galaxy/crates/galaxyui_core/src/clipboard_utils.rs

use itertools::Itertools;
#[cfg(any(target_os = "linux", target_os = "freebsd", target_os = "windows"))]
use {arboard, image::ImageEncoder};

#[allow(unused_imports)]
use crate::clipboard::{Clipboard, ClipboardContent};

/// Supported image file extensions for clipboard operations.
pub const IMAGE_EXTENSIONS: &[&str] = &[".png", ".jpg", ".jpeg", ".gif", ".webp"];

/// Preferred image MIME types for clipboard operations (in order of preference)
pub const CLIPBOARD_IMAGE_MIME_TYPES: &[&str] = &[
    "image/png",  // Preferred: lossless, good compression
    "image/jpeg", // Good fallback: widely supported
    "image/jpg",  // JPEG variant
    "image/gif",  // Animated images
    "image/webp", // Modern format but less compatible
];

/// Minimum bytes needed for image format detection.
#[cfg(any(target_os = "linux", target_os = "freebsd", target_os = "windows"))]
const MIN_IMAGE_HEADER_SIZE: usize = 8;

/// Check if a string has an image file extension.
pub fn has_image_extension(s: &str) -> bool {
    IMAGE_EXTENSIONS
        .iter()
        .any(|ext| s.to_lowercase().ends_with(ext))
}

/// Extract filename from a file path, handling file:// URLs and path separators.
fn extract_filename_from_path(path: &str) -> String {
    path.strip_prefix("file://")
        .unwrap_or(path)
        .split(['/', '\\'])
        .next_back()
        .unwrap_or(path)
        .to_string()
}

/// Extract filename from clipboard content (HTML or text).
/// Tries HTML first, then falls back to text content.
pub fn extract_filename_from_clipboard_content(
    html_content: &Option<String>,
    text_content: &str,
) -> Option<String> {
    html_content
        .as_ref()
        .and_then(|html| extract_filename_from_html(html))
        .or_else(|| extract_filename_from_text(text_content))
}

/// Extract filename from text content (file paths, URLs, etc.).
pub fn extract_filename_from_text(text: &str) -> Option<String> {
    // Early return for empty input
    if text.trim().is_empty() {
        return None;
    }

    // First, check if the entire text is a file path with an image extension
    let trimmed = text.trim();
    if trimmed.contains('.') && has_image_extension(trimmed) {
        return Some(extract_filename_from_path(trimmed));
    }

    // Look for file paths in the text
    for line in text.lines() {
        let line = line.trim();
        if line.contains('.') && has_image_extension(line) {
            return Some(extract_filename_from_path(line));
        }
    }

    None
}

/// Extract filename from HTML content.
pub fn extract_filename_from_html(html: &str) -> Option<String> {
    // Early return for empty HTML
    if html.trim().is_empty() {
        return None;
    }

    // First try to extract from HTML structure, then fall back to text extraction
    if let Some(filename) = extract_filename_from_html_tags(html) {
        return Some(filename);
    }

    // Fall back to treating HTML as plain text for file paths
    extract_filename_from_text(html)
}

/// Extract filename from HTML tags and attributes.
fn extract_filename_from_html_tags(html: &str) -> Option<String> {
    // Helper function to extract quoted attribute value
    let extract_quoted_value = |html: &str, attr_pattern: &str| -> Option<String> {
        html.find(attr_pattern)
            .and_then(|start| {
                let content_start = start + attr_pattern.len();
                html[content_start..].split('"').next()
            })
            .filter(|s| !s.is_empty())
            .map(|s| s.to_string())
    };

    // 1. Check src attribute in img tag (most common case)
    if let Some(src_content) = extract_quoted_value(html, "src=\"") {
        let filename = extract_filename_from_path(&src_content);
        if filename.contains('.') && has_image_extension(&filename) {
            return Some(filename);
        }
    }

    // 2. Check title attribute
    if let Some(title_content) = extract_quoted_value(html, "title=\"") {
        if title_content.contains('.') && has_image_extension(&title_content) {
            return Some(title_content);
        }
    }

    // 3. Check alt attribute
    if let Some(alt_content) = extract_quoted_value(html, "alt=\"") {
        if alt_content.contains('.') && has_image_extension(&alt_content) {
            return Some(alt_content);
        }
    }

    // 4. Look for any filename-like strings with image extensions in the entire HTML
    const TRIM_CHARS: &[char] = &['"', '\'', '<', '>', '(', ')', ',', ';'];

    for word in html.split_whitespace() {
        if word.contains('.') {
            let clean_word = word.trim_matches(TRIM_CHARS);
            if has_image_extension(clean_word) {
                let filename = extract_filename_from_path(clean_word);
                return Some(filename);
            }
        }
    }

    None
}

/// Best-effort conversion of HTML clipboard contents to plain text.
///
/// This is intentionally lightweight (no external HTML parser dependency). It strips tags,
/// decodes a small set of common entities, and collapses whitespace.
pub fn strip_html_to_plain_text(html: &str) -> String {
    if html.trim().is_empty() {
        return String::new();
    }

    // Fast path: if there are no obvious tag/entity markers, treat as plain text.
    if !html.contains('<') && !html.contains('&') {
        return html.split_whitespace().collect::<Vec<_>>().join(" ");
    }

    fn decode_entity(entity: &str) -> Option<char> {
        match entity {
            "nbsp" => Some(' '),
            "amp" => Some('&'),
            "lt" => Some('<'),
            "gt" => Some('>'),
            "quot" => Some('"'),
            "apos" => Some('\''),
            "#39" => Some('\''),
            _ if entity.starts_with("#x") || entity.starts_with("#X") => {
                u32::from_str_radix(&entity[2..], 16)
                    .ok()
                    .and_then(char::from_u32)
            }
            _ if entity.starts_with('#') => {
                entity[1..].parse::<u32>().ok().and_then(char::from_u32)
            }
            _ => None,
        }
    }

    let mut out = String::with_capacity(html.len());
    let mut in_tag = false;
    let mut in_entity = false;
    let mut entity_buf = String::new();
    let mut last_was_space = false;

    for ch in html.chars() {
        if in_tag {
            if ch == '>' {
                in_tag = false;
                // Treat tags as word boundaries.
                if !last_was_space {
                    out.push(' ');
                    last_was_space = true;
                }
            }
            continue;
        }

        if in_entity {
            if ch == ';' {
                let decoded = decode_entity(entity_buf.as_str());
                if let Some(decoded) = decoded {
                    if decoded.is_whitespace() {
                        if !last_was_space {
                            out.push(' ');
                            last_was_space = true;
                        }
                    } else {
                        out.push(decoded);
                        last_was_space = false;
                    }
                } else {
                    // Unknown entity; keep it as-is (best effort).
                    if !last_was_space {
                        out.push(' ');
                    }
                    out.push('&');
                    out.push_str(entity_buf.as_str());
                    out.push(';');
                    out.push(' ');
                    last_was_space = true;
                }
                entity_buf.clear();
                in_entity = false;
                continue;
            }

            // Guard against extremely long/unterminated entities.
            if entity_buf.len() >= 24 {
                in_entity = false;
                entity_buf.clear();
                if !last_was_space {
                    out.push(' ');
                    last_was_space = true;
                }
                continue;
            }

            entity_buf.push(ch);
            continue;
        }

        match ch {
            '<' => {
                in_tag = true;
                // Ensure words on either side of tags don't get glued together.
                if !last_was_space && !out.is_empty() {
                    out.push(' ');
                    last_was_space = true;
                }
            }
            '&' => {
                in_entity = true;
                entity_buf.clear();
            }
            ch if ch.is_whitespace() => {
                if !last_was_space {
                    out.push(' ');
                    last_was_space = true;
                }
            }
            _ => {
                out.push(ch);
                last_was_space = false;
            }
        }
    }

    out.split_whitespace().collect::<Vec<_>>().join(" ")
}

/// Process clipboard image data, preserving original format or converting to PNG.
#[cfg(any(target_os = "linux", target_os = "freebsd", target_os = "windows"))]
pub fn process_clipboard_image(
    arboard_image: &arboard::ImageData,
    filename: Option<String>,
) -> Option<crate::clipboard::ImageData> {
    let result =
        try_preserve_original_format(&arboard_image.bytes, filename.clone()).or_else(|| {
            convert_raw_bitmap_to_png(
                arboard_image.width,
                arboard_image.height,
                arboard_image.bytes.to_vec(),
                filename,
            )
        });

    if result.is_none() {
        log::warn!(
            "Failed to process clipboard image: format preservation and PNG conversion both failed"
        );
    }

    result
}

/// Read image data from clipboard, checking for images before expensive filename extraction.
#[cfg(any(target_os = "linux", target_os = "freebsd", target_os = "windows"))]
pub fn read_images_from_clipboard(
    clipboard: &mut arboard::Clipboard,
    html_content: &Option<String>,
    text_content: &str,
) -> Option<Vec<crate::clipboard::ImageData>> {
    // First, quickly check if there are any images in the clipboard
    // This is a fast operation that avoids filename extraction overhead
    match clipboard.get().image() {
        Ok(arboard_image) => {
            // Images found! Now extract filename from clipboard content
            let filename = extract_filename_from_clipboard_content(html_content, text_content);

            // Process the image with the extracted filename
            match process_clipboard_image(&arboard_image, filename) {
                Some(image_data) => Some(vec![image_data]),
                None => {
                    log::warn!("Failed to process clipboard image: format detection and conversion both failed");
                    None
                }
            }
        }
        Err(arboard::Error::ContentNotAvailable) => None,
        Err(err) => {
            log::warn!("Unable to read image from clipboard: {err:?}");
            None
        }
    }
}

/// Try to preserve original image format using infer crate for detection.
#[cfg(any(target_os = "linux", target_os = "freebsd", target_os = "windows"))]
pub fn try_preserve_original_format(
    bytes: &[u8],
    filename: Option<String>,
) -> Option<crate::clipboard::ImageData> {
    if bytes.len() < MIN_IMAGE_HEADER_SIZE {
        return None;
    }

    // Use infer crate to detect the image format
    if let Some(kind) = infer::get(bytes) {
        // Check if it's a supported image format
        match kind.mime_type() {
            "image/png" | "image/jpeg" | "image/gif" | "image/webp" => {
                return Some(crate::clipboard::ImageData {
                    data: bytes.to_vec(),
                    mime_type: kind.mime_type().to_string(),
                    filename,
                });
            }
            _ => {}
        }
    }
    None
}

/// Converts RGBA bitmap data to PNG format, returns None on invalid dimensions/encoding.
#[cfg(any(target_os = "linux", target_os = "freebsd", target_os = "windows"))]
pub fn convert_raw_bitmap_to_png(
    width: usize,
    height: usize,
    bytes: Vec<u8>,
    filename: Option<String>,
) -> Option<crate::clipboard::ImageData> {
    // Validate dimensions before processing
    let width_u32 = match width.try_into() {
        Ok(w) => w,
        Err(e) => {
            log::warn!("Invalid width for PNG conversion: {width} - {e}");
            return None;
        }
    };

    let height_u32 = match height.try_into() {
        Ok(h) => h,
        Err(e) => {
            log::warn!("Invalid height for PNG conversion: {height} - {e}");
            return None;
        }
    };

    // Create RGBA image buffer from raw data
    // Note: arboard should already provide data in RGBA format
    let img_buffer =
        image::ImageBuffer::<image::Rgba<u8>, Vec<u8>>::from_raw(width_u32, height_u32, bytes)?;

    // Encode as PNG with optimized settings for speed
    let mut png_data = Vec::new();
    let mut cursor = std::io::Cursor::new(&mut png_data);

    // Use fast compression settings to reduce encoding time
    let encoder = image::codecs::png::PngEncoder::new_with_quality(
        &mut cursor,
        image::codecs::png::CompressionType::Fast,
        image::codecs::png::FilterType::NoFilter,
    );

    let encode_result = encoder.write_image(
        &img_buffer,
        width_u32,
        height_u32,
        image::ColorType::Rgba8.into(),
    );

    match encode_result {
        Ok(_) => Some(crate::clipboard::ImageData {
            data: png_data,
            mime_type: "image/png".to_string(),
            filename,
        }),
        Err(err) => {
            log::warn!("PNG encoding failed: {err:?}");
            None
        }
    }
}

pub fn get_image_filepaths_from_paths(paths: &[String]) -> Vec<String> {
    paths
        .iter()
        .filter(|path| has_image_extension(path))
        .cloned()
        .collect()
}

/// Create escaped file paths text string for insertion into terminal.
pub fn escaped_paths_str(
    paths: &[String],
    shell_family: Option<galaxy_util::path::ShellFamily>,
) -> String {
    // Handle regular file paths as text
    #[allow(unused_mut)]
    let mut input = paths
        .iter()
        .map(|path| match shell_family {
            Some(shell_family) => shell_family.escape(path.as_ref()),
            None => std::borrow::Cow::Borrowed(path.as_ref()),
        })
        .join(" ");

    // Append a space in case of back-to-back drag-drops.
    input.push(' ');

    input
}

#[cfg(test)]
#[path = "clipboard_utils_tests.rs"]
mod tests;