Macro is a unified workspace for teams: email, chat, docs, tasks, agents, calls, and CRM — @-linked together with shared AI memory.
TypeScript agentaiai-agentsall-in-onecrmcrm-systememaillinearmcpmessagingnotesnotionofficerustslackslack-alternativestartupstartup-toolstypescriptworkspace
//! Deflate (zlib streams) and the prediction Photoshop applies before it:
//! differences between neighboring samples along each row, 32-bit rows
//! split into byte planes first.
use crate::error::{PsdError, Result};
use miniz_oxide::inflate::TINFLStatus;
/// Photoshop-like compression level (zlib's default).
const LEVEL: u8 = 6;
/// Inflates exactly `size` bytes (raw deflate is accepted too, as some
/// writers leave out the zlib header).
pub(super) fn inflate(data: &[u8], size: usize) -> Result<Vec<u8>> {
let inflated = miniz_oxide::inflate::decompress_to_vec_zlib_with_limit(data, size)
.or_else(|e| match e.status {
TINFLStatus::HasMoreOutput => Ok(e.output),
_ => miniz_oxide::inflate::decompress_to_vec_with_limit(data, size),
})
.map_err(|_| PsdError::corrupt("deflate data is damaged"))?;
if inflated.len() != size {
return Err(PsdError::corrupt("deflate data is cut off"));
}
Ok(inflated)
}
/// Deflates into a zlib stream.
pub(super) fn deflate(data: &[u8]) -> Vec<u8> {
miniz_oxide::deflate::compress_to_vec_zlib(data, LEVEL)
}
/// Undoes prediction over rows of `width` samples at `depth` bits.
pub(super) fn unpredict(data: &mut [u8], width: usize, depth: u16) {
if width == 0 {
return;
}
match depth {
16 => {
for row in data.chunks_exact_mut(width * 2) {
let mut prev = 0u16;
for s in row.chunks_exact_mut(2) {
prev = prev.wrapping_add(u16::from_be_bytes([s[0], s[1]]));
s.copy_from_slice(&prev.to_be_bytes());
}
}
}
32 => {
let mut planes = vec![0; width * 4];
for row in data.chunks_exact_mut(width * 4) {
let mut prev = 0u8;
for (p, &b) in planes.iter_mut().zip(row.iter()) {
prev = prev.wrapping_add(b);
*p = prev;
}
for (x, sample) in row.chunks_exact_mut(4).enumerate() {
for (k, b) in sample.iter_mut().enumerate() {
*b = planes[k * width + x];
}
}
}
}
_ => {
for row in data.chunks_exact_mut(crate::channels::row_bytes(width as u32, depth)) {
let mut prev = 0u8;
for b in row {
prev = prev.wrapping_add(*b);
*b = prev;
}
}
}
}
}
/// Applies prediction over rows of `width` samples at `depth` bits.
pub(super) fn predict(data: &mut [u8], width: usize, depth: u16) {
if width == 0 {
return;
}
match depth {
16 => {
for row in data.chunks_exact_mut(width * 2) {
let mut prev = 0u16;
for s in row.chunks_exact_mut(2) {
let v = u16::from_be_bytes([s[0], s[1]]);
s.copy_from_slice(&v.wrapping_sub(prev).to_be_bytes());
prev = v;
}
}
}
32 => {
let mut planes = vec![0; width * 4];
for row in data.chunks_exact_mut(width * 4) {
for (x, sample) in row.chunks_exact(4).enumerate() {
for (k, &b) in sample.iter().enumerate() {
planes[k * width + x] = b;
}
}
let mut prev = 0u8;
for (b, &p) in row.iter_mut().zip(planes.iter()) {
*b = p.wrapping_sub(prev);
prev = p;
}
}
}
_ => {
for row in data.chunks_exact_mut(crate::channels::row_bytes(width as u32, depth)) {
let mut prev = 0u8;
for b in row {
let v = *b;
*b = v.wrapping_sub(prev);
prev = v;
}
}
}
}
}