Merge branch 'swscanf_works' into 'master'

swscanf and vswscanf implementation proposal

See merge request redox-os/relibc!472
This commit is contained in:
Jeremy Soller
2024-08-01 18:41:14 +00:00
7 changed files with 687 additions and 3 deletions
+53
View File
@@ -0,0 +1,53 @@
use super::{fseek_locked, ftell_locked, FILE, SEEK_SET};
use crate::{
io::Read,
platform::types::{off_t, wint_t},
};
struct LookAheadBuffer {
buf: *const wint_t,
pos: isize,
look_ahead: isize,
}
impl LookAheadBuffer {
fn look_ahead(&mut self) -> Result<Option<wint_t>, i32> {
let wchar = unsafe { *self.buf.offset(self.look_ahead) };
if wchar == 0 {
Ok(None)
} else {
self.look_ahead += 1;
Ok(Some(wchar))
}
}
fn commit(&mut self) {
self.pos = self.look_ahead;
}
}
impl From<*const wint_t> for LookAheadBuffer {
fn from(buff: *const wint_t) -> LookAheadBuffer {
LookAheadBuffer {
buf: buff,
pos: 0,
look_ahead: 0,
}
}
}
pub struct LookAheadReader(LookAheadBuffer);
impl LookAheadReader {
pub fn lookahead1(&mut self) -> Result<Option<wint_t>, i32> {
self.0.look_ahead()
}
pub fn commit(&mut self) {
self.0.commit()
}
}
impl From<*const wint_t> for LookAheadReader {
fn from(buff: *const wint_t) -> LookAheadReader {
LookAheadReader(buff.into())
}
}
+19 -3
View File
@@ -15,8 +15,10 @@ use crate::{
platform::{self, types::*, ERRNO},
};
mod lookaheadreader;
mod utf8;
mod wprintf;
mod wscanf;
#[repr(C)]
#[derive(Clone, Copy)]
@@ -277,9 +279,23 @@ pub unsafe extern "C" fn putwchar(wc: wchar_t) -> wint_t {
fputwc(wc, &mut *stdout)
}
// #[no_mangle]
pub extern "C" fn swscanf(s: *const wchar_t, format: *const wchar_t, ap: va_list) -> c_int {
unimplemented!();
#[no_mangle]
pub unsafe extern "C" fn vswscanf(
s: *const wchar_t,
format: *const wchar_t,
__valist: va_list,
) -> c_int {
let reader = (s as *const wint_t).into();
wscanf::scanf(reader, format, __valist)
}
#[no_mangle]
pub unsafe extern "C" fn swscanf(
s: *const wchar_t,
format: *const wchar_t,
mut __valist: ...
) -> c_int {
vswscanf(s, format, __valist.as_va_list())
}
/// Push wide character `wc` back onto `stream` so it'll be read next
+518
View File
@@ -0,0 +1,518 @@
use super::lookaheadreader::LookAheadReader;
use crate::platform::types::*;
use alloc::{string::String, vec::Vec};
use core::ffi::VaList as va_list;
#[derive(PartialEq, Eq)]
enum IntKind {
Byte,
Short,
Int,
Long,
LongLong,
IntMax,
PtrDiff,
Size,
}
#[derive(PartialEq, Eq)]
enum CharKind {
Ascii,
Wide,
}
/// Helper function for progressing a C string
unsafe fn next_char(string: &mut *const wchar_t) -> Result<wint_t, c_int> {
let c = **string as wint_t;
*string = string.offset(1);
if c == 0 {
Err(-1)
} else {
Ok(c)
}
}
macro_rules! wc_as_char {
($c:ident) => {
char::try_from($c).map_err(|_| -1)?
};
}
unsafe fn inner_scanf(
mut r: LookAheadReader,
mut format: *const wchar_t,
mut ap: va_list,
) -> Result<c_int, c_int> {
let mut matched = 0;
let mut wchar = 0;
let mut skip_read = false;
let mut count = 0;
macro_rules! read {
() => {{
match r.lookahead1() {
Ok(None) => false,
Ok(Some(b)) => {
wchar = b;
count += 1;
true
}
Err(x) => return Err(x),
}
}};
}
macro_rules! maybe_read {
() => {
maybe_read!(inner false);
};
(noreset) => {
maybe_read!(inner);
};
(inner $($placeholder:expr)*) => {
if !skip_read && !read!() {
match matched {
0 => return Ok(-1),
a => return Ok(a),
}
}
$(else {
// Hacky way of having this optional
skip_read = $placeholder;
})*
}
}
while *format != 0 {
let mut c = next_char(&mut format)?;
if c as u8 == b' ' {
maybe_read!(noreset);
while (wc_as_char!(wchar)).is_whitespace() {
if !read!() {
return Ok(matched);
}
}
skip_read = true;
} else if c as u8 != b'%' {
maybe_read!();
if c != wchar {
return Ok(matched);
}
r.commit();
} else {
c = next_char(&mut format)?;
let mut ignore = false;
if c as u8 == b'*' {
ignore = true;
c = next_char(&mut format)?;
}
let mut width = String::new();
while c as u8 >= b'0' && c as u8 <= b'9' {
width.push(wc_as_char!(c));
c = next_char(&mut format)?;
}
let mut width = if width.is_empty() {
None
} else {
match width.parse::<usize>() {
Ok(n) => Some(n),
Err(_) => return Err(-1),
}
};
// When an EOF occurs, eof is set, stuff is marked matched
// as usual, and finally it is returned
let mut eof = false;
let mut kind = IntKind::Int;
let mut c_kind = CharKind::Ascii;
loop {
match c as u8 {
b'h' => {
if kind == IntKind::Short || kind == IntKind::Byte {
kind = IntKind::Byte;
} else {
kind = IntKind::Short;
}
}
b'j' => kind = IntKind::IntMax,
b'l' => {
if kind == IntKind::Long || kind == IntKind::LongLong {
kind = IntKind::LongLong;
} else {
kind = IntKind::Long;
}
}
b'q' | b'L' => kind = IntKind::LongLong,
b't' => kind = IntKind::PtrDiff,
b'z' => kind = IntKind::Size,
// If kind is Long, means we found a 'l' before finding 'c' or 's'. In this
// case the format corresponds to a wide char/string
b'c' | b's' if kind == IntKind::Long => {
c_kind = CharKind::Wide;
break;
}
_ => break,
}
c = next_char(&mut format)?;
}
if c as u8 != b'n' {
maybe_read!(noreset);
}
match c as u8 {
b'%' => {
while (wc_as_char!(wchar)).is_whitespace() {
if !read!() {
return Ok(matched);
}
}
if wchar as u8 != b'%' {
return Err(matched);
} else if !read!() {
return Ok(matched);
}
}
b'd' | b'i' | b'o' | b'u' | b'x' | b'X' | b'f' | b'e' | b'g' | b'E' | b'a'
| b'p' => {
while (wc_as_char!(wchar)).is_whitespace() {
if !read!() {
return Ok(matched);
}
}
let pointer = c as u8 == b'p';
// Pointers aren't automatic, but we do want to parse "0x"
let auto = c as u8 == b'i' || pointer;
let float = c as u8 == b'f'
|| c as u8 == b'e'
|| c as u8 == b'g'
|| c as u8 == b'E'
|| c as u8 == b'a';
let mut radix = match c as u8 {
b'o' => 8,
b'x' | b'X' | b'p' => 16,
_ => 10,
};
let mut n = String::new();
let mut dot = false;
while width.map(|w| w > 0).unwrap_or(true)
&& ((wchar as u8 >= b'0' && wchar as u8 <= b'7')
|| (radix >= 10 && (wchar as u8 >= b'8' && wchar as u8 <= b'9'))
|| (float && !dot && wchar as u8 == b'.')
|| (radix == 16
&& ((wchar as u8 >= b'a' && wchar as u8 <= b'f')
|| (wchar as u8 >= b'A' && wchar as u8 <= b'F'))))
{
if auto
&& n.is_empty()
&& wchar as u8 == b'0'
&& width.map(|w| w > 0).unwrap_or(true)
{
if !pointer {
radix = 8;
}
width = width.map(|w| w - 1);
if !read!() {
return Ok(matched);
}
if width.map(|w| w > 0).unwrap_or(true)
&& (wchar as u8 == b'x' || wchar as u8 == b'X')
{
radix = 16;
width = width.map(|w| w - 1);
if width.map(|w| w > 0).unwrap_or(true) && !read!() {
return Ok(matched);
}
}
continue;
}
if wchar as u8 == b'.' {
// Don't allow another dot
dot = true;
}
n.push(wc_as_char!(wchar));
r.commit();
width = width.map(|w| w - 1);
if width.map(|w| w > 0).unwrap_or(true) && !read!() {
break;
}
}
macro_rules! parse_type {
(noformat $type:ident) => {{
let n = if n.is_empty() {
0 as $type
} else {
n.parse::<$type>().map_err(|_| 0)?
};
if !ignore {
*ap.arg::<*mut $type>() = n;
matched += 1;
}
}};
(c_double) => {
parse_type!(noformat c_double)
};
(c_float) => {
parse_type!(noformat c_float)
};
($type:ident) => {
parse_type!($type, $type)
};
($type:ident, $final:ty) => {{
let n = if n.is_empty() {
0 as $type
} else {
$type::from_str_radix(&n, radix).map_err(|_| 0)?
};
if !ignore {
*ap.arg::<*mut $final>() = n as $final;
matched += 1;
}
}};
}
if float {
if kind == IntKind::Long || kind == IntKind::LongLong {
parse_type!(c_double);
} else {
parse_type!(c_float);
}
} else if c as u8 == b'p' {
parse_type!(size_t, *mut c_void);
} else {
let unsigned = c as u8 == b'o'
|| c as u8 == b'u'
|| c as u8 == b'x'
|| c as u8 == b'X';
match kind {
IntKind::Byte => {
if unsigned {
parse_type!(c_uchar);
} else {
parse_type!(c_char);
}
}
IntKind::Short => {
if unsigned {
parse_type!(c_ushort)
} else {
parse_type!(c_short)
}
}
IntKind::Int => {
if unsigned {
parse_type!(c_uint)
} else {
parse_type!(c_int)
}
}
IntKind::Long => {
if unsigned {
parse_type!(c_ulong)
} else {
parse_type!(c_long)
}
}
IntKind::LongLong => {
if unsigned {
parse_type!(c_ulonglong)
} else {
parse_type!(c_longlong)
}
}
IntKind::IntMax => {
if unsigned {
parse_type!(uintmax_t)
} else {
parse_type!(intmax_t)
}
}
IntKind::PtrDiff => parse_type!(ptrdiff_t),
IntKind::Size => {
if unsigned {
parse_type!(size_t)
} else {
parse_type!(ssize_t)
}
}
}
}
}
b's' => {
macro_rules! parse_string_type {
($type:ident) => {
while (wc_as_char!(wchar)).is_whitespace() {
if !read!() {
return Ok(matched);
}
}
let mut ptr: Option<*mut $type> =
if ignore { None } else { Some(ap.arg()) };
while width.map(|w| w > 0).unwrap_or(true)
&& !(wc_as_char!(wchar)).is_whitespace()
{
if let Some(ref mut ptr) = ptr {
**ptr = wchar as $type;
*ptr = ptr.offset(1);
}
width = width.map(|w| w - 1);
if width.map(|w| w > 0).unwrap_or(true) && !read!() {
eof = true;
break;
}
}
if let Some(ptr) = ptr {
*ptr = 0;
matched += 1;
r.commit();
}
};
}
if c_kind == CharKind::Ascii {
parse_string_type!(c_char);
} else {
parse_string_type!(wchar_t);
}
}
b'c' => {
macro_rules! parse_char_type {
($type:ident) => {
let ptr: Option<*mut $type> =
if ignore { None } else { Some(ap.arg()) };
for i in 0..width.unwrap_or(1) {
if let Some(ptr) = ptr {
*ptr.add(i) = wchar as $type;
}
width = width.map(|w| w - 1);
if width.map(|w| w > 0).unwrap_or(true) && !read!() {
eof = true;
break;
}
}
if ptr.is_some() {
matched += 1;
r.commit();
}
};
}
if c_kind == CharKind::Ascii {
parse_char_type!(c_char);
} else {
parse_char_type!(wchar_t);
}
}
b'[' => {
c = next_char(&mut format)?;
let mut matches = Vec::new();
let invert = if c as u8 == b'^' {
c = next_char(&mut format)?;
true
} else {
false
};
let mut prev;
loop {
matches.push(c);
prev = c;
c = next_char(&mut format)?;
if c as u8 == b'-' {
if prev as u8 == b']' {
continue;
}
c = next_char(&mut format)?;
if c as u8 == b']' {
matches.push('-' as wint_t);
break;
}
prev += 1;
while prev < c {
matches.push(prev);
prev += 1;
}
} else if c as u8 == b']' {
break;
}
}
let mut ptr: Option<*mut c_char> = if ignore { None } else { Some(ap.arg()) };
// While we haven't used up all the width, and it matches
let mut data_stored = false;
while width.map(|w| w > 0).unwrap_or(true)
&& !invert == matches.contains(&wchar)
{
if let Some(ref mut ptr) = ptr {
**ptr = wchar as c_char;
*ptr = ptr.offset(1);
data_stored = true;
}
r.commit();
// Decrease the width, and read a new character unless the width is 0
width = width.map(|w| w - 1);
if width.map(|w| w > 0).unwrap_or(true) && !read!() {
// Reading a new character has failed, return after
// actually marking this as matched
eof = true;
break;
}
}
if data_stored {
*ptr.unwrap() = 0;
matched += 1;
}
}
b'n' => {
if !ignore {
*ap.arg::<*mut c_int>() = count as c_int;
}
}
_ => return Err(-1),
}
if eof {
return Ok(matched);
}
if width != Some(0) && c as u8 != b'n' {
// It didn't hit the width, so an extra character was read and matched.
// But this character did not match so let's reuse it.
skip_read = true;
}
}
}
Ok(matched)
}
pub unsafe fn scanf(r: LookAheadReader, format: *const wchar_t, ap: va_list) -> c_int {
match inner_scanf(r, format, ap) {
Ok(n) => n,
Err(n) => n,
}
}
+1
View File
@@ -120,6 +120,7 @@ EXPECT_NAMES=\
wchar/mbsrtowcs \
wchar/printf-on-wchars \
wchar/putwchar \
wchar/wscanf \
wchar/ungetwc \
wchar/wprintf \
wchar/wcrtomb \
Binary file not shown.
+96
View File
@@ -0,0 +1,96 @@
/* swscanf example */
#include <wchar.h>
#include <stdio.h>
#include "test_helpers.h"
struct params {
short sa;
int ia;
int ib;
int ic;
float fa;
double da;
int *ptr;
char c;
wchar_t wc;
char string1[20];
char string2[20];
char string3[20];
char string4[20];
wchar_t wstring1[20];
wchar_t wstring2[20];
wchar_t wstring3[20];
wchar_t wstring4[20];
};
void test(wchar_t* fmt_in, wchar_t* input, struct params *p, ...) {
va_list args;
va_start(args, p);
wint_t ret = vswscanf(input, fmt_in, args);
va_end(args);
wprintf(
L"%d, { sa: %hhd, ia: %d, ib: %d, ic: %d, fa: %f, da: %lf, ptr: %p, char: %c, wide char: %lc, string1: %s, string2: %s, string3: %s, string4: %s, wstring1: %ls, wstring2: %ls, wstring3: %ls, wstring4: %ls }\n",
ret, p->sa, p->ia, p->ib, p->ic, p->fa, p->da, p->ptr, p->c, p->wc, p->string1, p->string2, p->string3, p->string4, p->wstring1, p->wstring2, p->wstring3, p->wstring4
);
}
int main ()
{
struct params p = { .c = 'a' };
test(L"%hd %d", L"12 345", &p, &p.sa, &p.ia);
test(L"%x %i %i", L"12 0x345 010", &p, &p.ia, &p.ib, &p.ic);
test(L"%f.%lf", L"0.1.0.2", &p, &p.fa, &p.da);
test(L"%p", L"0xABCDEF", &p, &p.ptr);
test(L"%s", L"Hello World", &p, &p.string1);
test(L"%3i", L"0xFF", &p, &p.ia);
test(L"%c%3c", L"hello", &p, &p.c, &p.string1);
test(L"%lc", L"β", &p, &p.wc);
test(L"%lc %f", L"π 3.14", &p, &p.wc, &p.fa);
test(L"test: %2i%n", L"test: 0xFF", &p, &p.ia, &p.ib);
test(L"hello world%%", L"hello world%", &p);
test(L"h%1[ae]ll%1[^a] wor%1[^\n]%[d]", L"hello world", &p, &p.string1, &p.string2, &p.string3, &p.string4);
test(L"h%1[ae]ll%1[^a] wor%1[^\n]%[d]", L"halle worfdddddd", &p, &p.string1, &p.string2, &p.string3, &p.string4);
test(L"%[^a]%[b]", L"testbbbb", &p, &p.string1, &p.string2);
test(L"%ls %ls", L"Привет мир", &p, &p.wstring1, &p.wstring2);
test(L"%ls %ls", L"こんにちは 世界", &p, &p.wstring1, &p.wstring2);
test(L"%ls %d %ls %d", L"αβγ 123 δεζ 456", &p, &p.wstring1, &p.ia, &p.wstring2, &p.ib);
test(L"%ls %s %ls %s", L"αβγ test1 δεζ test2", &p, &p.wstring1, &p.string1, &p.wstring2, &p.string2);
test(L"%ls %ls %ls %ls", L"z ß 水 🍌", &p, &p.wstring1, &p.wstring2, &p.wstring3, &p.wstring4);
// Scanf stolen from the url parsing in curl
wchar_t protobuf[16];
wchar_t slashbuf[4];
wchar_t hostbuf[100];
wchar_t pathbuf[100];
// don't push NUL, make sure scanf does that
memset(protobuf, 97, 16);
memset(slashbuf, 97, 4);
memset(hostbuf, 97, 100);
memset(pathbuf, 97, 100);
int ret = swscanf(
L"https://redox-os.org\0# extra garbage for nul test", L"%15[^\n/:]:%3[/]%[^\n/?#]%[^\n]",
&protobuf, &slashbuf, &hostbuf, &pathbuf
);
if (ret < 4) {
*pathbuf = 0;
}
if (ret < 3) {
*hostbuf = 0;
}
if (ret < 2) {
*slashbuf = 0;
}
if (ret < 1) {
*protobuf = 0;
}
wprintf(L"%d \"%s\" \"%s\" \"%s\" \"%s\"\n", ret, &protobuf, &slashbuf, &hostbuf, &pathbuf);
return 0;
}