nom-exif 3.1.1

use crate::error::{nom_error_to_parsing_error_with_state, ParsingError, ParsingErrorState};
use crate::file::MediaMimeImage;
use crate::parser::{BufParser, ParsingState, ShareBuf};
use crate::raf::RafInfo;
use crate::slice::SubsliceRange;
use crate::{cr3, heif, jpeg, MediaParser};
pub use exif_exif::Exif;
use exif_exif::TIFF_HEADER_LEN;
use exif_iter::input_into_iter;
pub use exif_iter::{ExifEntry, ExifIter, ExifIterEntry, IfdIndex};
pub use gps::{GPSInfo, LatLng};
pub use tags::{ExifTag, TagOrCode};

use std::io::Read;
use std::ops::Range;

pub(crate) mod ifd;
pub(crate) use exif_exif::{check_exif_header, check_exif_header2, TiffHeader};
pub(crate) use travel::IfdHeaderTravel;

mod exif_exif;
mod exif_iter;
pub mod gps;
mod tags;
mod travel;

#[tracing::instrument(skip(reader, skip_by_seek))]
pub(crate) fn parse_exif_iter<R: Read>(
    parser: &mut MediaParser,
    mime_img: MediaMimeImage,
    reader: &mut R,
    skip_by_seek: crate::parser::SkipBySeekFn<R>,
) -> Result<ExifIter, crate::Error> {
    // For CR3 files, we need special handling to get all CMT blocks
    if mime_img == MediaMimeImage::Cr3 {
        return parse_cr3_exif_iter(parser, reader, skip_by_seek);
    }

    let out = parser.load_and_parse(reader, skip_by_seek, |buf, state| {
        extract_exif_range(mime_img, buf, state)
    })?;

    let has_track = match mime_img {
        MediaMimeImage::Jpeg => detect_motion_photo(parser, reader),
        _ => false,
    };

    range_to_iter(parser, out, has_track)
}

/// Demand-driven scan for a Pixel/Google Motion Photo signal in a JPEG
/// buffer that may not yet hold all APP segments.
///
/// `load_and_parse` only fills enough bytes to read the EXIF segment;
/// for JPEGs with a large EXIF (Pixel/Galaxy thumbnails routinely push
/// it past 30 KB) the XMP segment that carries `GCamera:MotionPhoto`
/// can sit just past the buffer's edge. Try the scan first; if it
/// reports `NeedMoreBytes`, pull another small chunk and retry. Cap
/// the total extra reads at `MAX_EXTRA` so a malformed file can't loop
/// forever.
fn detect_motion_photo<R: Read>(parser: &mut MediaParser, reader: &mut R) -> bool {
    use crate::parser::{Buf, BufParser};
    const CHUNK: usize = 8 * 1024;
    const MAX_EXTRA: usize = 256 * 1024;
    let mut extra = 0;
    loop {
        match jpeg::scan_motion_photo(parser.buffer()) {
            jpeg::MotionPhotoScan::Found(_) => return true,
            jpeg::MotionPhotoScan::NotPresent => return false,
            jpeg::MotionPhotoScan::NeedMoreBytes => {
                if extra >= MAX_EXTRA {
                    return false;
                }
                let want = CHUNK.min(MAX_EXTRA - extra);
                if parser.fill_buf(reader, want).is_err() {
                    return false;
                }
                extra += want;
            }
        }
    }
}

/// Special parser for CR3 files that extracts all CMT blocks (CMT1, CMT2, CMT3)
/// and adds them as additional TIFF blocks to the ExifIter.
#[tracing::instrument(skip(reader, skip_by_seek))]
fn parse_cr3_exif_iter<R: Read>(
    parser: &mut MediaParser,
    reader: &mut R,
    skip_by_seek: crate::parser::SkipBySeekFn<R>,
) -> Result<ExifIter, crate::Error> {
    // First, parse to get all CMT ranges
    let cmt_ranges = parser.load_and_parse(reader, skip_by_seek, |buf, _state| {
        cr3::extract_all_cmt_ranges(buf)
    })?;

    let Some(cmt_ranges) = cmt_ranges else {
        return Err(crate::Error::Malformed {
            kind: crate::error::MalformedKind::IsoBmffBox,
            message: "cr3: no CMT data found".into(),
        });
    };

    if cmt_ranges.ranges.is_empty() {
        return Err(crate::Error::Malformed {
            kind: crate::error::MalformedKind::IsoBmffBox,
            message: "cr3: no CMT ranges available".into(),
        });
    }

    tracing::debug!(
        cmt_count = cmt_ranges.ranges.len(),
        "Found CMT ranges in CR3 file"
    );

    // Get the first CMT range (CMT1) to create the primary ExifIter
    let (first_block_id, first_range) = &cmt_ranges.ranges[0];
    tracing::debug!(
        block_id = first_block_id,
        range = ?first_range,
        "Creating primary ExifIter from first CMT block"
    );

    // Take ownership of the parser's full buffer once. All CMT block ranges
    // are relative to the parser's position-adjusted buffer view; absolute
    // ranges within `full` are obtained by adding `position`.
    let (full, position) = parser.share_buf();

    // Invariant: parse_moov_box uses streaming::take(box_size) for every box,
    // so when extract_all_cmt_ranges returns Some, all child boxes (including
    // CMT1/2/3 data) are fully loaded into `full`. Step 3a hardens this from
    // a soft warning to a structured ParsingError, so reaching here always
    // means every range fits within `full`.
    debug_assert!(
        cmt_ranges
            .ranges
            .iter()
            .all(|(_, r)| r.end + position <= full.len()),
        "CMT range extends beyond loaded buffer; parse_moov_box invariant violated"
    );

    let primary_abs = (first_range.start + position)..(first_range.end + position);
    let primary_view = full.slice(primary_abs);
    let mut iter = input_into_iter(primary_view, None)?;

    for (block_id, range) in cmt_ranges.ranges.iter().skip(1) {
        if *block_id == "CMT3" {
            tracing::debug!(block_id, "Skipping CMT3 (MakerNotes) - proprietary format");
            continue;
        }
        let abs = (range.start + position)..(range.end + position);
        tracing::debug!(
            block_id,
            original_range = ?range,
            absolute_range = ?abs,
            "Adding additional CMT block"
        );
        iter.add_tiff_block(block_id.to_string(), full.slice(abs), None);
    }

    Ok(iter)
}

type ExifRangeResult = Result<Option<(Range<usize>, Option<TiffHeader>)>, ParsingErrorState>;

fn extract_exif_range(
    img: MediaMimeImage,
    buf: &[u8],
    state: Option<ParsingState>,
) -> ExifRangeResult {
    let (exif_data, state) = extract_exif_with_mime(img, buf, state)?;
    let header = state.and_then(|x| match x {
        ParsingState::TiffHeader(h) => Some(h),
        ParsingState::HeifExifSize(_) => None,
        ParsingState::Cr3ExifSize(_) => None,
    });
    Ok(exif_data
        .and_then(|x| buf.subslice_in_range(x))
        .map(|x| (x, header)))
}

fn range_to_iter(
    parser: &mut impl ShareBuf,
    out: Option<(Range<usize>, Option<TiffHeader>)>,
    has_embedded_track: bool,
) -> Result<ExifIter, crate::Error> {
    if let Some((range, header)) = out {
        tracing::debug!(?range, ?header, "Got Exif data");
        let (full, position) = parser.share_buf();
        let abs = (range.start + position)..(range.end + position);
        let view = full.slice(abs);
        let mut iter = input_into_iter(view, header)?;
        iter.set_has_embedded_track(has_embedded_track);
        Ok(iter)
    } else {
        tracing::debug!("Exif not found");
        Err(crate::Error::ExifNotFound)
    }
}

#[cfg(feature = "tokio")]
#[tracing::instrument(skip(parser, reader, skip_by_seek))]
pub(crate) async fn parse_exif_iter_async<P, R: AsyncRead + Unpin + Send>(
    parser: &mut P,
    mime_img: MediaMimeImage,
    reader: &mut R,
    skip_by_seek: crate::parser_async::AsyncSkipBySeekFn<R>,
) -> Result<ExifIter, crate::Error>
where
    P: crate::parser_async::AsyncBufParser + crate::parser::ShareBuf,
{
    let out = parser
        .load_and_parse(reader, skip_by_seek, |buf, state| {
            extract_exif_range(mime_img, buf, state)
        })
        .await?;

    let has_track = match mime_img {
        MediaMimeImage::Jpeg => detect_motion_photo_async(parser, reader).await,
        _ => false,
    };

    range_to_iter(parser, out, has_track)
}

/// Async twin of [`detect_motion_photo`].
#[cfg(feature = "tokio")]
async fn detect_motion_photo_async<P, R>(parser: &mut P, reader: &mut R) -> bool
where
    P: crate::parser_async::AsyncBufParser + crate::parser::Buf,
    R: AsyncRead + Unpin + Send,
{
    const CHUNK: usize = 8 * 1024;
    const MAX_EXTRA: usize = 256 * 1024;
    let mut extra = 0;
    loop {
        match jpeg::scan_motion_photo(parser.buffer()) {
            jpeg::MotionPhotoScan::Found(_) => return true,
            jpeg::MotionPhotoScan::NotPresent => return false,
            jpeg::MotionPhotoScan::NeedMoreBytes => {
                if extra >= MAX_EXTRA {
                    return false;
                }
                let want = CHUNK.min(MAX_EXTRA - extra);
                if parser.fill_buf(reader, want).await.is_err() {
                    return false;
                }
                extra += want;
            }
        }
    }
}

#[tracing::instrument(skip(buf))]
pub(crate) fn extract_exif_with_mime(
    img_type: crate::file::MediaMimeImage,
    buf: &[u8],
    state: Option<ParsingState>,
) -> Result<(Option<&[u8]>, Option<ParsingState>), ParsingErrorState> {
    let (exif_data, state) = match img_type {
        MediaMimeImage::Jpeg => jpeg::extract_exif_data(buf)
            .map(|res| (res.1, state.clone()))
            .map_err(|e| nom_error_to_parsing_error_with_state(e, state))?,
        MediaMimeImage::Heic | crate::file::MediaMimeImage::Heif => heif_extract_exif(state, buf)?,
        MediaMimeImage::Tiff => {
            let header = match state {
                Some(ParsingState::TiffHeader(ref h)) => h.to_owned(),
                None => {
                    let (_, header) = TiffHeader::parse(buf)
                        .map_err(|e| nom_error_to_parsing_error_with_state(e, None))?;
                    if header.ifd0_offset as usize > buf.len() {
                        let clear_and_skip =
                            ParsingError::Need(header.ifd0_offset as usize - TIFF_HEADER_LEN + 2);
                        let state = Some(ParsingState::TiffHeader(header));
                        return Err(ParsingErrorState::new(clear_and_skip, state));
                    }
                    header
                }
                _ => {
                    return Err(ParsingErrorState::new(
                        ParsingError::Failed("unexpected parsing state for tiff".into()),
                        None,
                    ))
                }
            };

            // full fill TIFF data
            tracing::debug!("full fill TIFF data");
            let mut iter = IfdHeaderTravel::new(
                buf,
                header.ifd0_offset as usize,
                TagOrCode::Unknown(0x2a),
                header.endian,
            );
            iter.travel_ifd(0)
                .map_err(|e| ParsingErrorState::new(e, state.clone()))?;
            tracing::debug!("full fill TIFF data done");

            (Some(buf), state)
        }
        MediaMimeImage::Raf => RafInfo::parse(buf)
            .map(|res| (res.1.exif_data, state.clone()))
            .map_err(|e| nom_error_to_parsing_error_with_state(e, state))?,
        MediaMimeImage::Cr3 => cr3_extract_exif(state, buf)?,
    };
    Ok((exif_data, state))
}

fn heif_extract_exif(
    state: Option<ParsingState>,
    buf: &[u8],
) -> Result<(Option<&[u8]>, Option<ParsingState>), ParsingErrorState> {
    heif::extract_exif_data(state, buf)
}

fn cr3_extract_exif(
    state: Option<ParsingState>,
    buf: &[u8],
) -> Result<(Option<&[u8]>, Option<ParsingState>), ParsingErrorState> {
    cr3::extract_exif_data(state, buf)
}

#[cfg(feature = "tokio")]
use tokio::io::AsyncRead;

#[cfg(test)]
mod tests {
    use crate::{
        exif::gps::{Altitude, LatRef, LonRef, Speed},
        file::MediaMimeImage,
        testkit::read_sample,
        values::URational,
    };
    use test_case::test_case;

    use super::*;

    #[test_case(
        "exif.jpg",
        LatRef::North,
        LatLng::new(
            URational::new(22, 1),
            URational::new(31, 1),
            URational::new(5208, 100)
        ),
        LonRef::East,
        LatLng::new(
            URational::new(114, 1),
            URational::new(1, 1),
            URational::new(1733, 100)
        ),
        Altitude::AboveSeaLevel(URational::new(0, 1)),
        None
    )]
    fn gps_info(
        path: &str,
        latitude_ref: LatRef,
        latitude: LatLng,
        longitude_ref: LonRef,
        longitude: LatLng,
        altitude: Altitude,
        speed: Option<Speed>,
    ) {
        let _ = tracing_subscriber::fmt().with_test_writer().try_init();

        let buf = read_sample(path).unwrap();
        let (data, _) = extract_exif_with_mime(MediaMimeImage::Jpeg, &buf, None).unwrap();
        let data = data.unwrap();

        let subslice_in_range = buf.subslice_in_range(data).unwrap();
        let iter = input_into_iter(bytes::Bytes::from(buf).slice(subslice_in_range), None).unwrap();
        let exif: Exif = iter.into();

        let gps = exif.gps_info().unwrap();
        assert_eq!(
            *gps,
            GPSInfo {
                latitude_ref,
                latitude,
                longitude_ref,
                longitude,
                altitude,
                speed,
            }
        )
    }
}