armature-core 0.9.0

//! Form processing and multipart support

use crate::Error;
use serde::de::DeserializeOwned;
use std::collections::HashMap;

/// Parse URL-encoded form data
pub fn parse_form<T: DeserializeOwned>(body: &[u8]) -> Result<T, Error> {
    serde_urlencoded::from_bytes(body)
        .map_err(|e| Error::BadRequest(format!("Failed to parse form data: {}", e)))
}

/// Parse URL-encoded form data into a HashMap
pub fn parse_form_map(body: &[u8]) -> Result<HashMap<String, String>, Error> {
    let form_data: Vec<(String, String)> = serde_urlencoded::from_bytes(body)
        .map_err(|e| Error::BadRequest(format!("Failed to parse form data: {}", e)))?;

    Ok(form_data.into_iter().collect())
}

/// Multipart form field
#[derive(Debug, Clone)]
pub struct FormField {
    /// Field name
    pub name: String,

    /// Field value (for text fields)
    pub value: Option<String>,

    /// File data (for file fields)
    pub file: Option<FormFile>,
}

/// Uploaded file data
#[derive(Debug, Clone)]
pub struct FormFile {
    /// Original filename
    pub filename: String,

    /// Content type (MIME type)
    pub content_type: String,

    /// File size in bytes
    pub size: usize,

    /// File data
    pub data: Vec<u8>,
}

impl FormFile {
    /// Create a new form file
    pub fn new(filename: String, content_type: String, data: Vec<u8>) -> Self {
        let size = data.len();
        Self {
            filename,
            content_type,
            size,
            data,
        }
    }

    /// Get file extension
    pub fn extension(&self) -> Option<&str> {
        self.filename.rsplit('.').next()
    }

    /// Check if file is an image
    pub fn is_image(&self) -> bool {
        self.content_type.starts_with("image/")
    }

    /// Check if file size exceeds limit
    pub fn exceeds_size(&self, max_bytes: usize) -> bool {
        self.size > max_bytes
    }

    /// Save file to disk
    pub fn save_to(&self, path: &str) -> Result<(), Error> {
        std::fs::write(path, &self.data)
            .map_err(|e| Error::Internal(format!("Failed to save file: {}", e)))
    }

    /// Save file to disk asynchronously
    pub async fn save_to_async(&self, path: &str) -> Result<(), Error> {
        tokio::fs::write(path, &self.data)
            .await
            .map_err(|e| Error::Internal(format!("Failed to save file: {}", e)))
    }
}

/// Save multiple files in parallel
///
/// This function saves multiple uploaded files concurrently, providing
/// significant performance improvements over sequential saves.
///
/// # Arguments
///
/// * `files` - Vector of tuples containing (file, destination_path)
///
/// # Returns
///
/// Returns a vector of saved file paths in the same order as input.
///
/// # Performance
///
/// - **Sequential:** O(n * disk_write_time)
/// - **Parallel:** O(max(disk_write_times))
/// - **Speedup:** 5-10x for batch file uploads
///
/// # Examples
///
/// ```no_run
/// # use armature_core::form::*;
/// # async fn example(files: Vec<FormFile>) -> Result<(), armature_core::Error> {
/// // Save 10 files in parallel (5-10x faster)
/// let file_paths: Vec<_> = files.iter()
///     .enumerate()
///     .map(|(i, file)| (file, format!("uploads/file_{}.dat", i)))
///     .collect();
///
/// let saved = save_files_parallel(file_paths).await?;
/// println!("Saved {} files", saved.len());
/// # Ok(())
/// # }
/// ```
pub async fn save_files_parallel(files: Vec<(&FormFile, String)>) -> Result<Vec<String>, Error> {
    use tokio::task::JoinSet;

    let mut set = JoinSet::new();

    for (file, path) in files {
        let data = file.data.clone();
        let path_clone = path.clone();

        set.spawn(async move {
            tokio::fs::write(&path_clone, &data)
                .await
                .map_err(|e| Error::Internal(format!("Failed to save file: {}", e)))?;
            Ok::<_, Error>(path_clone)
        });
    }

    let mut saved_paths = Vec::new();
    while let Some(result) = set.join_next().await {
        saved_paths.push(result.map_err(|e| Error::Internal(e.to_string()))??);
    }

    Ok(saved_paths)
}

/// Multipart form data parser
pub struct MultipartParser {
    boundary: String,
}

impl MultipartParser {
    /// Create a new multipart parser from Content-Type header
    pub fn from_content_type(content_type: &str) -> Result<Self, Error> {
        // Extract boundary from Content-Type header
        // Example: "multipart/form-data; boundary=----WebKitFormBoundary7MA4YWxkTrZu0gW"
        let boundary = content_type
            .split(';')
            .find_map(|part| {
                let part = part.trim();
                if part.starts_with("boundary=") {
                    Some(
                        part.trim_start_matches("boundary=")
                            .trim_matches('"')
                            .to_string(),
                    )
                } else {
                    None
                }
            })
            .ok_or_else(|| Error::BadRequest("Missing boundary in Content-Type".to_string()))?;

        Ok(Self { boundary })
    }

    /// Parse multipart form data
    pub fn parse(&self, body: &[u8]) -> Result<Vec<FormField>, Error> {
        let mut fields = Vec::new();
        let boundary_marker = format!("--{}", self.boundary);
        let delimiter = boundary_marker.as_bytes();

        // Locate every boundary delimiter with SIMD-accelerated byte search.
        let finder = memchr::memmem::Finder::new(delimiter);
        let positions: Vec<usize> = finder.find_iter(body).collect();

        // Each part lives between two consecutive delimiters.
        for pair in positions.windows(2) {
            let mut part = &body[pair[0] + delimiter.len()..pair[1]];

            // The closing delimiter ("--boundary--") is not a part.
            if part.starts_with(b"--") {
                continue;
            }

            // Strip the CRLF terminating the boundary line...
            if let Some(rest) = part.strip_prefix(b"\r\n") {
                part = rest;
            } else if let Some(rest) = part.strip_prefix(b"\n") {
                part = rest;
            }

            // ...and the CRLF preceding the next boundary.
            if let Some(rest) = part.strip_suffix(b"\r\n") {
                part = rest;
            } else if let Some(rest) = part.strip_suffix(b"\n") {
                part = rest;
            }

            // Parse each part
            if let Some(field) = self.parse_part(part)? {
                fields.push(field);
            }
        }

        Ok(fields)
    }

    /// Parse a single multipart part
    fn parse_part(&self, part: &[u8]) -> Result<Option<FormField>, Error> {
        if part.is_empty() {
            return Ok(None);
        }

        // Split the header block from the body at the first blank line.
        // The body must stay a raw byte slice: file uploads are arbitrary binary.
        let (header_block, content) = match memchr::memmem::find(part, b"\r\n\r\n") {
            Some(pos) => (&part[..pos], &part[pos + 4..]),
            None => match memchr::memmem::find(part, b"\n\n") {
                Some(pos) => (&part[..pos], &part[pos + 2..]),
                None => (part, &part[part.len()..]),
            },
        };

        // Headers are ASCII per RFC 7578, so parsing them as UTF-8 is safe.
        let headers = String::from_utf8_lossy(header_block);

        // Parse headers
        let mut name = None;
        let mut filename = None;
        let mut content_type = None;

        for line in headers.lines() {
            if line.starts_with("Content-Disposition:") {
                // Parse name and filename
                for attr in line.split(';') {
                    let attr = attr.trim();
                    if attr.starts_with("name=") {
                        name = Some(
                            attr.trim_start_matches("name=")
                                .trim_matches('"')
                                .to_string(),
                        );
                    } else if attr.starts_with("filename=") {
                        filename = Some(
                            attr.trim_start_matches("filename=")
                                .trim_matches('"')
                                .to_string(),
                        );
                    }
                }
            } else if line.starts_with("Content-Type:") {
                content_type = Some(line.trim_start_matches("Content-Type:").trim().to_string());
            }
        }

        let name = name.ok_or_else(|| Error::BadRequest("Missing field name".to_string()))?;

        // Create field
        if let Some(filename) = filename {
            // File field: keep the body bytes exactly as uploaded
            let file = FormFile::new(
                filename,
                content_type.unwrap_or_else(|| "application/octet-stream".to_string()),
                content.to_vec(),
            );
            Ok(Some(FormField {
                name,
                value: None,
                file: Some(file),
            }))
        } else {
            // Text field
            Ok(Some(FormField {
                name,
                value: Some(String::from_utf8_lossy(content).into_owned()),
                file: None,
            }))
        }
    }

    /// Convert parsed fields to HashMap
    pub fn to_map(fields: Vec<FormField>) -> HashMap<String, String> {
        fields
            .into_iter()
            .filter_map(|field| field.value.map(|value| (field.name, value)))
            .collect()
    }

    /// Get files from parsed fields
    pub fn get_files(fields: &[FormField]) -> Vec<(String, &FormFile)> {
        fields
            .iter()
            .filter_map(|field| field.file.as_ref().map(|file| (field.name.clone(), file)))
            .collect()
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_parse_form_map() {
        let body = b"name=John+Doe&email=john%40example.com&age=30";
        let form = parse_form_map(body).unwrap();

        assert_eq!(form.get("name"), Some(&"John Doe".to_string()));
        assert_eq!(form.get("email"), Some(&"john@example.com".to_string()));
        assert_eq!(form.get("age"), Some(&"30".to_string()));
    }

    #[test]
    fn test_form_file_extension() {
        let file = FormFile::new(
            "document.pdf".to_string(),
            "application/pdf".to_string(),
            vec![1, 2, 3],
        );

        assert_eq!(file.extension(), Some("pdf"));
    }

    #[test]
    fn test_form_file_is_image() {
        let image = FormFile::new("photo.jpg".to_string(), "image/jpeg".to_string(), vec![]);
        assert!(image.is_image());

        let doc = FormFile::new("doc.pdf".to_string(), "application/pdf".to_string(), vec![]);
        assert!(!doc.is_image());
    }

    #[test]
    fn test_form_file_size_check() {
        let file = FormFile::new(
            "file.txt".to_string(),
            "text/plain".to_string(),
            vec![0; 1024], // 1KB
        );

        assert!(!file.exceeds_size(2048)); // 2KB limit
        assert!(file.exceeds_size(512)); // 512B limit
    }

    #[test]
    fn test_multipart_parser_from_content_type() {
        let content_type = "multipart/form-data; boundary=----WebKitFormBoundary7MA4YWxkTrZu0gW";
        let parser = MultipartParser::from_content_type(content_type).unwrap();

        assert_eq!(parser.boundary, "----WebKitFormBoundary7MA4YWxkTrZu0gW");
    }

    #[test]
    fn test_multipart_binary_roundtrip() {
        // Non-UTF-8 bytes, embedded CRLFs, and leading/trailing whitespace
        // must all survive the parse exactly as uploaded.
        let file_data: Vec<u8> = vec![
            b' ', b'\t', 0x00, 0xFF, 0xFE, b'\r', b'\n', 0x80, 0xC3, 0x01, b'\n', b' ',
        ];

        let mut body = Vec::new();
        body.extend_from_slice(b"--XBOUNDARY\r\n");
        body.extend_from_slice(
            b"Content-Disposition: form-data; name=\"file\"; filename=\"blob.bin\"\r\n",
        );
        body.extend_from_slice(b"Content-Type: application/octet-stream\r\n\r\n");
        body.extend_from_slice(&file_data);
        body.extend_from_slice(b"\r\n--XBOUNDARY\r\n");
        body.extend_from_slice(b"Content-Disposition: form-data; name=\"note\"\r\n\r\n");
        body.extend_from_slice(b"hello world");
        body.extend_from_slice(b"\r\n--XBOUNDARY--\r\n");

        let parser =
            MultipartParser::from_content_type("multipart/form-data; boundary=XBOUNDARY").unwrap();
        let fields = parser.parse(&body).unwrap();

        assert_eq!(fields.len(), 2);

        let file = fields[0].file.as_ref().unwrap();
        assert_eq!(fields[0].name, "file");
        assert_eq!(file.filename, "blob.bin");
        assert_eq!(file.content_type, "application/octet-stream");
        assert_eq!(file.data, file_data);
        assert_eq!(file.size, file_data.len());

        assert_eq!(fields[1].name, "note");
        assert_eq!(fields[1].value.as_deref(), Some("hello world"));
    }

    #[test]
    fn test_multipart_parser_missing_boundary() {
        let content_type = "multipart/form-data";
        let result = MultipartParser::from_content_type(content_type);

        assert!(result.is_err());
    }
}