1 | //! This module contains functionality for decompression. |
2 | |
3 | #[cfg (feature = "with-alloc" )] |
4 | use crate::alloc::{boxed::Box, vec, vec::Vec}; |
5 | use ::core::usize; |
6 | #[cfg (all(feature = "std" , feature = "with-alloc" ))] |
7 | use std::error::Error; |
8 | |
9 | pub mod core; |
10 | mod output_buffer; |
11 | pub mod stream; |
12 | use self::core::*; |
13 | |
14 | const TINFL_STATUS_FAILED_CANNOT_MAKE_PROGRESS: i32 = -4; |
15 | const TINFL_STATUS_BAD_PARAM: i32 = -3; |
16 | const TINFL_STATUS_ADLER32_MISMATCH: i32 = -2; |
17 | const TINFL_STATUS_FAILED: i32 = -1; |
18 | const TINFL_STATUS_DONE: i32 = 0; |
19 | const TINFL_STATUS_NEEDS_MORE_INPUT: i32 = 1; |
20 | const TINFL_STATUS_HAS_MORE_OUTPUT: i32 = 2; |
21 | |
22 | /// Return status codes. |
23 | #[repr (i8)] |
24 | #[derive (Debug, Copy, Clone, PartialEq, Eq, Hash)] |
25 | pub enum TINFLStatus { |
26 | /// More input data was expected, but the caller indicated that there was no more data, so the |
27 | /// input stream is likely truncated. |
28 | /// |
29 | /// This can't happen if you have provided the |
30 | /// [`TINFL_FLAG_HAS_MORE_INPUT`][core::inflate_flags::TINFL_FLAG_HAS_MORE_INPUT] flag to the |
31 | /// decompression. By setting that flag, you indicate more input exists but is not provided, |
32 | /// and so reaching the end of the input data without finding the end of the compressed stream |
33 | /// would instead return a [`NeedsMoreInput`][Self::NeedsMoreInput] status. |
34 | FailedCannotMakeProgress = TINFL_STATUS_FAILED_CANNOT_MAKE_PROGRESS as i8, |
35 | |
36 | /// The output buffer is an invalid size; consider the `flags` parameter. |
37 | BadParam = TINFL_STATUS_BAD_PARAM as i8, |
38 | |
39 | /// The decompression went fine, but the adler32 checksum did not match the one |
40 | /// provided in the header. |
41 | Adler32Mismatch = TINFL_STATUS_ADLER32_MISMATCH as i8, |
42 | |
43 | /// Failed to decompress due to invalid data. |
44 | Failed = TINFL_STATUS_FAILED as i8, |
45 | |
46 | /// Finished decompression without issues. |
47 | /// |
48 | /// This indicates the end of the compressed stream has been reached. |
49 | Done = TINFL_STATUS_DONE as i8, |
50 | |
51 | /// The decompressor needs more input data to continue decompressing. |
52 | /// |
53 | /// This occurs when there's no more consumable input, but the end of the stream hasn't been |
54 | /// reached, and you have supplied the |
55 | /// [`TINFL_FLAG_HAS_MORE_INPUT`][core::inflate_flags::TINFL_FLAG_HAS_MORE_INPUT] flag to the |
56 | /// decompressor. Had you not supplied that flag (which would mean you were asserting that you |
57 | /// believed all the data was available) you would have gotten a |
58 | /// [`FailedCannotMakeProcess`][Self::FailedCannotMakeProgress] instead. |
59 | NeedsMoreInput = TINFL_STATUS_NEEDS_MORE_INPUT as i8, |
60 | |
61 | /// There is still pending data that didn't fit in the output buffer. |
62 | HasMoreOutput = TINFL_STATUS_HAS_MORE_OUTPUT as i8, |
63 | } |
64 | |
65 | impl TINFLStatus { |
66 | pub fn from_i32(value: i32) -> Option<TINFLStatus> { |
67 | use self::TINFLStatus::*; |
68 | match value { |
69 | TINFL_STATUS_FAILED_CANNOT_MAKE_PROGRESS => Some(FailedCannotMakeProgress), |
70 | TINFL_STATUS_BAD_PARAM => Some(BadParam), |
71 | TINFL_STATUS_ADLER32_MISMATCH => Some(Adler32Mismatch), |
72 | TINFL_STATUS_FAILED => Some(Failed), |
73 | TINFL_STATUS_DONE => Some(Done), |
74 | TINFL_STATUS_NEEDS_MORE_INPUT => Some(NeedsMoreInput), |
75 | TINFL_STATUS_HAS_MORE_OUTPUT => Some(HasMoreOutput), |
76 | _ => None, |
77 | } |
78 | } |
79 | } |
80 | |
81 | /// Struct return when decompress_to_vec functions fail. |
82 | #[cfg (feature = "with-alloc" )] |
83 | #[derive (Debug)] |
84 | pub struct DecompressError { |
85 | /// Decompressor status on failure. See [TINFLStatus] for details. |
86 | pub status: TINFLStatus, |
87 | /// The currently decompressed data if any. |
88 | pub output: Vec<u8>, |
89 | } |
90 | |
91 | #[cfg (feature = "with-alloc" )] |
92 | impl alloc::fmt::Display for DecompressError { |
93 | #[cold ] |
94 | fn fmt(&self, f: &mut ::core::fmt::Formatter<'_>) -> ::core::fmt::Result { |
95 | f.write_str(data:match self.status { |
96 | TINFLStatus::FailedCannotMakeProgress => "Truncated input stream" , |
97 | TINFLStatus::BadParam => "Invalid output buffer size" , |
98 | TINFLStatus::Adler32Mismatch => "Adler32 checksum mismatch" , |
99 | TINFLStatus::Failed => "Invalid input data" , |
100 | TINFLStatus::Done => "" , // Unreachable |
101 | TINFLStatus::NeedsMoreInput => "Truncated input stream" , |
102 | TINFLStatus::HasMoreOutput => "Output size exceeded the specified limit" , |
103 | }) |
104 | } |
105 | } |
106 | |
107 | /// Implement Error trait only if std feature is requested as it requires std. |
108 | #[cfg (all(feature = "std" , feature = "with-alloc" ))] |
109 | impl Error for DecompressError {} |
110 | |
111 | #[cfg (feature = "with-alloc" )] |
112 | fn decompress_error(status: TINFLStatus, output: Vec<u8>) -> Result<Vec<u8>, DecompressError> { |
113 | Err(DecompressError { status, output }) |
114 | } |
115 | |
116 | /// Decompress the deflate-encoded data in `input` to a vector. |
117 | /// |
118 | /// NOTE: This function will not bound the output, so if the output is large enough it can result in an out of memory error. |
119 | /// It is therefore suggested to not use this for anything other than test programs, use the functions with a specified limit, or |
120 | /// ideally streaming decompression via the [flate2](https://github.com/alexcrichton/flate2-rs) library instead. |
121 | /// |
122 | /// Returns a [`Result`] containing the [`Vec`] of decompressed data on success, and a [struct][DecompressError] containing the status and so far decompressed data if any on failure. |
123 | #[inline ] |
124 | #[cfg (feature = "with-alloc" )] |
125 | pub fn decompress_to_vec(input: &[u8]) -> Result<Vec<u8>, DecompressError> { |
126 | decompress_to_vec_inner(input, flags:0, max_output_size:usize::max_value()) |
127 | } |
128 | |
129 | /// Decompress the deflate-encoded data (with a zlib wrapper) in `input` to a vector. |
130 | /// |
131 | /// NOTE: This function will not bound the output, so if the output is large enough it can result in an out of memory error. |
132 | /// It is therefore suggested to not use this for anything other than test programs, use the functions with a specified limit, or |
133 | /// ideally streaming decompression via the [flate2](https://github.com/alexcrichton/flate2-rs) library instead. |
134 | /// |
135 | /// Returns a [`Result`] containing the [`Vec`] of decompressed data on success, and a [struct][DecompressError] containing the status and so far decompressed data if any on failure. |
136 | #[inline ] |
137 | #[cfg (feature = "with-alloc" )] |
138 | pub fn decompress_to_vec_zlib(input: &[u8]) -> Result<Vec<u8>, DecompressError> { |
139 | decompress_to_vec_inner( |
140 | input, |
141 | flags:inflate_flags::TINFL_FLAG_PARSE_ZLIB_HEADER, |
142 | max_output_size:usize::max_value(), |
143 | ) |
144 | } |
145 | |
146 | /// Decompress the deflate-encoded data in `input` to a vector. |
147 | /// |
148 | /// The vector is grown to at most `max_size` bytes; if the data does not fit in that size, |
149 | /// the error [struct][DecompressError] will contain the status [`TINFLStatus::HasMoreOutput`] and the data that was decompressed on failure. |
150 | /// |
151 | /// As this function tries to decompress everything in one go, it's not ideal for general use outside of tests or where the output size is expected to be small. |
152 | /// It is suggested to use streaming decompression via the [flate2](https://github.com/alexcrichton/flate2-rs) library instead. |
153 | /// |
154 | /// Returns a [`Result`] containing the [`Vec`] of decompressed data on success, and a [struct][DecompressError] on failure. |
155 | #[inline ] |
156 | #[cfg (feature = "with-alloc" )] |
157 | pub fn decompress_to_vec_with_limit( |
158 | input: &[u8], |
159 | max_size: usize, |
160 | ) -> Result<Vec<u8>, DecompressError> { |
161 | decompress_to_vec_inner(input, flags:0, max_output_size:max_size) |
162 | } |
163 | |
164 | /// Decompress the deflate-encoded data (with a zlib wrapper) in `input` to a vector. |
165 | /// The vector is grown to at most `max_size` bytes; if the data does not fit in that size, |
166 | /// the error [struct][DecompressError] will contain the status [`TINFLStatus::HasMoreOutput`] and the data that was decompressed on failure. |
167 | /// |
168 | /// As this function tries to decompress everything in one go, it's not ideal for general use outside of tests or where the output size is expected to be small. |
169 | /// It is suggested to use streaming decompression via the [flate2](https://github.com/alexcrichton/flate2-rs) library instead. |
170 | /// |
171 | /// Returns a [`Result`] containing the [`Vec`] of decompressed data on success, and a [struct][DecompressError] on failure. |
172 | #[inline ] |
173 | #[cfg (feature = "with-alloc" )] |
174 | pub fn decompress_to_vec_zlib_with_limit( |
175 | input: &[u8], |
176 | max_size: usize, |
177 | ) -> Result<Vec<u8>, DecompressError> { |
178 | decompress_to_vec_inner(input, flags:inflate_flags::TINFL_FLAG_PARSE_ZLIB_HEADER, max_output_size:max_size) |
179 | } |
180 | |
181 | /// Backend of various to-[`Vec`] decompressions. |
182 | /// |
183 | /// Returns [`Vec`] of decompressed data on success and the [error struct][DecompressError] with details on failure. |
184 | #[cfg (feature = "with-alloc" )] |
185 | fn decompress_to_vec_inner( |
186 | mut input: &[u8], |
187 | flags: u32, |
188 | max_output_size: usize, |
189 | ) -> Result<Vec<u8>, DecompressError> { |
190 | let flags = flags | inflate_flags::TINFL_FLAG_USING_NON_WRAPPING_OUTPUT_BUF; |
191 | let mut ret: Vec<u8> = vec![0; input.len().saturating_mul(2).min(max_output_size)]; |
192 | |
193 | let mut decomp = Box::<DecompressorOxide>::default(); |
194 | |
195 | let mut out_pos = 0; |
196 | loop { |
197 | // Wrap the whole output slice so we know we have enough of the |
198 | // decompressed data for matches. |
199 | let (status, in_consumed, out_consumed) = |
200 | decompress(&mut decomp, input, &mut ret, out_pos, flags); |
201 | out_pos += out_consumed; |
202 | |
203 | match status { |
204 | TINFLStatus::Done => { |
205 | ret.truncate(out_pos); |
206 | return Ok(ret); |
207 | } |
208 | |
209 | TINFLStatus::HasMoreOutput => { |
210 | // in_consumed is not expected to be out of bounds, |
211 | // but the check eliminates a panicking code path |
212 | if in_consumed > input.len() { |
213 | return decompress_error(TINFLStatus::HasMoreOutput, ret); |
214 | } |
215 | input = &input[in_consumed..]; |
216 | |
217 | // if the buffer has already reached the size limit, return an error |
218 | if ret.len() >= max_output_size { |
219 | return decompress_error(TINFLStatus::HasMoreOutput, ret); |
220 | } |
221 | // calculate the new length, capped at `max_output_size` |
222 | let new_len = ret.len().saturating_mul(2).min(max_output_size); |
223 | ret.resize(new_len, 0); |
224 | } |
225 | |
226 | _ => return decompress_error(status, ret), |
227 | } |
228 | } |
229 | } |
230 | |
231 | /// Decompress one or more source slices from an iterator into the output slice. |
232 | /// |
233 | /// * On success, returns the number of bytes that were written. |
234 | /// * On failure, returns the failure status code. |
235 | /// |
236 | /// This will fail if the output buffer is not large enough, but in that case |
237 | /// the output buffer will still contain the partial decompression. |
238 | /// |
239 | /// * `out` the output buffer. |
240 | /// * `it` the iterator of input slices. |
241 | /// * `zlib_header` if the first slice out of the iterator is expected to have a |
242 | /// Zlib header. Otherwise the slices are assumed to be the deflate data only. |
243 | /// * `ignore_adler32` if the adler32 checksum should be calculated or not. |
244 | pub fn decompress_slice_iter_to_slice<'out, 'inp>( |
245 | out: &'out mut [u8], |
246 | it: impl Iterator<Item = &'inp [u8]>, |
247 | zlib_header: bool, |
248 | ignore_adler32: bool, |
249 | ) -> Result<usize, TINFLStatus> { |
250 | use self::core::inflate_flags::*; |
251 | |
252 | let mut it = it.peekable(); |
253 | let r = &mut DecompressorOxide::new(); |
254 | let mut out_pos = 0; |
255 | while let Some(in_buf) = it.next() { |
256 | let has_more = it.peek().is_some(); |
257 | let flags = { |
258 | let mut f = TINFL_FLAG_USING_NON_WRAPPING_OUTPUT_BUF; |
259 | if zlib_header { |
260 | f |= TINFL_FLAG_PARSE_ZLIB_HEADER; |
261 | } |
262 | if ignore_adler32 { |
263 | f |= TINFL_FLAG_IGNORE_ADLER32; |
264 | } |
265 | if has_more { |
266 | f |= TINFL_FLAG_HAS_MORE_INPUT; |
267 | } |
268 | f |
269 | }; |
270 | let (status, _input_read, bytes_written) = decompress(r, in_buf, out, out_pos, flags); |
271 | out_pos += bytes_written; |
272 | match status { |
273 | TINFLStatus::NeedsMoreInput => continue, |
274 | TINFLStatus::Done => return Ok(out_pos), |
275 | e => return Err(e), |
276 | } |
277 | } |
278 | // If we ran out of source slices without getting a `Done` from the |
279 | // decompression we can call it a failure. |
280 | Err(TINFLStatus::FailedCannotMakeProgress) |
281 | } |
282 | |
283 | #[cfg (all(test, feature = "with-alloc" ))] |
284 | mod test { |
285 | use super::{ |
286 | decompress_slice_iter_to_slice, decompress_to_vec_zlib, decompress_to_vec_zlib_with_limit, |
287 | DecompressError, TINFLStatus, |
288 | }; |
289 | const ENCODED: [u8; 20] = [ |
290 | 120, 156, 243, 72, 205, 201, 201, 215, 81, 168, 202, 201, 76, 82, 4, 0, 27, 101, 4, 19, |
291 | ]; |
292 | |
293 | #[test ] |
294 | fn decompress_vec() { |
295 | let res = decompress_to_vec_zlib(&ENCODED[..]).unwrap(); |
296 | assert_eq!(res.as_slice(), &b"Hello, zlib!" [..]); |
297 | } |
298 | |
299 | #[test ] |
300 | fn decompress_vec_with_high_limit() { |
301 | let res = decompress_to_vec_zlib_with_limit(&ENCODED[..], 100_000).unwrap(); |
302 | assert_eq!(res.as_slice(), &b"Hello, zlib!" [..]); |
303 | } |
304 | |
305 | #[test ] |
306 | fn fail_to_decompress_with_limit() { |
307 | let res = decompress_to_vec_zlib_with_limit(&ENCODED[..], 8); |
308 | match res { |
309 | Err(DecompressError { |
310 | status: TINFLStatus::HasMoreOutput, |
311 | .. |
312 | }) => (), // expected result |
313 | _ => panic!("Decompression output size limit was not enforced" ), |
314 | } |
315 | } |
316 | |
317 | #[test ] |
318 | fn test_decompress_slice_iter_to_slice() { |
319 | // one slice |
320 | let mut out = [0_u8; 12_usize]; |
321 | let r = |
322 | decompress_slice_iter_to_slice(&mut out, Some(&ENCODED[..]).into_iter(), true, false); |
323 | assert_eq!(r, Ok(12)); |
324 | assert_eq!(&out[..12], &b"Hello, zlib!" [..]); |
325 | |
326 | // some chunks at a time |
327 | for chunk_size in 1..13 { |
328 | // Note: because of https://github.com/Frommi/miniz_oxide/issues/110 our |
329 | // out buffer needs to have +1 byte available when the chunk size cuts |
330 | // the adler32 data off from the last actual data. |
331 | let mut out = [0_u8; 12_usize + 1]; |
332 | let r = |
333 | decompress_slice_iter_to_slice(&mut out, ENCODED.chunks(chunk_size), true, false); |
334 | assert_eq!(r, Ok(12)); |
335 | assert_eq!(&out[..12], &b"Hello, zlib!" [..]); |
336 | } |
337 | |
338 | // output buffer too small |
339 | let mut out = [0_u8; 3_usize]; |
340 | let r = decompress_slice_iter_to_slice(&mut out, ENCODED.chunks(7), true, false); |
341 | assert!(r.is_err()); |
342 | } |
343 | } |
344 | |