form_net/multipart/
parser.rs

1//! The sans-IO multipart parser: a state machine over the delimiter matcher
2//! and the header reader.
3
4use core::fmt;
5
6use crate::error::{EncodingError, FormError, Limit, Location, PolicyRefusal, SyntaxError};
7use crate::limits::{CHARSET_CAPTURE, EmptyFilePolicy, Limits, Policy, TokenPlacement};
8use crate::multipart::boundary::Boundary;
9use crate::multipart::filename::FileName;
10use crate::multipart::header::{Areas, BlockScan, HeadFacts, Text, read_block};
11
12/// Whether a part is a text field or a file.
13///
14/// # Examples
15///
16/// ```
17/// # let body = b"--X\r\nContent-Disposition: form-data; name=a\r\n\r\n1\r\n--X--";
18/// let mp = form_net::parse_multipart(b"multipart/form-data; boundary=X", body)?;
19/// assert_eq!(mp.parts()[0].kind(), form_net::PartKind::Field);
20/// # Ok::<(), form_net::FormError>(())
21/// ```
22#[non_exhaustive]
23#[derive(Clone, Copy, Debug, PartialEq, Eq)]
24pub enum PartKind {
25    /// A text field: no `filename` parameter.
26    Field,
27    /// A file: a `filename` parameter (even an empty one) or an accepted
28    /// `filename*`.
29    File,
30}
31
32/// The `charset` a part declared.
33///
34/// `Unspecified` means no `charset` parameter: UTF-8 is assumed unless a
35/// `_charset_` field before the part said otherwise. `Other` is any label
36/// that is not UTF-8 or US-ASCII; its text is refused by every text accessor.
37///
38/// # Examples
39///
40/// ```
41/// use form_net::Charset;
42///
43/// # let body = b"--X\r\nContent-Disposition: form-data; name=a\r\nContent-Type: text/plain; charset=UTF-8\r\n\r\n1\r\n--X--";
44/// # let mp = form_net::parse_multipart(b"multipart/form-data; boundary=X", body)?;
45/// assert_eq!(mp.parts()[0].content_type().unwrap().charset(), Charset::Utf8);
46/// # Ok::<(), form_net::FormError>(())
47/// ```
48#[non_exhaustive]
49#[derive(Clone, Copy, Debug, PartialEq, Eq)]
50pub enum Charset {
51    /// `utf-8`.
52    Utf8,
53    /// `us-ascii`.
54    Ascii,
55    /// No `charset` parameter.
56    Unspecified,
57    /// Any other label.
58    Other,
59}
60
61impl Charset {
62    pub(crate) fn from_label(label: &[u8]) -> Charset {
63        if label.eq_ignore_ascii_case(b"utf-8") {
64            Charset::Utf8
65        } else if label.eq_ignore_ascii_case(b"us-ascii") {
66            Charset::Ascii
67        } else {
68            Charset::Other
69        }
70    }
71}
72
73/// One step of the parser: how many input bytes it consumed, and what
74/// happened.
75#[non_exhaustive]
76#[derive(Debug)]
77pub struct Step<'p> {
78    /// Bytes of the input consumed; pass the rest (and later new bytes) to
79    /// the next call.
80    pub consumed: usize,
81    /// What happened.
82    pub event: Event<'p>,
83}
84
85/// What the parser found.
86///
87/// Per part the order is always `PartStart`, any number of `Data` (none for
88/// an empty part), `PartEnd`. A `Data` slice is never empty. After the last
89/// part comes `End`. `NeedInput` can appear anywhere between.
90///
91/// `Debug` prints the length of `Data`, never the bytes.
92#[non_exhaustive]
93pub enum Event<'p> {
94    /// A part's headers were read and checked.
95    PartStart(PartHead<'p>),
96    /// Bytes of the current part's data.
97    Data(&'p [u8]),
98    /// The current part ended.
99    PartEnd(PartSummary),
100    /// The close delimiter was read: the form is complete.
101    End,
102    /// Every byte given was consumed; give more.
103    NeedInput,
104}
105
106impl fmt::Debug for Event<'_> {
107    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
108        match self {
109            Self::PartStart(head) => f.debug_tuple("PartStart").field(head).finish(),
110            Self::Data(bytes) => write!(f, "Data({} bytes)", bytes.len()),
111            Self::PartEnd(summary) => f.debug_tuple("PartEnd").field(summary).finish(),
112            Self::End => f.write_str("End"),
113            Self::NeedInput => f.write_str("NeedInput"),
114        }
115    }
116}
117
118/// What one part was.
119#[non_exhaustive]
120#[derive(Clone, Copy, Debug, PartialEq, Eq)]
121pub struct PartSummary {
122    /// The part index, from 0.
123    pub index: u32,
124    /// Field or file.
125    pub kind: PartKind,
126    /// The length of the part's data.
127    pub bytes: u64,
128    /// Whether the part was skipped (by `skip_part` or an inspector).
129    pub skipped: bool,
130}
131
132/// The validated header block of one part, borrowed.
133///
134/// A head is only produced after every header budget and rule passed.
135#[derive(Clone, Copy)]
136pub struct PartHead<'p> {
137    index: u32,
138    block: &'p [u8],
139    name: &'p str,
140    name_raw: &'p [u8],
141    kind: PartKind,
142    filename: Option<FileName<'p>>,
143    content_type: Option<&'p [u8]>,
144    charset: Charset,
145}
146
147impl fmt::Debug for PartHead<'_> {
148    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
149        f.debug_struct("PartHead")
150            .field("index", &self.index)
151            .field("name", &self.name)
152            .field("kind", &self.kind)
153            .field("filename", &self.filename)
154            .finish_non_exhaustive()
155    }
156}
157
158impl<'p> PartHead<'p> {
159    /// A head over bytes kept elsewhere (an owned copy).
160    #[cfg(feature = "alloc")]
161    #[allow(clippy::too_many_arguments)]
162    pub(crate) const fn from_parts(
163        index: u32,
164        block: &'p [u8],
165        name: &'p str,
166        name_raw: &'p [u8],
167        kind: PartKind,
168        filename: Option<FileName<'p>>,
169        content_type: Option<&'p [u8]>,
170        charset: Charset,
171    ) -> Self {
172        Self {
173            index,
174            block,
175            name,
176            name_raw,
177            kind,
178            filename,
179            content_type,
180            charset,
181        }
182    }
183
184    /// The header block, the final CRLF CRLF included.
185    #[cfg(feature = "alloc")]
186    pub(crate) const fn block(&self) -> &'p [u8] {
187        self.block
188    }
189
190    /// The part index, from 0, in body order.
191    #[must_use]
192    pub const fn index(&self) -> u32 {
193        self.index
194    }
195
196    /// The field name, with `%0A`, `%0D`, `%22` decoded to LF, CR, `"` (no
197    /// other escape is decoded). Always valid UTF-8 and never empty.
198    #[must_use]
199    pub const fn name(&self) -> &'p str {
200        self.name
201    }
202
203    /// The field name as sent.
204    #[must_use]
205    pub const fn name_raw(&self) -> &'p [u8] {
206        self.name_raw
207    }
208
209    /// Field or file.
210    #[must_use]
211    pub const fn kind(&self) -> PartKind {
212        self.kind
213    }
214
215    /// Whether this is a file part.
216    #[must_use]
217    pub const fn is_file(&self) -> bool {
218        matches!(self.kind, PartKind::File)
219    }
220
221    /// The client's file name, for file parts.
222    #[must_use]
223    pub const fn filename(&self) -> Option<FileName<'p>> {
224        self.filename
225    }
226
227    /// The declared `Content-Type`: a hint, never trusted.
228    #[must_use]
229    pub fn content_type(&self) -> Option<ContentType<'p>> {
230        self.content_type.and_then(ContentType::parse)
231    }
232
233    /// The part's own `charset` parameter (`Unspecified` when absent).
234    #[must_use]
235    pub const fn charset(&self) -> Charset {
236        self.charset
237    }
238
239    /// Every header line, in order, as raw `(name, value)` bytes.
240    #[must_use]
241    pub fn headers(&self) -> Headers<'p> {
242        Headers { rest: self.block }
243    }
244
245    /// The size of the header block, the final CRLF CRLF included.
246    #[must_use]
247    pub const fn header_bytes(&self) -> u32 {
248        self.block.len() as u32
249    }
250}
251
252/// A part's declared `Content-Type`, already checked. **A hint**: the client
253/// chose it.
254///
255/// # Examples
256///
257/// ```
258/// # let body = b"--X\r\nContent-Disposition: form-data; name=f; filename=a.png\r\nContent-Type: Image/PNG; q=\"x y\"\r\n\r\n\r\n--X--";
259/// # let mp = form_net::parse_multipart(b"multipart/form-data; boundary=X", body)?;
260/// let ct = mp.file("f").unwrap().content_type().unwrap();
261/// assert!(ct.is("image", "png"));
262/// assert_eq!((ct.type_(), ct.subtype()), ("Image", "PNG"));
263/// assert_eq!(ct.parameter("Q"), Some(&b"x y"[..]));
264/// # Ok::<(), form_net::FormError>(())
265/// ```
266#[derive(Clone, Copy)]
267pub struct ContentType<'p> {
268    raw: &'p [u8],
269    media: parse_net::MediaType<'p>,
270}
271
272impl fmt::Debug for ContentType<'_> {
273    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
274        f.debug_struct("ContentType")
275            .field("type", &self.type_())
276            .field("subtype", &self.subtype())
277            .finish_non_exhaustive()
278    }
279}
280
281impl<'p> ContentType<'p> {
282    pub(crate) fn parse(raw: &'p [u8]) -> Option<ContentType<'p>> {
283        parse_net::media::parse(raw, &crate::kind::HEADER_LIMITS)
284            .ok()
285            .map(|media| ContentType { raw, media })
286    }
287
288    /// The value as sent.
289    #[must_use]
290    pub const fn raw(&self) -> &'p [u8] {
291        self.raw
292    }
293
294    /// The type, as sent.
295    #[must_use]
296    pub fn type_(&self) -> &'p str {
297        core::str::from_utf8(self.media.type_()).unwrap_or("")
298    }
299
300    /// The subtype, as sent.
301    #[must_use]
302    pub fn subtype(&self) -> &'p str {
303        core::str::from_utf8(self.media.subtype()).unwrap_or("")
304    }
305
306    /// Whether the type and subtype are these, ASCII case-insensitively.
307    #[must_use]
308    pub fn is(&self, type_: &str, subtype: &str) -> bool {
309        self.media.is(type_, subtype)
310    }
311
312    /// A parameter's value by name (ASCII case-insensitive), without
313    /// surrounding quotes; backslashes are literal, as browsers send them.
314    #[must_use]
315    pub fn parameter(&self, name: &str) -> Option<&'p [u8]> {
316        self.media
317            .params()
318            .get(name)
319            .map(|w| crate::kind::word_raw(&w))
320    }
321
322    /// The `charset` parameter.
323    #[must_use]
324    pub fn charset(&self) -> Charset {
325        self.parameter("charset")
326            .map_or(Charset::Unspecified, Charset::from_label)
327    }
328}
329
330/// Every header line of a part, in order, as raw `(name, value)` bytes with
331/// the value trimmed.
332///
333/// `Debug` prints the header names only: values carry file names.
334#[derive(Clone)]
335pub struct Headers<'p> {
336    rest: &'p [u8],
337}
338
339impl fmt::Debug for Headers<'_> {
340    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
341        let mut list = f.debug_list();
342        for (name, _) in self.clone() {
343            list.entry(&core::str::from_utf8(name).unwrap_or("?"));
344        }
345        list.finish()
346    }
347}
348
349impl<'p> Iterator for Headers<'p> {
350    type Item = (&'p [u8], &'p [u8]);
351
352    fn next(&mut self) -> Option<Self::Item> {
353        let lf = self.rest.iter().position(|&b| b == b'\n')?;
354        let line = self.rest.get(..lf.saturating_sub(1)).unwrap_or(&[]);
355        self.rest = self.rest.get(lf.wrapping_add(1)..).unwrap_or(&[]);
356        if line.is_empty() {
357            self.rest = &[];
358            return None;
359        }
360        let colon = line.iter().position(|&b| b == b':').unwrap_or(line.len());
361        let name = line.get(..colon).unwrap_or(&[]);
362        let value = line.get(colon.wrapping_add(1)..).unwrap_or(&[]);
363        Some((name, parse_net::text::trim_ows(value)))
364    }
365}
366
367#[derive(Clone, Copy, Debug, PartialEq, Eq)]
368enum State {
369    Preamble,
370    /// After a delimiter: expecting `--`, padding, or CRLF.
371    AfterDelimiter(After),
372    Headers,
373    Body,
374    /// The delimiter that ends the current part was found; `PartEnd` is due.
375    EndOfPart,
376    Epilogue,
377    Poisoned,
378}
379
380#[derive(Clone, Copy, Debug, PartialEq, Eq)]
381enum After {
382    Start,
383    Dash,
384    Padding(u8),
385    Cr,
386}
387
388/// The token field's progress.
389#[derive(Clone, Copy, Debug, PartialEq, Eq)]
390enum Token {
391    NotSeen,
392    Capturing,
393    Captured,
394}
395
396const SKIPPING: u8 = 1;
397const IS_TOKEN: u8 = 2;
398const IS_CHARSET: u8 = 4;
399const DEFERRED: u8 = 8;
400
401/// Where a head's block is.
402#[derive(Clone, Copy, Debug)]
403pub(crate) enum Source {
404    /// In the input, from this index.
405    Input(usize),
406    /// In the parser buffer, from 0.
407    Buffer,
408}
409
410/// What a call to `advance` produced, as positions only.
411#[derive(Clone, Copy, Debug)]
412pub(crate) enum Outcome {
413    PartStart(HeadFacts, Source),
414    /// `input[start..end]`.
415    DataInput(usize, usize),
416    /// The first bytes of the delimiter.
417    DataDelimiter(usize),
418    PartEnd(PartSummary),
419    End,
420    NeedInput,
421}
422
423/// The sans-IO `multipart/form-data` parser.
424///
425/// Feed it bytes as they arrive, in chunks of any size split at any byte;
426/// it returns one [`Event`] per call and how many bytes it consumed. The
427/// events are identical for every split of the same body. It never
428/// allocates after construction and never rescans a byte: the delimiter
429/// search is a streaming Knuth-Morris-Pratt automaton with an eight-byte skip
430/// to the next CR.
431///
432/// The buffer holds the header block, the decoded field name, the decoded
433/// `filename*` (under `AcceptAlone`), the token, and the `_charset_` value:
434/// [`Limits::multipart_buffer_len`] bytes. It is wiped on `reset` and on
435/// drop.
436///
437/// # Examples
438///
439/// ```
440/// use form_net::{Event, FormKind, Limits, MultipartParser, Policy};
441///
442/// let content_type = b"multipart/form-data; boundary=X";
443/// let body = b"--X\r\nContent-Disposition: form-data; name=\"title\"\r\n\r\nHoliday\r\n--X--\r\n";
444/// let FormKind::Multipart(boundary) = FormKind::detect(content_type)? else { unreachable!() };
445/// let mut parser = MultipartParser::new(boundary, &Limits::DEFAULT, &Policy::DEFAULT)?;
446///
447/// let mut events = Vec::new();
448/// let mut chunks = body.chunks(7).peekable();     // any split gives the same events
449/// while let Some(chunk) = chunks.next() {
450///     let last = chunks.peek().is_none();
451///     let mut rest = chunk;
452///     loop {
453///         let step = parser.parse(rest, last)?;
454///         rest = &rest[step.consumed..];
455///         match step.event {
456///             Event::PartStart(head) => events.push(format!("start {}", head.name())),
457///             Event::Data(bytes) => events.push(format!("data {}", bytes.len())),
458///             Event::PartEnd(summary) => events.push(format!("end {}", summary.bytes)),
459///             Event::End => break,
460///             Event::NeedInput => break,
461///             _ => {}
462///         }
463///     }
464/// }
465/// assert!(parser.is_done());
466/// assert_eq!(events.first().map(String::as_str), Some("start title"));
467/// assert_eq!(events.last().map(String::as_str), Some("end 7"));
468/// # Ok::<(), form_net::FormError>(())
469/// ```
470pub struct MultipartParser<B: AsMut<[u8]> + AsRef<[u8]>> {
471    boundary: Boundary,
472    limits: Limits,
473    policy: Policy,
474    buffer: B,
475    state: State,
476    /// Delimiter bytes matched and not yet known to be data.
477    k: u8,
478    flags: u8,
479    token: Token,
480    token_len: u16,
481    charset_len: u8,
482    charset_done: bool,
483    scan: BlockScan,
484    /// Header bytes copied into the buffer.
485    copied: u32,
486    parts: u32,
487    files: u32,
488    fields: u32,
489    kind: PartKind,
490    part_bytes: u64,
491    text_total: u64,
492    data_start: u64,
493    preamble: u64,
494    epilogue: u64,
495    /// Where the current header block starts in the body.
496    head_start: u64,
497    /// Where the epilogue starts in the body.
498    epilogue_start: u64,
499    bytes_seen: u64,
500}
501
502impl<B: AsMut<[u8]> + AsRef<[u8]>> fmt::Debug for MultipartParser<B> {
503    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
504        f.debug_struct("MultipartParser")
505            .field("boundary", &self.boundary)
506            .field("state", &self.state)
507            .field("bytes_seen", &self.bytes_seen)
508            .field("parts_seen", &self.parts)
509            .finish_non_exhaustive()
510    }
511}
512
513#[cfg(feature = "alloc")]
514impl MultipartParser<alloc::vec::Vec<u8>> {
515    /// A parser with its own buffer of
516    /// [`Limits::multipart_buffer_len`] bytes, allocated once.
517    ///
518    /// # Errors
519    ///
520    /// `InvalidLimits` from validating the limits with the policy.
521    ///
522    /// # Examples
523    ///
524    /// ```
525    /// use form_net::{Boundary, Limits, MultipartParser, Policy};
526    ///
527    /// let parser = MultipartParser::new(Boundary::new(b"X")?, &Limits::DEFAULT, &Policy::DEFAULT)?;
528    /// assert_eq!(parser.bytes_seen(), 0);
529    /// # Ok::<(), form_net::FormError>(())
530    /// ```
531    pub fn new(boundary: Boundary, limits: &Limits, policy: &Policy) -> Result<Self, FormError> {
532        limits.validate_with(policy)?;
533        let buffer = alloc::vec![0u8; limits.multipart_buffer_len(policy)];
534        Self::with_buffer(boundary, limits, policy, buffer)
535    }
536}
537
538impl<B: AsMut<[u8]> + AsRef<[u8]>> MultipartParser<B> {
539    /// A parser over a buffer the caller provides; it never allocates.
540    ///
541    /// # Errors
542    ///
543    /// `InvalidLimits`; `BufferTooSmall` when `buffer` is shorter than
544    /// [`Limits::multipart_buffer_len`].
545    ///
546    /// # Examples
547    ///
548    /// ```
549    /// use form_net::{Boundary, Limits, MultipartParser, Policy};
550    ///
551    /// const LIMITS: Limits = Limits::DEFAULT.with_max_part_header_bytes(1024);
552    /// const LEN: usize = LIMITS.multipart_buffer_len(&Policy::DEFAULT);
553    /// let mut buffer = [0u8; LEN];
554    /// let parser = MultipartParser::with_buffer(Boundary::new(b"X")?, &LIMITS, &Policy::DEFAULT, &mut buffer[..])?;
555    /// # Ok::<(), form_net::FormError>(())
556    /// ```
557    pub fn with_buffer(
558        boundary: Boundary,
559        limits: &Limits,
560        policy: &Policy,
561        buffer: B,
562    ) -> Result<Self, FormError> {
563        limits.validate_with(policy)?;
564        let needed = limits.multipart_buffer_len(policy);
565        if buffer.as_ref().len() < needed {
566            return Err(FormError::BufferTooSmall { needed });
567        }
568        Ok(Self {
569            boundary,
570            limits: *limits,
571            policy: *policy,
572            buffer,
573            state: State::Preamble,
574            // The body may start with `--boundary`: start as if CRLF had
575            // been seen.
576            k: 2,
577            flags: 0,
578            token: Token::NotSeen,
579            token_len: 0,
580            charset_len: 0,
581            charset_done: false,
582            scan: BlockScan::default(),
583            copied: 0,
584            parts: 0,
585            files: 0,
586            fields: 0,
587            kind: PartKind::Field,
588            part_bytes: 0,
589            text_total: 0,
590            data_start: 0,
591            preamble: 0,
592            epilogue: 0,
593            head_start: 0,
594            epilogue_start: 0,
595            bytes_seen: 0,
596        })
597    }
598
599    /// Consumes a prefix of `input` and returns one event. `eof = true` says
600    /// `input` ends the body.
601    ///
602    /// # Errors
603    ///
604    /// Every multipart error (see the crate's `docs/API.md`); `Poisoned`
605    /// after any error until [`reset`](Self::reset).
606    pub fn parse<'p>(&'p mut self, input: &'p [u8], eof: bool) -> Result<Step<'p>, FormError> {
607        let (consumed, outcome) = self.advance(input, eof)?;
608        Ok(Step {
609            consumed,
610            event: self.view(outcome, input),
611        })
612    }
613
614    /// Discards the rest of the current part: its bytes are consumed and
615    /// counted against the budgets but produce no `Data` events, and its
616    /// summary has `skipped = true`. Outside a part it does nothing.
617    pub fn skip_part(&mut self) {
618        if matches!(self.state, State::Body) {
619            self.flags |= SKIPPING;
620        }
621    }
622
623    /// Returns the parser to its initial state for a new body, wipes its
624    /// buffer, and clears the poisoned state. It never allocates.
625    pub fn reset(&mut self, boundary: Boundary) {
626        crate::wipe::wipe(self.buffer.as_mut());
627        let limits = self.limits;
628        let policy = self.policy;
629        self.boundary = boundary;
630        self.state = State::Preamble;
631        self.k = 2;
632        self.flags = 0;
633        self.token = Token::NotSeen;
634        self.token_len = 0;
635        self.charset_len = 0;
636        self.charset_done = false;
637        self.scan = BlockScan::default();
638        self.copied = 0;
639        self.parts = 0;
640        self.files = 0;
641        self.fields = 0;
642        self.kind = PartKind::Field;
643        self.part_bytes = 0;
644        self.text_total = 0;
645        self.data_start = 0;
646        self.preamble = 0;
647        self.epilogue = 0;
648        self.head_start = 0;
649        self.epilogue_start = 0;
650        self.bytes_seen = 0;
651        self.limits = limits;
652        self.policy = policy;
653    }
654
655    /// Body bytes consumed so far.
656    #[must_use]
657    pub const fn bytes_seen(&self) -> u64 {
658        self.bytes_seen
659    }
660
661    /// Parts started so far.
662    #[must_use]
663    pub const fn parts_seen(&self) -> u32 {
664        self.parts
665    }
666
667    /// File parts counted so far.
668    #[must_use]
669    pub const fn files_seen(&self) -> u32 {
670        self.files
671    }
672
673    /// Whether the close delimiter was read.
674    #[must_use]
675    pub fn is_done(&self) -> bool {
676        self.state == State::Epilogue
677    }
678
679    /// The limits the parser was built with.
680    #[must_use]
681    pub const fn limits(&self) -> &Limits {
682        &self.limits
683    }
684
685    /// The policy the parser was built with.
686    #[must_use]
687    pub const fn policy(&self) -> &Policy {
688        &self.policy
689    }
690
691    /// The token field's value, once its part has ended.
692    #[must_use]
693    pub fn captured_token(&self) -> Option<&str> {
694        if self.token != Token::Captured {
695            return None;
696        }
697        let start = self.token_area();
698        let bytes = self
699            .buffer
700            .as_ref()
701            .get(start..start.saturating_add(usize::from(self.token_len)))?;
702        core::str::from_utf8(bytes).ok()
703    }
704
705    /// The value of the last complete `_charset_` field.
706    #[must_use]
707    pub fn declared_charset(&self) -> Option<&[u8]> {
708        if !self.charset_done {
709            return None;
710        }
711        let start = self.charset_area();
712        self.buffer
713            .as_ref()
714            .get(start..start.saturating_add(usize::from(self.charset_len)))
715    }
716
717    /// The decoded name of the last head, `len` bytes in the name area.
718    #[cfg(feature = "alloc")]
719    pub(crate) fn area_name(&self, len: u32) -> &str {
720        let start = self.name_area();
721        let bytes = self
722            .buffer
723            .as_ref()
724            .get(start..start.saturating_add(len as usize))
725            .unwrap_or(&[]);
726        core::str::from_utf8(bytes).unwrap_or("")
727    }
728
729    /// The decoded `filename*` of the last head, `len` bytes.
730    #[cfg(feature = "alloc")]
731    pub(crate) fn area_filename(&self, len: u32) -> &[u8] {
732        let start = self.filename_area();
733        self.buffer
734            .as_ref()
735            .get(start..start.saturating_add(len as usize))
736            .unwrap_or(&[])
737    }
738
739    /// Where the current part's data starts in the body.
740    #[cfg(feature = "alloc")]
741    pub(crate) const fn data_start(&self) -> u64 {
742        self.data_start
743    }
744
745    // Buffer layout: header block, name, filename (AcceptAlone), token, charset.
746
747    fn name_area(&self) -> usize {
748        self.limits.max_part_header_bytes as usize
749    }
750
751    fn filename_area(&self) -> usize {
752        self.name_area().saturating_add(self.limits.name() as usize)
753    }
754
755    fn token_area(&self) -> usize {
756        let filename = if matches!(
757            self.policy.extended_filename,
758            crate::ExtendedFilenamePolicy::AcceptAlone
759        ) {
760            self.limits.filename() as usize
761        } else {
762            0
763        };
764        self.filename_area().saturating_add(filename)
765    }
766
767    fn charset_area(&self) -> usize {
768        let token = if self.policy.token_field.is_some() {
769            usize::from(self.limits.token())
770        } else {
771            0
772        };
773        self.token_area().saturating_add(token)
774    }
775
776    /// The location of the current part's header block.
777    fn head_at(&self) -> Location {
778        Location::in_part(self.parts.checked_sub(1), self.head_start)
779    }
780
781    /// The location of the first data byte past a budget of `limit` of
782    /// which `used` is spent, in the current part.
783    fn excess(&self, used: u64, limit: u64) -> Location {
784        let into_part = self.part_bytes.saturating_add(limit.saturating_sub(used));
785        Location::in_part(
786            self.parts.checked_sub(1),
787            self.data_start.saturating_add(into_part),
788        )
789    }
790
791    fn at(&self, pos: usize) -> Location {
792        let part = if matches!(self.state, State::Preamble | State::Epilogue) {
793            None
794        } else {
795            self.parts.checked_sub(1)
796        };
797        Location::in_part(part, self.bytes_seen.saturating_add(pos as u64))
798    }
799
800    /// Runs the state machine and returns positions only, so callers that
801    /// own the input for longer (the buffered parser) can slice it
802    /// themselves.
803    pub(crate) fn advance(
804        &mut self,
805        input: &[u8],
806        eof: bool,
807    ) -> Result<(usize, Outcome), FormError> {
808        if self.state == State::Poisoned {
809            return Err(FormError::Poisoned);
810        }
811        match self.run(input, eof) {
812            Ok((consumed, outcome)) => {
813                self.bytes_seen = self.bytes_seen.saturating_add(consumed as u64);
814                Ok((consumed, outcome))
815            }
816            Err(e) => {
817                self.state = State::Poisoned;
818                Err(e)
819            }
820        }
821    }
822
823    fn run(&mut self, input: &[u8], eof: bool) -> Result<(usize, Outcome), FormError> {
824        if self.bytes_seen.saturating_add(input.len() as u64) > self.limits.body() {
825            return Err(FormError::limit(
826                Limit::Body,
827                Location::at(self.limits.body()),
828            ));
829        }
830        let mut pos = 0usize;
831        loop {
832            let hay = input.get(pos..).unwrap_or(&[]);
833            match self.state {
834                State::Poisoned => return Err(FormError::Poisoned),
835                State::Preamble => {
836                    if let Some(after) = self.preamble_step(hay)? {
837                        pos = pos.wrapping_add(after);
838                        self.state = State::AfterDelimiter(After::Start);
839                        continue;
840                    }
841                    if eof {
842                        return Err(FormError::syntax(
843                            SyntaxError::NoInitialDelimiter,
844                            self.at(input.len()),
845                        ));
846                    }
847                    return Ok((input.len(), Outcome::NeedInput));
848                }
849                State::AfterDelimiter(sub) => {
850                    let Some(&c) = hay.first() else {
851                        return self.need_input(input.len(), eof);
852                    };
853                    pos = pos.wrapping_add(1);
854                    let at = self.at(pos);
855                    self.state = match (sub, c) {
856                        (After::Start, b'-') => State::AfterDelimiter(After::Dash),
857                        (After::Dash, b'-') => {
858                            self.state = State::Epilogue;
859                            self.epilogue_start = self.bytes_seen.saturating_add(pos as u64);
860                            return Ok((pos, Outcome::End));
861                        }
862                        (After::Start, b' ' | b'\t') => {
863                            if self.limits.max_padding_bytes == 0 {
864                                return Err(FormError::limit(Limit::Padding, at));
865                            }
866                            State::AfterDelimiter(After::Padding(1))
867                        }
868                        (After::Padding(n), b' ' | b'\t') => {
869                            if n >= self.limits.max_padding_bytes {
870                                return Err(FormError::limit(Limit::Padding, at));
871                            }
872                            State::AfterDelimiter(After::Padding(n.saturating_add(1)))
873                        }
874                        (After::Start | After::Padding(_), b'\r') => {
875                            State::AfterDelimiter(After::Cr)
876                        }
877                        (After::Cr, b'\n') => {
878                            self.parts = self.parts.saturating_add(1);
879                            if self.parts > self.limits.max_parts {
880                                return Err(FormError::limit(Limit::Parts, self.at(pos)));
881                            }
882                            self.scan = BlockScan::default();
883                            self.copied = 0;
884                            self.head_start = self.bytes_seen.saturating_add(pos as u64);
885                            State::Headers
886                        }
887                        (After::Start | After::Padding(_), b'\n') => {
888                            return Err(FormError::syntax(SyntaxError::BareLineFeed, at));
889                        }
890                        _ => return Err(FormError::syntax(SyntaxError::BoundaryInContent, at)),
891                    };
892                }
893                State::Headers => {
894                    let at = self.head_at();
895                    match self.scan.feed(hay, &self.limits, at)? {
896                        Some(end) => {
897                            let source = if self.copied == 0 {
898                                Source::Input(pos)
899                            } else {
900                                self.copy_header(hay.get(..end).unwrap_or(&[]), at)?;
901                                Source::Buffer
902                            };
903                            let block_start_in_input = pos;
904                            pos = pos.wrapping_add(end);
905                            let facts =
906                                self.read_head(input, source, block_start_in_input, end, at)?;
907                            let (is_token, is_charset) = {
908                                let block = self.block(&facts, source, input);
909                                let name = self.name_bytes(&facts, block);
910                                let token = self
911                                    .policy
912                                    .token_field
913                                    .is_some_and(|t| name == t.name().as_bytes());
914                                (token, name == b"_charset_")
915                            };
916                            self.start_part(&facts, is_token, is_charset, at)?;
917                            self.data_start = self.bytes_seen.saturating_add(pos as u64);
918                            return Ok((pos, Outcome::PartStart(facts, source)));
919                        }
920                        None => {
921                            self.copy_header(hay, at)?;
922                            return self.need_input(input.len(), eof);
923                        }
924                    }
925                }
926                State::Body => {
927                    if hay.is_empty() {
928                        return self.need_input(input.len(), eof);
929                    }
930                    match self.body_step(hay, pos, input.len(), eof)? {
931                        BodyStep::Emit(consumed, outcome) => return Ok((consumed, outcome)),
932                        BodyStep::Continue(next) => pos = next,
933                    }
934                }
935                State::EndOfPart => {
936                    let summary = self.end_part(self.at(pos))?;
937                    self.state = State::AfterDelimiter(After::Start);
938                    return Ok((pos, Outcome::PartEnd(summary)));
939                }
940                State::Epilogue => {
941                    self.epilogue = self.epilogue.saturating_add(hay.len() as u64);
942                    if self.epilogue > self.limits.epilogue() {
943                        let at = self.epilogue_start.saturating_add(self.limits.epilogue());
944                        return Err(FormError::limit(Limit::Epilogue, Location::at(at)));
945                    }
946                    return Ok((input.len(), Outcome::End));
947                }
948            }
949        }
950    }
951
952    fn need_input(&self, len: usize, eof: bool) -> Result<(usize, Outcome), FormError> {
953        if eof {
954            Err(FormError::syntax(SyntaxError::Truncated, self.at(len)))
955        } else {
956            Ok((len, Outcome::NeedInput))
957        }
958    }
959
960    /// Scans the preamble. Returns the index in `hay` just after the first
961    /// delimiter, or `None` when all of `hay` was consumed.
962    fn preamble_step(&mut self, hay: &[u8]) -> Result<Option<usize>, FormError> {
963        let len = self.boundary.len() as u64;
964        // A delimiter must end within max_preamble_bytes + its own length.
965        let budget = self.limits.preamble().saturating_add(len);
966        let room = usize::try_from(budget.saturating_sub(self.preamble)).unwrap_or(usize::MAX);
967        let window = hay.get(..room.min(hay.len())).unwrap_or(hay);
968        let (k, found) = self.boundary.run(usize::from(self.k), window);
969        self.k = k as u8;
970        match found {
971            Some(after) => {
972                // Real bytes up to the end of the delimiter; a delimiter that
973                // includes the two virtual bytes (CRLF) starts the body.
974                let end = self.preamble.saturating_add(after as u64);
975                let preamble = end.saturating_sub(len);
976                if preamble > self.limits.preamble() {
977                    return Err(FormError::limit(
978                        Limit::Preamble,
979                        Location::at(self.limits.preamble()),
980                    ));
981                }
982                self.k = 0;
983                Ok(Some(after))
984            }
985            None => {
986                self.preamble = self.preamble.saturating_add(window.len() as u64);
987                if window.len() < hay.len() || self.preamble >= budget {
988                    return Err(FormError::limit(
989                        Limit::Preamble,
990                        Location::at(self.limits.preamble()),
991                    ));
992                }
993                Ok(None)
994            }
995        }
996    }
997
998    fn copy_header(&mut self, bytes: &[u8], at: Location) -> Result<(), FormError> {
999        let start = self.copied as usize;
1000        let end = start.saturating_add(bytes.len());
1001        if end > self.limits.max_part_header_bytes as usize {
1002            return Err(FormError::limit(Limit::PartHeaderBytes, at));
1003        }
1004        match self.buffer.as_mut().get_mut(start..end) {
1005            Some(dst) => dst.copy_from_slice(bytes),
1006            None => return Err(FormError::limit(Limit::PartHeaderBytes, at)),
1007        }
1008        self.copied = end as u32;
1009        Ok(())
1010    }
1011
1012    fn read_head(
1013        &mut self,
1014        input: &[u8],
1015        source: Source,
1016        start: usize,
1017        end: usize,
1018        at: Location,
1019    ) -> Result<HeadFacts, FormError> {
1020        let name_area = self.name_area();
1021        let filename_area = self.filename_area();
1022        let filename_end = self.token_area();
1023        let limits = self.limits;
1024        let policy = self.policy;
1025        let copied = self.copied as usize;
1026        let buf = self.buffer.as_mut();
1027        let (head, rest) = buf.split_at_mut(name_area.min(buf.len()));
1028        let (name, rest) =
1029            rest.split_at_mut(filename_area.saturating_sub(name_area).min(rest.len()));
1030        let filename = rest
1031            .get_mut(..filename_end.saturating_sub(filename_area))
1032            .unwrap_or(&mut []);
1033        let block = match source {
1034            Source::Input(_) => input.get(start..start.saturating_add(end)).unwrap_or(&[]),
1035            Source::Buffer => head.get(..copied).unwrap_or(&[]),
1036        };
1037        read_block(block, &limits, &policy, Areas { name, filename }, at)
1038    }
1039
1040    /// The budgets and rules that apply when a part starts.
1041    fn start_part(
1042        &mut self,
1043        facts: &HeadFacts,
1044        is_token: bool,
1045        is_charset: bool,
1046        at: Location,
1047    ) -> Result<(), FormError> {
1048        self.state = State::Body;
1049        self.k = 0;
1050        self.flags = 0;
1051        self.kind = facts.kind;
1052        self.part_bytes = 0;
1053        match facts.kind {
1054            PartKind::Field => {
1055                self.fields = self.fields.saturating_add(1);
1056                if self.fields > self.limits.max_fields {
1057                    return Err(FormError::limit(Limit::Fields, at));
1058                }
1059                if is_token {
1060                    if self.token != Token::NotSeen {
1061                        return Err(FormError::policy(PolicyRefusal::DuplicateToken, at));
1062                    }
1063                    self.token = Token::Capturing;
1064                    self.token_len = 0;
1065                    self.flags |= IS_TOKEN;
1066                }
1067                if is_charset {
1068                    self.flags |= IS_CHARSET;
1069                    self.charset_len = 0;
1070                    self.charset_done = false;
1071                }
1072            }
1073            PartKind::File => {
1074                let empty = matches!(facts.filename, Some((Text::Block(span), _)) if span.len == 0)
1075                    || matches!(facts.filename, Some((Text::Area(0), _)));
1076                if empty && self.policy.empty_files == EmptyFilePolicy::Skip {
1077                    self.flags |= DEFERRED;
1078                } else {
1079                    self.count_file(at)?;
1080                }
1081            }
1082        }
1083        Ok(())
1084    }
1085
1086    /// The block of a head, wherever it is.
1087    fn block<'a>(&'a self, facts: &HeadFacts, source: Source, input: &'a [u8]) -> &'a [u8] {
1088        let len = facts.block_len as usize;
1089        match source {
1090            Source::Input(start) => input.get(start..start.saturating_add(len)).unwrap_or(&[]),
1091            Source::Buffer => self.buffer.as_ref().get(..len).unwrap_or(&[]),
1092        }
1093    }
1094
1095    /// The decoded name of a head.
1096    fn name_bytes<'a>(&'a self, facts: &HeadFacts, block: &'a [u8]) -> &'a [u8] {
1097        match facts.name {
1098            Text::Block(span) => span.get(block),
1099            Text::Area(len) => {
1100                let start = self.name_area();
1101                self.buffer
1102                    .as_ref()
1103                    .get(start..start.saturating_add(len as usize))
1104                    .unwrap_or(&[])
1105            }
1106        }
1107    }
1108
1109    /// The borrowed view of a head.
1110    fn head_view<'p>(&'p self, facts: &HeadFacts, source: Source, input: &'p [u8]) -> PartHead<'p> {
1111        let block = self.block(facts, source, input);
1112        let name = core::str::from_utf8(self.name_bytes(facts, block)).unwrap_or("");
1113        let filename = facts.filename.map(|(text, origin)| {
1114            let raw = match text {
1115                Text::Block(span) => span.get(block),
1116                Text::Area(len) => {
1117                    let start = self.filename_area();
1118                    self.buffer
1119                        .as_ref()
1120                        .get(start..start.saturating_add(len as usize))
1121                        .unwrap_or(&[])
1122                }
1123            };
1124            FileName::new(raw, origin)
1125        });
1126        PartHead {
1127            index: self.parts.saturating_sub(1),
1128            block,
1129            name,
1130            name_raw: facts.name_raw.get(block),
1131            kind: facts.kind,
1132            filename,
1133            content_type: facts.content_type.map(|span| span.get(block)),
1134            charset: facts.charset,
1135        }
1136    }
1137
1138    /// Turns positions into borrowed events.
1139    fn view<'p>(&'p self, outcome: Outcome, input: &'p [u8]) -> Event<'p> {
1140        match outcome {
1141            Outcome::PartStart(facts, source) => {
1142                Event::PartStart(self.head_view(&facts, source, input))
1143            }
1144            Outcome::DataInput(start, end) => Event::Data(input.get(start..end).unwrap_or(&[])),
1145            Outcome::DataDelimiter(n) => {
1146                Event::Data(self.boundary.delimiter().get(..n).unwrap_or(&[]))
1147            }
1148            Outcome::PartEnd(summary) => Event::PartEnd(summary),
1149            Outcome::End => Event::End,
1150            Outcome::NeedInput => Event::NeedInput,
1151        }
1152    }
1153
1154    fn count_file(&mut self, at: Location) -> Result<(), FormError> {
1155        self.files = self.files.saturating_add(1);
1156        if self.files > self.limits.max_files {
1157            return Err(FormError::limit(Limit::Files, at));
1158        }
1159        if let Some(t) = self.policy.token_field {
1160            if t.placement() == TokenPlacement::BeforeFirstFile && self.token != Token::Captured {
1161                return Err(FormError::policy(PolicyRefusal::TokenFieldLate, at));
1162            }
1163        }
1164        Ok(())
1165    }
1166
1167    /// Checks budgets for `n` more data bytes of the current part, and
1168    /// captures the token or `_charset_` value.
1169    fn take_data(&mut self, bytes: &[u8]) -> Result<(), FormError> {
1170        let n = bytes.len() as u64;
1171        if n == 0 {
1172            return Ok(());
1173        }
1174        if self.flags & DEFERRED != 0 {
1175            self.flags &= !DEFERRED;
1176            let at = Location::in_part(self.parts.checked_sub(1), self.data_start);
1177            self.count_file(at)?;
1178        }
1179        let total = self.part_bytes.saturating_add(n);
1180        match self.kind {
1181            PartKind::File => {
1182                let limit = self.limits.file();
1183                if total > limit {
1184                    let at = self.excess(self.part_bytes, limit);
1185                    return Err(FormError::limit(Limit::FileSize, at));
1186                }
1187            }
1188            PartKind::Field => {
1189                let limit = u64::from(self.limits.value());
1190                if total > limit {
1191                    let at = self.excess(self.part_bytes, limit);
1192                    return Err(FormError::limit(Limit::ValueLength, at));
1193                }
1194                let text = self.text_total.saturating_add(n);
1195                let limit = self.limits.text();
1196                if text > limit {
1197                    let at = self.excess(self.text_total, limit);
1198                    return Err(FormError::limit(Limit::TextTotal, at));
1199                }
1200                self.text_total = text;
1201            }
1202        }
1203        if self.flags & IS_TOKEN != 0 {
1204            let start = self
1205                .token_area()
1206                .saturating_add(usize::from(self.token_len));
1207            let end = start.saturating_add(bytes.len());
1208            let limit = u64::from(self.limits.token());
1209            let at = self.excess(u64::from(self.token_len), limit);
1210            if end
1211                > self
1212                    .token_area()
1213                    .saturating_add(usize::from(self.limits.token()))
1214            {
1215                return Err(FormError::limit(Limit::TokenLength, at));
1216            }
1217            match self.buffer.as_mut().get_mut(start..end) {
1218                Some(dst) => dst.copy_from_slice(bytes),
1219                None => return Err(FormError::limit(Limit::TokenLength, at)),
1220            }
1221            self.token_len = self.token_len.saturating_add(bytes.len() as u16);
1222        }
1223        if self.flags & IS_CHARSET != 0 {
1224            let start = self
1225                .charset_area()
1226                .saturating_add(usize::from(self.charset_len));
1227            let end = start.saturating_add(bytes.len());
1228            let at = self.excess(u64::from(self.charset_len), CHARSET_CAPTURE as u64);
1229            if end > self.charset_area().saturating_add(CHARSET_CAPTURE) {
1230                return Err(FormError::encoding(EncodingError::UnsupportedCharset, at));
1231            }
1232            match self.buffer.as_mut().get_mut(start..end) {
1233                Some(dst) => dst.copy_from_slice(bytes),
1234                None => return Err(FormError::encoding(EncodingError::UnsupportedCharset, at)),
1235            }
1236            self.charset_len = self.charset_len.saturating_add(bytes.len() as u8);
1237        }
1238        self.part_bytes = total;
1239        Ok(())
1240    }
1241
1242    fn end_part(&mut self, at: Location) -> Result<PartSummary, FormError> {
1243        if self.flags & IS_TOKEN != 0 {
1244            let start = self.token_area();
1245            let bytes = self
1246                .buffer
1247                .as_ref()
1248                .get(start..start.saturating_add(usize::from(self.token_len)))
1249                .unwrap_or(&[]);
1250            if core::str::from_utf8(bytes).is_err() {
1251                return Err(FormError::encoding(EncodingError::InvalidUtf8, at));
1252            }
1253            self.token = Token::Captured;
1254        }
1255        if self.flags & IS_CHARSET != 0 {
1256            self.charset_done = true;
1257        }
1258        let summary = PartSummary {
1259            index: self.parts.saturating_sub(1),
1260            kind: self.kind,
1261            bytes: self.part_bytes,
1262            skipped: self.flags & SKIPPING != 0,
1263        };
1264        self.flags = 0;
1265        self.k = 0;
1266        Ok(summary)
1267    }
1268
1269    /// One step in a part's data. `pos` is where `hay` starts in the input.
1270    fn body_step(
1271        &mut self,
1272        hay: &[u8],
1273        pos: usize,
1274        input_len: usize,
1275        eof: bool,
1276    ) -> Result<BodyStep, FormError> {
1277        let len = self.boundary.len();
1278        let k0 = usize::from(self.k);
1279        if k0 > 0 {
1280            // Bytes from earlier chunks that matched the start of the
1281            // delimiter: resolve them before anything else.
1282            let mut k = k0;
1283            for (i, &c) in hay.iter().enumerate() {
1284                k = self.boundary.step(k, c);
1285                let n = i.wrapping_add(1);
1286                if k == len {
1287                    // The delimiter started inside the held bytes.
1288                    let data = k0.saturating_add(n).saturating_sub(len);
1289                    self.k = 0;
1290                    self.state = State::EndOfPart;
1291                    return self.emit_delimiter_bytes(data, pos.wrapping_add(n));
1292                }
1293                if k <= n {
1294                    // Every held byte is data; scan the input again from 0.
1295                    self.k = 0;
1296                    return self.emit_delimiter_bytes(k0, pos);
1297                }
1298            }
1299            // All of `hay` extended the partial match.
1300            let n = hay.len();
1301            if eof {
1302                return Err(FormError::syntax(
1303                    SyntaxError::Truncated,
1304                    self.at(input_len),
1305                ));
1306            }
1307            let data = k0.saturating_add(n).saturating_sub(k);
1308            self.k = k as u8;
1309            return self.emit_delimiter_bytes(data, input_len);
1310        }
1311        let (k, found) = self.boundary.run(0, hay);
1312        match found {
1313            Some(after) => {
1314                let data = after.saturating_sub(len);
1315                self.k = 0;
1316                self.state = State::EndOfPart;
1317                let bytes = hay.get(..data).unwrap_or(&[]);
1318                self.emit_input_bytes(bytes, pos, pos.wrapping_add(data), pos.wrapping_add(after))
1319            }
1320            None => {
1321                if eof {
1322                    return Err(FormError::syntax(
1323                        SyntaxError::Truncated,
1324                        self.at(input_len),
1325                    ));
1326                }
1327                self.k = k as u8;
1328                let data_end = hay.len().saturating_sub(k);
1329                let bytes = hay.get(..data_end).unwrap_or(&[]);
1330                self.emit_input_bytes(bytes, pos, pos.wrapping_add(data_end), input_len)
1331            }
1332        }
1333    }
1334
1335    /// Emits `delimiter[..n]` as data (or skips it), consuming up to
1336    /// `consumed`.
1337    fn emit_delimiter_bytes(&mut self, n: usize, consumed: usize) -> Result<BodyStep, FormError> {
1338        let boundary = self.boundary;
1339        let bytes = boundary.delimiter().get(..n).unwrap_or(&[]);
1340        self.take_data(bytes)?;
1341        if n == 0 || self.flags & SKIPPING != 0 {
1342            return Ok(BodyStep::Continue(consumed));
1343        }
1344        Ok(BodyStep::Emit(consumed, Outcome::DataDelimiter(n)))
1345    }
1346
1347    /// Emits `input[start..end]` as data (or skips it), consuming up to
1348    /// `consumed`.
1349    fn emit_input_bytes(
1350        &mut self,
1351        bytes: &[u8],
1352        start: usize,
1353        end: usize,
1354        consumed: usize,
1355    ) -> Result<BodyStep, FormError> {
1356        self.take_data(bytes)?;
1357        if start == end || self.flags & SKIPPING != 0 {
1358            return Ok(BodyStep::Continue(consumed));
1359        }
1360        Ok(BodyStep::Emit(consumed, Outcome::DataInput(start, end)))
1361    }
1362}
1363
1364/// What a body step did.
1365enum BodyStep {
1366    Emit(usize, Outcome),
1367    Continue(usize),
1368}
1369
1370impl<B: AsMut<[u8]> + AsRef<[u8]>> Drop for MultipartParser<B> {
1371    fn drop(&mut self) {
1372        crate::wipe::wipe(self.buffer.as_mut());
1373    }
1374}
1375
1376#[cfg(test)]
1377mod tests {
1378    use super::*;
1379
1380    #[test]
1381    fn test_parser_size_is_bounded() {
1382        // NFR-18 as measured: the boundary and its matcher table are 149 of
1383        // these bytes; the buffer is separate.
1384        assert!(core::mem::size_of::<MultipartParser<&mut [u8]>>() <= 384);
1385        assert!(core::mem::size_of::<Boundary>() <= 150);
1386    }
1387
1388    #[cfg(feature = "alloc")]
1389    #[test]
1390    fn test_debug_never_shows_data_or_file_names() {
1391        let body = b"--X\r\nContent-Disposition: form-data; name=f; filename=\"secret.pdf\"\r\n\r\nhunter2\r\n--X--";
1392        let mut p = MultipartParser::new(
1393            Boundary::new(b"X").unwrap(),
1394            &Limits::DEFAULT,
1395            &Policy::DEFAULT,
1396        )
1397        .unwrap();
1398        let mut rest: &[u8] = body;
1399        let mut seen = alloc::string::String::new();
1400        loop {
1401            let step = p.parse(rest, true).unwrap();
1402            rest = &rest[step.consumed..];
1403            if let Event::PartStart(head) = &step.event {
1404                seen.push_str(&alloc::format!("{:?}", head.headers()));
1405            }
1406            seen.push_str(&alloc::format!("{step:?}"));
1407            if matches!(step.event, Event::End) {
1408                break;
1409            }
1410        }
1411        assert!(seen.contains("Data(7 bytes)"), "{seen}");
1412        assert!(
1413            !seen.contains("hunter2") && !seen.contains("secret"),
1414            "{seen}"
1415        );
1416    }
1417
1418    #[test]
1419    fn test_charset_labels() {
1420        assert_eq!(Charset::from_label(b"UTF-8"), Charset::Utf8);
1421        assert_eq!(Charset::from_label(b"us-ascii"), Charset::Ascii);
1422        assert_eq!(Charset::from_label(b"utf8"), Charset::Other);
1423    }
1424}
form_net/multipart/parser.rs

form_net/multipart/
parser.rs
