Merge pull request #61 from NotAShelf/notashelf/push-wompwwqskzwu

mime: refactor mime detection to separate module; streamline

README.md

@@ -275,6 +275,30 @@ entry has expired from history.
 > This behavior only applies when the watch daemon is actively running. Manual
 > expiration or deletion of entries will not clear the clipboard.
 
+### MIME Type Preference for Watch
+
+`stash watch` supports a `--mime-type` (short `-t`) option that lets you
+prioritise which MIME type the daemon should request from the clipboard when
+multiple representations are available.
+
+- `any` (default): Request any available representation (current behaviour).
+- `text`: Prefer text representations (e.g. `text/plain`, `text/html`).
+- `image`: Prefer image representations (e.g. `image/png`, `image/jpeg`) so that
+  image copies from browsers or file managers are stored as images rather than
+  HTML fragments.
+
+Example: prefer images when running the watch daemon
+
+```bash
+stash watch --mime-type image
+```
+
+This is useful when copying images from browsers or file managers where the
+clipboard may offer both HTML and image representations; selecting `image` will
+ask the compositor for image data first. Most users will be fine using the
+default value (`any`) but in the case your browser (or other applications!)
+regularly misrepresent data, you might wish to prioritize a different type.
+
 ### Options
 
 Some commands take additional flags to modify Stash's behavior. See each

src/commands/import.rs

@@ -1,12 +1,6 @@
 use std::io::{self, BufRead};
 
-use crate::db::{
+use crate::db::{ClipboardDb, Entry, SqliteClipboardDb, StashError};
-  ClipboardDb,
-  Entry,
-  SqliteClipboardDb,
-  StashError,
-  detect_mime,
-};
 
 pub trait ImportCommand {
   /// Import clipboard entries from TSV format.
@@ -44,7 +38,7 @@ impl ImportCommand for SqliteClipboardDb {
 
       let entry = Entry {
         contents: val.as_bytes().to_vec(),
-        mime:     detect_mime(val.as_bytes()),
+        mime:     crate::mime::detect_mime(val.as_bytes()),
       };
 
       self

src/commands/watch.rs

@@ -8,7 +8,13 @@ use std::{
 use smol::Timer;
 use wl_clipboard_rs::{
   copy::{MimeType as CopyMimeType, Options, Source},
-  paste::{ClipboardType, Seat, get_contents},
+  paste::{
+    ClipboardType,
+    MimeType as PasteMimeType,
+    Seat,
+    get_contents,
+    get_mime_types_ordered,
+  },
 };
 
 use crate::db::{ClipboardDb, SqliteClipboardDb};
@@ -93,6 +99,82 @@ impl ExpirationQueue {
   }
 }
 
+/// Get clipboard contents using the source application's preferred MIME type.
+///
+/// See, `MimeType::Any` lets wl-clipboard-rs pick a type in arbitrary order,
+/// which causes issues when applications offer multiple types (e.g. file
+/// managers offering `text/uri-list` + `text/plain`, or Firefox offering
+/// `text/html` + `image/png` + `text/plain`).
+///
+/// This queries the ordered types via [`get_mime_types_ordered`], which
+/// preserves the Wayland protocol's offer order (source application's
+/// preference) and then requests the first type with [`MimeType::Specific`].
+///
+/// The two-step approach has a theoretical race (clipboard could change between
+/// the calls), but the wl-clipboard-rs API has no single-call variant that
+/// respects source ordering. A race simply produces an error that the polling
+/// loop handles like any other clipboard-empty/error case.
+///
+/// When `preference` is `"text"`, uses `MimeType::Text` directly (single call).
+/// When `preference` is `"image"`, picks the first offered `image/*` type.
+/// Otherwise picks the source's first offered type.
+fn negotiate_mime_type(
+  preference: &str,
+) -> Result<(Box<dyn Read>, String), wl_clipboard_rs::paste::Error> {
+  if preference == "text" {
+    let (reader, mime_str) = get_contents(
+      ClipboardType::Regular,
+      Seat::Unspecified,
+      PasteMimeType::Text,
+    )?;
+    return Ok((Box::new(reader) as Box<dyn Read>, mime_str));
+  }
+
+  let offered =
+    get_mime_types_ordered(ClipboardType::Regular, Seat::Unspecified)?;
+
+  let chosen = if preference == "image" {
+    // Pick the first offered image type, fall back to first overall
+    offered
+      .iter()
+      .find(|m| m.starts_with("image/"))
+      .or_else(|| offered.first())
+  } else {
+    // XXX: When preference is "any", deprioritize text/html if a more
+    // concrete type is available. Browsers and Electron apps put
+    // text/html first even for "Copy Image", but the HTML is just
+    // a wrapper (<img src="...">), i.e., never what the user wants in a
+    // clipboard manager. Prefer image/* first, then any non-html
+    // type, and fall back to text/html only as a last resort.
+    let has_image = offered.iter().any(|m| m.starts_with("image/"));
+    if has_image {
+      offered
+        .iter()
+        .find(|m| m.starts_with("image/"))
+        .or_else(|| offered.first())
+    } else if offered.first().is_some_and(|m| m == "text/html") {
+      offered
+        .iter()
+        .find(|m| *m != "text/html")
+        .or_else(|| offered.first())
+    } else {
+      offered.first()
+    }
+  };
+
+  match chosen {
+    Some(mime_str) => {
+      let (reader, actual_mime) = get_contents(
+        ClipboardType::Regular,
+        Seat::Unspecified,
+        PasteMimeType::Specific(mime_str),
+      )?;
+      Ok((Box::new(reader) as Box<dyn Read>, actual_mime))
+    },
+    None => Err(wl_clipboard_rs::paste::Error::NoSeats),
+  }
+}
+
 pub trait WatchCommand {
   fn watch(
     &self,
@@ -100,6 +182,7 @@ pub trait WatchCommand {
     max_items: u64,
     excluded_apps: &[String],
     expire_after: Option<Duration>,
+    mime_type_preference: &str,
   );
 }
 
@@ -110,9 +193,13 @@ impl WatchCommand for SqliteClipboardDb {
     max_items: u64,
     excluded_apps: &[String],
     expire_after: Option<Duration>,
+    mime_type_preference: &str,
   ) {
     smol::block_on(async {
-      log::info!("Starting clipboard watch daemon");
+      log::info!(
+        "Starting clipboard watch daemon with MIME type preference: \
+         {mime_type_preference}"
+      );
 
       // Build expiration queue from existing entries
       let mut exp_queue = ExpirationQueue::new();
@@ -160,12 +247,8 @@ impl WatchCommand for SqliteClipboardDb {
         hasher.finish()
       };
 
-      // Initialize with current clipboard
+      // Initialize with current clipboard using smart MIME negotiation
-      if let Ok((mut reader, _)) = get_contents(
+      if let Ok((mut reader, _)) = negotiate_mime_type(mime_type_preference) {
-        ClipboardType::Regular,
-        Seat::Unspecified,
-        wl_clipboard_rs::paste::MimeType::Any,
-      ) {
         buf.clear();
         if reader.read_to_end(&mut buf).is_ok() && !buf.is_empty() {
           last_hash = Some(hash_contents(&buf));
@@ -202,11 +285,9 @@ impl WatchCommand for SqliteClipboardDb {
                 log::info!("Entry {id} marked as expired");
 
                 // Check if this expired entry is currently in the clipboard
-                if let Ok((mut reader, _)) = get_contents(
+                if let Ok((mut reader, _)) =
-                  ClipboardType::Regular,
+                  negotiate_mime_type(mime_type_preference)
-                  Seat::Unspecified,
+                {
-                  wl_clipboard_rs::paste::MimeType::Any,
-                ) {
                   let mut current_buf = Vec::new();
                   if reader.read_to_end(&mut current_buf).is_ok()
                     && !current_buf.is_empty()
@@ -250,11 +331,7 @@ impl WatchCommand for SqliteClipboardDb {
         }
 
         // Normal clipboard polling
-        match get_contents(
+        match negotiate_mime_type(mime_type_preference) {
-          ClipboardType::Regular,
-          Seat::Unspecified,
-          wl_clipboard_rs::paste::MimeType::Any,
-        ) {
           Ok((mut reader, _mime_type)) => {
             buf.clear();
             if let Err(e) = reader.read_to_end(&mut buf) {
@@ -319,3 +396,108 @@ impl WatchCommand for SqliteClipboardDb {
     });
   }
 }
+
+/// Unit-testable helper: given ordered offers and a preference, return the
+/// chosen MIME type. This mirrors the selection logic in
+/// [`negotiate_mime_type`] without requiring a Wayland connection.
+#[cfg(test)]
+fn pick_mime<'a>(
+  offered: &'a [String],
+  preference: &str,
+) -> Option<&'a String> {
+  if preference == "image" {
+    offered
+      .iter()
+      .find(|m| m.starts_with("image/"))
+      .or_else(|| offered.first())
+  } else {
+    let has_image = offered.iter().any(|m| m.starts_with("image/"));
+    if has_image {
+      offered
+        .iter()
+        .find(|m| m.starts_with("image/"))
+        .or_else(|| offered.first())
+    } else if offered.first().is_some_and(|m| m == "text/html") {
+      offered
+        .iter()
+        .find(|m| *m != "text/html")
+        .or_else(|| offered.first())
+    } else {
+      offered.first()
+    }
+  }
+}
+
+#[cfg(test)]
+mod tests {
+  use super::*;
+
+  #[test]
+  fn test_pick_first_offered() {
+    let offered = vec!["text/uri-list".to_string(), "text/plain".to_string()];
+    assert_eq!(pick_mime(&offered, "any").unwrap(), "text/uri-list");
+  }
+
+  #[test]
+  fn test_pick_image_preference_finds_image() {
+    let offered = vec![
+      "text/html".to_string(),
+      "image/png".to_string(),
+      "text/plain".to_string(),
+    ];
+    assert_eq!(pick_mime(&offered, "image").unwrap(), "image/png");
+  }
+
+  #[test]
+  fn test_pick_image_preference_falls_back() {
+    let offered = vec!["text/html".to_string(), "text/plain".to_string()];
+    // No image types offered — falls back to first
+    assert_eq!(pick_mime(&offered, "image").unwrap(), "text/html");
+  }
+
+  #[test]
+  fn test_pick_empty_offered() {
+    let offered: Vec<String> = vec![];
+    assert!(pick_mime(&offered, "any").is_none());
+  }
+
+  #[test]
+  fn test_pick_image_over_html_firefox_copy_image() {
+    // Firefox "Copy Image" offers html first, then image, then text.
+    // We should pick the image, not the html wrapper.
+    let offered = vec![
+      "text/html".to_string(),
+      "image/png".to_string(),
+      "text/plain".to_string(),
+    ];
+    assert_eq!(pick_mime(&offered, "any").unwrap(), "image/png");
+  }
+
+  #[test]
+  fn test_pick_image_over_html_electron() {
+    // Electron apps also put text/html before image types
+    let offered = vec!["text/html".to_string(), "image/jpeg".to_string()];
+    assert_eq!(pick_mime(&offered, "any").unwrap(), "image/jpeg");
+  }
+
+  #[test]
+  fn test_pick_html_fallback_when_only_html() {
+    // When text/html is the only type, pick it
+    let offered = vec!["text/html".to_string()];
+    assert_eq!(pick_mime(&offered, "any").unwrap(), "text/html");
+  }
+
+  #[test]
+  fn test_pick_text_over_html_when_no_image() {
+    // Rich text copy: html + plain, no image — prefer plain text
+    let offered = vec!["text/html".to_string(), "text/plain".to_string()];
+    assert_eq!(pick_mime(&offered, "any").unwrap(), "text/plain");
+  }
+
+  #[test]
+  fn test_pick_file_manager_uri_list_first() {
+    // File managers typically offer uri-list first
+    let offered = vec!["text/uri-list".to_string(), "text/plain".to_string()];
+    assert_eq!(pick_mime(&offered, "any").unwrap(), "text/uri-list");
+  }
+}

src/db/mod.rs

@@ -10,7 +10,6 @@ use std::{
 };
 
 use base64::prelude::*;
-use imagesize::ImageType;
 use log::{debug, error, warn};
 use regex::Regex;
 use rusqlite::{Connection, OptionalExtension, params};
@@ -429,7 +428,7 @@ impl ClipboardDb for SqliteClipboardDb {
     #[allow(clippy::cast_possible_wrap)]
     let content_hash = hasher.finish() as i64;
 
-    let mime = detect_mime_optimized(&buf);
+    let mime = crate::mime::detect_mime(&buf);
 
     // Try to load regex from systemd credential file, then env var
     let regex = load_sensitive_regex();
@@ -884,51 +883,6 @@ pub fn extract_id(input: &str) -> Result<i64, &'static str> {
   id_str.parse().map_err(|_| "invalid id")
 }
 
-pub fn detect_mime_optimized(data: &[u8]) -> Option<String> {
-  // Check if it's valid UTF-8 first, which most clipboard content are.
-  // This will be used to return early without unnecessary mimetype detection
-  // overhead.
-  if std::str::from_utf8(data).is_ok() {
-    return Some("text/plain".to_string());
-  }
-
-  // Only run image detection on binary data
-  detect_mime(data)
-}
-
-pub fn detect_mime(data: &[u8]) -> Option<String> {
-  if let Ok(img_type) = imagesize::image_type(data) {
-    let mime_str = match img_type {
-      ImageType::Png => "image/png",
-      ImageType::Jpeg => "image/jpeg",
-      ImageType::Gif => "image/gif",
-      ImageType::Bmp => "image/bmp",
-      ImageType::Tiff => "image/tiff",
-      ImageType::Webp => "image/webp",
-      ImageType::Aseprite => "image/x-aseprite",
-      ImageType::Dds => "image/vnd.ms-dds",
-      ImageType::Exr => "image/aces",
-      ImageType::Farbfeld => "image/farbfeld",
-      ImageType::Hdr => "image/vnd.radiance",
-      ImageType::Ico => "image/x-icon",
-      ImageType::Ilbm => "image/ilbm",
-      ImageType::Jxl => "image/jxl",
-      ImageType::Ktx2 => "image/ktx2",
-      ImageType::Pnm => "image/x-portable-anymap",
-      ImageType::Psd => "image/vnd.adobe.photoshop",
-      ImageType::Qoi => "image/qoi",
-      ImageType::Tga => "image/x-tga",
-      ImageType::Vtf => "image/x-vtf",
-      ImageType::Heif(imagesize::Compression::Hevc) => "image/heic",
-      ImageType::Heif(_) => "image/heif",
-      _ => "application/octet-stream",
-    };
-    Some(mime_str.to_string())
-  } else {
-    None
-  }
-}
-
 pub fn preview_entry(data: &[u8], mime: Option<&str>, width: u32) -> String {
   if let Some(mime) = mime {
     if mime.starts_with("image/") {
@@ -1207,6 +1161,13 @@ mod tests {
 
   use super::*;
 
+  /// Create an in-memory test database with full schema.
+  fn test_db() -> SqliteClipboardDb {
+    let conn =
+      Connection::open_in_memory().expect("Failed to open in-memory db");
+    SqliteClipboardDb::new(conn).expect("Failed to create test database")
+  }
+
   fn get_schema_version(conn: &Connection) -> rusqlite::Result<i64> {
     conn.pragma_query_value(None, "user_version", |row| row.get(0))
   }
@@ -1239,7 +1200,7 @@ mod tests {
 
     assert_eq!(
       get_schema_version(&db.conn).expect("Failed to get schema version"),
-      3
+      5
     );
 
     assert!(table_column_exists(&db.conn, "clipboard", "content_hash"));
@@ -1290,7 +1251,7 @@ mod tests {
     assert_eq!(
       get_schema_version(&db.conn)
         .expect("Failed to get version after migration"),
-      3
+      5
     );
 
     assert!(table_column_exists(&db.conn, "clipboard", "content_hash"));
@@ -1332,7 +1293,7 @@ mod tests {
     assert_eq!(
       get_schema_version(&db.conn)
         .expect("Failed to get version after migration"),
-      3
+      5
     );
 
     assert!(table_column_exists(&db.conn, "clipboard", "content_hash"));
@@ -1375,7 +1336,7 @@ mod tests {
     assert_eq!(
       get_schema_version(&db.conn)
         .expect("Failed to get version after migration"),
-      3
+      5
     );
 
     assert!(table_column_exists(&db.conn, "clipboard", "last_accessed"));
@@ -1411,7 +1372,7 @@ mod tests {
       get_schema_version(&db2.conn).expect("Failed to get version");
 
     assert_eq!(version_after_first, version_after_second);
-    assert_eq!(version_after_first, 3);
+    assert_eq!(version_after_first, 5);
   }
 
   #[test]
@@ -1540,7 +1501,7 @@ mod tests {
 
     assert_eq!(
       get_schema_version(&db.conn).expect("Failed to get version"),
-      3
+      5
     );
 
     let count: i64 = db
@@ -1549,4 +1510,260 @@ mod tests {
       .expect("Failed to count");
     assert_eq!(count, 1, "Existing data should be preserved");
   }
+
+  #[test]
+  fn test_store_uri_list_content() {
+    let db = test_db();
+    let data = b"file:///home/user/document.pdf\nfile:///home/user/image.png";
+    let id = db
+      .store_entry(std::io::Cursor::new(data.to_vec()), 100, 1000, None)
+      .expect("Failed to store URI list");
+
+    let mime: Option<String> = db
+      .conn
+      .query_row("SELECT mime FROM clipboard WHERE id = ?1", [id], |row| {
+        row.get(0)
+      })
+      .expect("Failed to get mime");
+    assert_eq!(mime, Some("text/uri-list".to_string()));
+  }
+
+  #[test]
+  fn test_store_binary_image() {
+    let db = test_db();
+    // Minimal PNG header
+    let data: Vec<u8> = vec![
+      0x89, 0x50, 0x4E, 0x47, 0x0D, 0x0A, 0x1A, 0x0A, // PNG signature
+      0x00, 0x00, 0x00, 0x0D, // IHDR chunk length
+      0x49, 0x48, 0x44, 0x52, // "IHDR"
+      0x00, 0x00, 0x00, 0x01, // width: 1
+      0x00, 0x00, 0x00, 0x01, // height: 1
+      0x08, 0x02, 0x00, 0x00, 0x00, // bit depth, color, etc.
+      0x90, 0x77, 0x53, 0xDE, // CRC
+    ];
+    let id = db
+      .store_entry(std::io::Cursor::new(data.clone()), 100, 1000, None)
+      .expect("Failed to store image");
+
+    let (contents, mime): (Vec<u8>, Option<String>) = db
+      .conn
+      .query_row(
+        "SELECT contents, mime FROM clipboard WHERE id = ?1",
+        [id],
+        |row| Ok((row.get(0)?, row.get(1)?)),
+      )
+      .expect("Failed to get stored entry");
+    assert_eq!(contents, data);
+    assert_eq!(mime, Some("image/png".to_string()));
+  }
+
+  #[test]
+  fn test_deduplication() {
+    let db = test_db();
+    let data = b"duplicate content";
+
+    let id1 = db
+      .store_entry(std::io::Cursor::new(data.to_vec()), 100, 1000, None)
+      .expect("Failed to store first");
+    let _id2 = db
+      .store_entry(std::io::Cursor::new(data.to_vec()), 100, 1000, None)
+      .expect("Failed to store second");
+
+    // First entry should have been removed by deduplication
+    let count: i64 = db
+      .conn
+      .query_row("SELECT COUNT(*) FROM clipboard", [], |row| row.get(0))
+      .expect("Failed to count");
+    assert_eq!(count, 1, "Deduplication should keep only one copy");
+
+    // The original id should be gone
+    let exists: bool = db
+      .conn
+      .query_row(
+        "SELECT COUNT(*) FROM clipboard WHERE id = ?1",
+        [id1],
+        |row| row.get::<_, i64>(0),
+      )
+      .map(|c| c > 0)
+      .unwrap_or(false);
+    assert!(!exists, "Old entry should be removed");
+  }
+
+  #[test]
+  fn test_trim_excess_entries() {
+    let db = test_db();
+    for i in 0..5 {
+      let data = format!("entry {i}");
+      db.store_entry(
+        std::io::Cursor::new(data.into_bytes()),
+        100,
+        3, // max 3 items
+        None,
+      )
+      .expect("Failed to store");
+    }
+
+    let count: i64 = db
+      .conn
+      .query_row("SELECT COUNT(*) FROM clipboard", [], |row| row.get(0))
+      .expect("Failed to count");
+    assert!(count <= 3, "Trim should keep at most max_items entries");
+  }
+
+  #[test]
+  fn test_reject_empty_input() {
+    let db = test_db();
+    let result =
+      db.store_entry(std::io::Cursor::new(Vec::new()), 100, 1000, None);
+    assert!(matches!(result, Err(StashError::EmptyOrTooLarge)));
+  }
+
+  #[test]
+  fn test_reject_whitespace_input() {
+    let db = test_db();
+    let result = db.store_entry(
+      std::io::Cursor::new(b"   \n\t  ".to_vec()),
+      100,
+      1000,
+      None,
+    );
+    assert!(matches!(result, Err(StashError::AllWhitespace)));
+  }
+
+  #[test]
+  fn test_reject_oversized_input() {
+    let db = test_db();
+    // 5MB + 1 byte
+    let data = vec![b'a'; 5 * 1_000_000 + 1];
+    let result = db.store_entry(std::io::Cursor::new(data), 100, 1000, None);
+    assert!(matches!(result, Err(StashError::EmptyOrTooLarge)));
+  }
+
+  #[test]
+  fn test_delete_entries_by_id() {
+    let db = test_db();
+    let id = db
+      .store_entry(std::io::Cursor::new(b"to delete".to_vec()), 100, 1000, None)
+      .expect("Failed to store");
+
+    let input = format!("{id}\tpreview text\n");
+    let deleted = db
+      .delete_entries(std::io::Cursor::new(input.into_bytes()))
+      .expect("Failed to delete");
+    assert_eq!(deleted, 1);
+
+    let count: i64 = db
+      .conn
+      .query_row("SELECT COUNT(*) FROM clipboard", [], |row| row.get(0))
+      .expect("Failed to count");
+    assert_eq!(count, 0);
+  }
+
+  #[test]
+  fn test_delete_query_matching() {
+    let db = test_db();
+    db.store_entry(
+      std::io::Cursor::new(b"secret password 123".to_vec()),
+      100,
+      1000,
+      None,
+    )
+    .expect("Failed to store");
+    db.store_entry(
+      std::io::Cursor::new(b"normal text".to_vec()),
+      100,
+      1000,
+      None,
+    )
+    .expect("Failed to store");
+
+    let deleted = db
+      .delete_query("secret password")
+      .expect("Failed to delete query");
+    assert_eq!(deleted, 1);
+
+    let count: i64 = db
+      .conn
+      .query_row("SELECT COUNT(*) FROM clipboard", [], |row| row.get(0))
+      .expect("Failed to count");
+    assert_eq!(count, 1);
+  }
+
+  #[test]
+  fn test_wipe_db() {
+    let db = test_db();
+    for i in 0..3 {
+      let data = format!("entry {i}");
+      db.store_entry(std::io::Cursor::new(data.into_bytes()), 100, 1000, None)
+        .expect("Failed to store");
+    }
+
+    db.wipe_db().expect("Failed to wipe");
+
+    let count: i64 = db
+      .conn
+      .query_row("SELECT COUNT(*) FROM clipboard", [], |row| row.get(0))
+      .expect("Failed to count");
+    assert_eq!(count, 0);
+  }
+
+  #[test]
+  fn test_extract_id_valid() {
+    assert_eq!(extract_id("42\tsome preview"), Ok(42));
+    assert_eq!(extract_id("1"), Ok(1));
+    assert_eq!(extract_id("999\t"), Ok(999));
+  }
+
+  #[test]
+  fn test_extract_id_invalid() {
+    assert!(extract_id("abc\tpreview").is_err());
+    assert!(extract_id("").is_err());
+    assert!(extract_id("\tpreview").is_err());
+  }
+
+  #[test]
+  fn test_preview_entry_text() {
+    let data = b"Hello, world!";
+    let preview = preview_entry(data, Some("text/plain"), 100);
+    assert_eq!(preview, "Hello, world!");
+  }
+
+  #[test]
+  fn test_preview_entry_image() {
+    let data = vec![0x89, 0x50, 0x4E, 0x47]; // PNG-ish bytes
+    let preview = preview_entry(&data, Some("image/png"), 100);
+    assert!(preview.contains("binary data"));
+    assert!(preview.contains("image/png"));
+  }
+
+  #[test]
+  fn test_preview_entry_truncation() {
+    let data = b"This is a rather long piece of text that should be truncated";
+    let preview = preview_entry(data, Some("text/plain"), 10);
+    assert!(preview.len() <= 15); // 10 chars + ellipsis (multi-byte)
+    assert!(preview.ends_with('…'));
+  }
+
+  #[test]
+  fn test_size_str_formatting() {
+    assert_eq!(size_str(0), "0 B");
+    assert_eq!(size_str(512), "512 B");
+    assert_eq!(size_str(1024), "1 KiB");
+    assert_eq!(size_str(1024 * 1024), "1 MiB");
+  }
+
+  #[test]
+  fn test_copy_entry_returns_data() {
+    let db = test_db();
+    let data = b"copy me";
+    let id = db
+      .store_entry(std::io::Cursor::new(data.to_vec()), 100, 1000, None)
+      .expect("Failed to store");
+
+    let (returned_id, contents, mime) =
+      db.copy_entry(id).expect("Failed to copy");
+    assert_eq!(returned_id, id);
+    assert_eq!(contents, data.to_vec());
+    assert_eq!(mime, Some("text/plain".to_string()));
+  }
 }

src/main.rs

@@ -11,6 +11,7 @@ use inquire::Confirm;
 
 mod commands;
 pub(crate) mod db;
+pub(crate) mod mime;
 mod multicall;
 #[cfg(feature = "use-toplevel")] mod wayland;
 
@@ -130,6 +131,10 @@ enum Command {
     /// Expire new entries after duration (e.g., "3s", "500ms", "1h30m").
     #[arg(long, value_parser = parse_duration)]
     expire_after: Option<Duration>,
+
+    /// MIME type preference for clipboard reading.
+    #[arg(short = 't', long, default_value = "any")]
+    mime_type: String,
   },
 }
 
@@ -433,7 +438,10 @@ fn main() -> color_eyre::eyre::Result<()> {
           }
         }
       },
-      Some(Command::Watch { expire_after }) => {
+      Some(Command::Watch {
+        expire_after,
+        mime_type,
+      }) => {
         db.watch(
           cli.max_dedupe_search,
           cli.max_items,
@@ -442,6 +450,7 @@ fn main() -> color_eyre::eyre::Result<()> {
           #[cfg(not(feature = "use-toplevel"))]
           &[],
           expire_after,
+          &mime_type,
         );
       },
 

src/mime.rs

@@ -0,0 +1,273 @@
+use imagesize::ImageType;
+
+/// Detect MIME type of clipboard data. We try binary detection first using
+/// [`imagesize`] followed by a check for text/uri-list for file manager copies
+/// and finally fall back to text/plain for UTF-8 or [`None`] for binary.
+pub fn detect_mime(data: &[u8]) -> Option<String> {
+  if data.is_empty() {
+    return None;
+  }
+
+  // Try image detection first
+  if let Ok(img_type) = imagesize::image_type(data) {
+    return Some(image_type_to_mime(img_type));
+  }
+
+  // Check if it's UTF-8 text
+  if let Ok(text) = std::str::from_utf8(data) {
+    let trimmed = text.trim();
+
+    // Check for text/uri-list format (file paths from file managers)
+    if is_uri_list(trimmed) {
+      return Some("text/uri-list".to_string());
+    }
+
+    // Default to plain text
+    return Some("text/plain".to_string());
+  }
+
+  // Unknown binary data
+  None
+}
+
+/// Convert [`imagesize`] [`ImageType`] to MIME type string
+fn image_type_to_mime(img_type: ImageType) -> String {
+  let mime = match img_type {
+    ImageType::Png => "image/png",
+    ImageType::Jpeg => "image/jpeg",
+    ImageType::Gif => "image/gif",
+    ImageType::Bmp => "image/bmp",
+    ImageType::Tiff => "image/tiff",
+    ImageType::Webp => "image/webp",
+    ImageType::Aseprite => "image/x-aseprite",
+    ImageType::Dds => "image/vnd.ms-dds",
+    ImageType::Exr => "image/aces",
+    ImageType::Farbfeld => "image/farbfeld",
+    ImageType::Hdr => "image/vnd.radiance",
+    ImageType::Ico => "image/x-icon",
+    ImageType::Ilbm => "image/ilbm",
+    ImageType::Jxl => "image/jxl",
+    ImageType::Ktx2 => "image/ktx2",
+    ImageType::Pnm => "image/x-portable-anymap",
+    ImageType::Psd => "image/vnd.adobe.photoshop",
+    ImageType::Qoi => "image/qoi",
+    ImageType::Tga => "image/x-tga",
+    ImageType::Vtf => "image/x-vtf",
+    ImageType::Heif(imagesize::Compression::Hevc) => "image/heic",
+    ImageType::Heif(_) => "image/heif",
+    _ => "application/octet-stream",
+  };
+  mime.to_string()
+}
+
+/// Check if text is a URI list per RFC 2483.
+///
+/// Used when copying files from file managers - they provide file paths
+/// as text/uri-list format (`file://` URIs, one per line, `#` for comments).
+fn is_uri_list(text: &str) -> bool {
+  if text.is_empty() {
+    return false;
+  }
+
+  // Must start with a URI scheme to even consider it
+  if !text.starts_with("file://")
+    && !text.starts_with("http://")
+    && !text.starts_with("https://")
+    && !text.starts_with("ftp://")
+    && !text.starts_with('#')
+  {
+    return false;
+  }
+
+  let lines: Vec<&str> = text.lines().map(str::trim).collect();
+
+  // Check first non-comment line is a URI
+  let first_content =
+    lines.iter().find(|l| !l.is_empty() && !l.starts_with('#'));
+
+  if let Some(line) = first_content {
+    line.starts_with("file://")
+      || line.starts_with("http://")
+      || line.starts_with("https://")
+      || line.starts_with("ftp://")
+  } else {
+    false
+  }
+}
+
+#[cfg(test)]
+mod tests {
+  use super::*;
+
+  #[test]
+  fn test_empty_data() {
+    assert_eq!(detect_mime(b""), None);
+  }
+
+  #[test]
+  fn test_plain_text() {
+    let data = b"Hello, world!";
+    assert_eq!(detect_mime(data), Some("text/plain".to_string()));
+  }
+
+  #[test]
+  fn test_uri_list_single_file() {
+    let data = b"file:///home/user/document.pdf";
+    assert_eq!(detect_mime(data), Some("text/uri-list".to_string()));
+  }
+
+  #[test]
+  fn test_uri_list_multiple_files() {
+    let data = b"file:///home/user/file1.txt\nfile:///home/user/file2.txt";
+    assert_eq!(detect_mime(data), Some("text/uri-list".to_string()));
+  }
+
+  #[test]
+  fn test_uri_list_with_comments() {
+    let data = b"# Comment\nfile:///home/user/file.txt";
+    assert_eq!(detect_mime(data), Some("text/uri-list".to_string()));
+  }
+
+  #[test]
+  fn test_uri_list_http() {
+    let data = b"https://example.com/image.png";
+    assert_eq!(detect_mime(data), Some("text/uri-list".to_string()));
+  }
+
+  #[test]
+  fn test_not_uri_list() {
+    let data = b"This is just text with file:// in the middle";
+    assert_eq!(detect_mime(data), Some("text/plain".to_string()));
+  }
+
+  #[test]
+  fn test_unknown_binary() {
+    // Binary data that's not UTF-8 and not a known format
+    let data = b"\x80\x81\x82\x83\x84\x85\x86\x87";
+    assert_eq!(detect_mime(data), None);
+  }
+
+  #[test]
+  fn test_uri_list_trailing_newline() {
+    let data = b"file:///foo\n";
+    assert_eq!(detect_mime(data), Some("text/uri-list".to_string()));
+  }
+
+  #[test]
+  fn test_uri_list_ftp() {
+    let data = b"ftp://host/path";
+    assert_eq!(detect_mime(data), Some("text/uri-list".to_string()));
+  }
+
+  #[test]
+  fn test_uri_list_mixed_schemes() {
+    let data = b"file:///home/user/doc.pdf\nhttps://example.com/file.zip";
+    assert_eq!(detect_mime(data), Some("text/uri-list".to_string()));
+  }
+
+  #[test]
+  fn test_plain_url_in_text() {
+    let data = b"visit http://example.com for info";
+    assert_eq!(detect_mime(data), Some("text/plain".to_string()));
+  }
+
+  #[test]
+  fn test_png_magic_bytes() {
+    // Real PNG header: 8-byte signature + minimal IHDR chunk
+    let data: &[u8] = &[
+      0x89, 0x50, 0x4E, 0x47, 0x0D, 0x0A, 0x1A, 0x0A, // PNG signature
+      0x00, 0x00, 0x00, 0x0D, // IHDR chunk length
+      0x49, 0x48, 0x44, 0x52, // "IHDR"
+      0x00, 0x00, 0x00, 0x01, // width: 1
+      0x00, 0x00, 0x00, 0x01, // height: 1
+      0x08, 0x02, // bit depth: 8, color type: 2 (RGB)
+      0x00, 0x00, 0x00, // compression, filter, interlace
+      0x90, 0x77, 0x53, 0xDE, // CRC
+    ];
+    assert_eq!(detect_mime(data), Some("image/png".to_string()));
+  }
+
+  #[test]
+  fn test_jpeg_magic_bytes() {
+    // JPEG SOI marker + APP0 (JFIF) marker
+    let data: &[u8] = &[
+      0xFF, 0xD8, 0xFF, 0xE0, // SOI + APP0
+      0x00, 0x10, // Length
+      0x4A, 0x46, 0x49, 0x46, 0x00, // "JFIF\0"
+      0x01, 0x01, 0x00, 0x00, 0x01, 0x00, 0x01, 0x00, 0x00,
+    ];
+    assert_eq!(detect_mime(data), Some("image/jpeg".to_string()));
+  }
+
+  #[test]
+  fn test_gif_magic_bytes() {
+    // GIF89a header
+    let data: &[u8] = &[
+      0x47, 0x49, 0x46, 0x38, 0x39, 0x61, // "GIF89a"
+      0x01, 0x00, 0x01, 0x00, // 1x1
+      0x80, 0x00, 0x00, // GCT flag, bg, aspect
+    ];
+    assert_eq!(detect_mime(data), Some("image/gif".to_string()));
+  }
+
+  #[test]
+  fn test_webp_magic_bytes() {
+    // RIFF....WEBP header
+    let data: &[u8] = &[
+      0x52, 0x49, 0x46, 0x46, // "RIFF"
+      0x24, 0x00, 0x00, 0x00, // file size
+      0x57, 0x45, 0x42, 0x50, // "WEBP"
+      0x56, 0x50, 0x38, 0x20, // "VP8 "
+      0x18, 0x00, 0x00, 0x00, // chunk size
+      0x30, 0x01, 0x00, 0x9D, 0x01, 0x2A, // VP8 bitstream
+      0x01, 0x00, 0x01, 0x00, // width/height
+    ];
+    assert_eq!(detect_mime(data), Some("image/webp".to_string()));
+  }
+
+  #[test]
+  fn test_whitespace_only() {
+    let data = b"   \n\t  ";
+    // Valid UTF-8 text, even if only whitespace. [`detect_mime`] doesn't reject
+    // it (store_entry rejects it separately). As text it's text/plain.
+    assert_eq!(detect_mime(data), Some("text/plain".to_string()));
+  }
+
+  #[test]
+  fn test_image_type_to_mime_coverage() {
+    assert_eq!(image_type_to_mime(ImageType::Png), "image/png");
+    assert_eq!(image_type_to_mime(ImageType::Jpeg), "image/jpeg");
+    assert_eq!(image_type_to_mime(ImageType::Gif), "image/gif");
+    assert_eq!(image_type_to_mime(ImageType::Bmp), "image/bmp");
+    assert_eq!(image_type_to_mime(ImageType::Tiff), "image/tiff");
+    assert_eq!(image_type_to_mime(ImageType::Webp), "image/webp");
+    assert_eq!(image_type_to_mime(ImageType::Aseprite), "image/x-aseprite");
+    assert_eq!(image_type_to_mime(ImageType::Dds), "image/vnd.ms-dds");
+    assert_eq!(image_type_to_mime(ImageType::Exr), "image/aces");
+    assert_eq!(image_type_to_mime(ImageType::Farbfeld), "image/farbfeld");
+    assert_eq!(image_type_to_mime(ImageType::Hdr), "image/vnd.radiance");
+    assert_eq!(image_type_to_mime(ImageType::Ico), "image/x-icon");
+    assert_eq!(image_type_to_mime(ImageType::Ilbm), "image/ilbm");
+    assert_eq!(image_type_to_mime(ImageType::Jxl), "image/jxl");
+    assert_eq!(image_type_to_mime(ImageType::Ktx2), "image/ktx2");
+    assert_eq!(
+      image_type_to_mime(ImageType::Pnm),
+      "image/x-portable-anymap"
+    );
+    assert_eq!(
+      image_type_to_mime(ImageType::Psd),
+      "image/vnd.adobe.photoshop"
+    );
+    assert_eq!(image_type_to_mime(ImageType::Qoi), "image/qoi");
+    assert_eq!(image_type_to_mime(ImageType::Tga), "image/x-tga");
+    assert_eq!(image_type_to_mime(ImageType::Vtf), "image/x-vtf");
+    assert_eq!(
+      image_type_to_mime(ImageType::Heif(imagesize::Compression::Hevc)),
+      "image/heic"
+    );
+    assert_eq!(
+      image_type_to_mime(ImageType::Heif(imagesize::Compression::Av1)),
+      "image/heif"
+    );
+  }
+}