.similar_vectors.len()); for vec_file_entry in &self.similar_vectors { for file_entry in vec_file_entry { println!( "{} - {} - {} - {}", file_entry.path.display(), file_entry.dimensions, format_size(file_entry.size, BINARY), get_string_from_similarity(&file_entry.similarity, self.hash_size) ); } println!(); } } } } pub fn save_hashes_to_file( hashmap: &HashMap, text_messages: &mut Messages, save_also_as_json: bool, hash_size: u8, hash_alg: HashAlg, image_filter: FilterType, ) { if let Some(((file_handler, cache_file), (file_handler_json, cache_file_json))) = open_cache_folder(&get_cache_file(&hash_size, &hash_alg, &image_filter), true, save_also_as_json, &mut text_messages.warnings) { { let writer = BufWriter::new(file_handler.unwrap()); // Unwrap because cannot fail here if let Err(e) = bincode::serialize_into(writer, hashmap) { text_messages .warnings .push(format!("Cannot write data to cache file {}, reason {}", cache_file.display(), e)); return; } } if save_also_as_json { if let Some(file_handler_json) = file_handler_json { let writer = BufWriter::new(file_handler_json); if let Err(e) = serde_json::to_writer(writer, hashmap) { text_messages .warnings .push(format!("Cannot write data to cache file {}, reason {}", cache_file_json.display(), e)); return; } } } text_messages.messages.push(format!("Properly saved to file {} cache entries.", hashmap.len())); } } pub fn load_hashes_from_file( text_messages: &mut Messages, delete_outdated_cache: bool, hash_size: u8, hash_alg: HashAlg, image_filter: FilterType, ) -> Option> { if let Some(((file_handler, cache_file), (file_handler_json, cache_file_json))) = open_cache_folder(&get_cache_file(&hash_size, &hash_alg, &image_filter), false, true, &mut text_messages.warnings) { let mut hashmap_loaded_entries: HashMap; if let Some(file_handler) = file_handler { let reader = BufReader::new(file_handler); hashmap_loaded_entries = match bincode::deserialize_from(reader) { Ok(t) => t, Err(e) => { text_messages .warnings .push(format!("Failed to load data from cache file {}, reason {}", cache_file.display(), e)); return None; } }; } else { let reader = BufReader::new(file_handler_json.unwrap()); // Unwrap cannot fail, because at least one file must be valid hashmap_loaded_entries = match serde_json::from_reader(reader) { Ok(t) => t, Err(e) => { text_messages .warnings .push(format!("Failed to load data from cache file {}, reason {}", cache_file_json.display(), e)); return None; } }; } // Don't load cache data if destination file not exists if delete_outdated_cache { hashmap_loaded_entries.retain(|src_path, _file_entry| Path::new(src_path).exists()); } text_messages.messages.push(format!("Properly loaded {} cache entries.", hashmap_loaded_entries.len())); return Some(hashmap_loaded_entries); } None } fn get_cache_file(hash_size: &u8, hash_alg: &HashAlg, image_filter: &FilterType) -> String { format!( "cache_similar_images_{}_{}_{}_50.bin", hash_size, convert_algorithm_to_string(hash_alg), convert_filters_to_string(image_filter), ) } #[must_use] pub fn get_string_from_similarity(similarity: &u32, hash_size: u8) -> String { let index_preset = match hash_size { 8 => 0, 16 => 1, 32 => 2, 64 => 3, _ => panic!(), }; if *similarity == 0 { flc!("core_similarity_original") } else if *similarity <= SIMILAR_VALUES[index_preset][0] { flc!("core_similarity_very_high") } else if *similarity <= SIMILAR_VALUES[index_preset][1] { flc!("core_similarity_high") } else if *similarity <= SIMILAR_VALUES[index_preset][2] { flc!("core_similarity_medium") } else if *similarity <= SIMILAR_VALUES[index_preset][3] { flc!("core_similarity_small") } else if *similarity <= SIMILAR_VALUES[index_preset][4] { flc!("core_similarity_very_small") } else if *similarity <= SIMILAR_VALUES[index_preset][5] { flc!("core_similarity_minimal") } else { panic!(); } } #[must_use] pub fn return_similarity_from_similarity_preset(similarity_preset: &SimilarityPreset, hash_size: u8) -> u32 { let index_preset = match hash_size { 8 => 0, 16 => 1, 32 => 2, 64 => 3, _ => panic!(), }; match similarity_preset { SimilarityPreset::Original => 0, SimilarityPreset::VeryHigh => SIMILAR_VALUES[index_preset][0], SimilarityPreset::High => SIMILAR_VALUES[index_preset][1], SimilarityPreset::Medium => SIMILAR_VALUES[index_preset][2], SimilarityPreset::Small => SIMILAR_VALUES[index_preset][3], SimilarityPreset::VerySmall => SIMILAR_VALUES[index_preset][4], SimilarityPreset::Minimal => SIMILAR_VALUES[index_preset][5], SimilarityPreset::None => panic!(""), } } fn convert_filters_to_string(image_filter: &FilterType) -> String { match image_filter { FilterType::Lanczos3 => "Lanczos3", FilterType::Nearest => "Nearest", FilterType::Triangle => "Triangle", FilterType::Gaussian => "Gaussian", FilterType::CatmullRom => "CatmullRom", } .to_string() } fn convert_algorithm_to_string(hash_alg: &HashAlg) -> String { match hash_alg { HashAlg::Mean => "Mean", HashAlg::Gradient => "Gradient", HashAlg::Blockhash => "Blockhash", HashAlg::VertGradient => "VertGradient", HashAlg::DoubleGradient => "DoubleGradient", } .to_string() } pub fn test_image_conversion_speed() { let file_name: &str = "test.jpg"; let file_path = Path::new(file_name); match image::open(file_path) { Ok(img_open) => { for alg in [HashAlg::Blockhash, HashAlg::Gradient, HashAlg::DoubleGradient, HashAlg::VertGradient, HashAlg::Mean] { for filter in [ FilterType::Lanczos3, FilterType::CatmullRom, FilterType::Gaussian, FilterType::Nearest, FilterType::Triangle, ] { for size in [8, 16, 32, 64] { let hasher_config = HasherConfig::new().hash_alg(alg).resize_filter(filter).hash_size(size, size); let start = SystemTime::now(); let hasher = hasher_config.to_hasher(); let _hash = hasher.hash_image(&img_open); let end = SystemTime::now(); println!("{:?} us {:?} {:?} {}x{}", end.duration_since(start).unwrap().as_micros(), alg, filter, size, size); } } } } Err(e) => { println!( "Failed to open test file {}, reason {}", match file_path.canonicalize() { Ok(t) => t.to_string_lossy().to_string(), Err(_inspected) => file_name.to_string(), }, e ); } } } #[allow(dead_code)] #[allow(unreachable_code)] #[allow(unused_variables)] // Function to validate if after first check there are any duplicated entries // E.g. /a.jpg is used also as master and similar image which is forbidden, because may // cause accidentally delete more pictures that user wanted fn debug_check_for_duplicated_things( use_reference_folders: bool, hashes_parents: &HashMap, hashes_similarity: &HashMap, all_hashed_images: &HashMap>, numm: &str, ) { #[cfg(not(debug_assertions))] return; if use_reference_folders { return; } let mut found_broken_thing = false; let mut hashmap_hashes: HashSet<_> = Default::default(); let mut hashmap_names: HashSet<_> = Default::default(); for (hash, number_of_children) in hashes_parents { if *number_of_children > 0 { if hashmap_hashes.contains(hash) { println!("------1--HASH--{} {:?}", numm, all_hashed_images.get(hash).unwrap()); found_broken_thing = true; } hashmap_hashes.insert((*hash).clone()); for i in all_hashed_images.get(hash).unwrap() { let name = i.path.to_string_lossy().to_string(); if hashmap_names.contains(&name) { println!("------1--NAME--{numm} {name:?}"); found_broken_thing = true; } hashmap_names.insert(name); } } } for hash in hashes_similarity.keys() { if hashmap_hashes.contains(hash) { println!("------2--HASH--{} {:?}", numm, all_hashed_images.get(hash).unwrap()); found_broken_thing = true; } hashmap_hashes.insert((*hash).clone()); for i in all_hashed_images.get(hash).unwrap() { let name = i.path.to_string_lossy().to_string(); if hashmap_names.contains(&name) { println!("------2--NAME--{numm} {name:?}"); found_broken_thing = true; } hashmap_names.insert(name); } } assert!(!found_broken_thing); }