| 1 |
#include <Python.h> |
| 2 |
#include "pycore_bytesobject.h" // _PyBytes_DecodeEscape() |
| 3 |
#include "pycore_unicodeobject.h" // _PyUnicode_DecodeUnicodeEscapeInternal() |
| 4 |
|
| 5 |
#include "lexer/state.h" |
| 6 |
#include "pegen.h" |
| 7 |
#include "string_parser.h" |
| 8 |
|
| 9 |
#include <stdbool.h> |
| 10 |
|
| 11 |
//// STRING HANDLING FUNCTIONS //// |
| 12 |
|
| 13 |
static int |
| 14 |
warn_invalid_escape_sequence(Parser *p, const char* buffer, const char *first_invalid_escape, Token *t) |
| 15 |
{ |
| 16 |
if (p->call_invalid_rules) { |
| 17 |
// Do not report warnings if we are in the second pass of the parser |
| 18 |
// to avoid showing the warning twice. |
| 19 |
return 0; |
| 20 |
} |
| 21 |
unsigned char c = (unsigned char)*first_invalid_escape; |
| 22 |
if ((t->type == FSTRING_MIDDLE || t->type == FSTRING_END) && (c == '{' || c == '}')) { |
| 23 |
// in this case the tokenizer has already emitted a warning, |
| 24 |
// see Parser/tokenizer/helpers.c:warn_invalid_escape_sequence |
| 25 |
return 0; |
| 26 |
} |
| 27 |
|
| 28 |
int octal = ('4' <= c && c <= '7'); |
| 29 |
PyObject *msg = |
| 30 |
octal |
| 31 |
? PyUnicode_FromFormat("invalid octal escape sequence '\\%.3s'", |
| 32 |
first_invalid_escape) |
| 33 |
: PyUnicode_FromFormat("invalid escape sequence '\\%c'", c); |
| 34 |
if (msg == NULL) { |
| 35 |
return -1; |
| 36 |
} |
| 37 |
PyObject *category; |
| 38 |
if (p->feature_version >= 12) { |
| 39 |
category = PyExc_SyntaxWarning; |
| 40 |
} |
| 41 |
else { |
| 42 |
category = PyExc_DeprecationWarning; |
| 43 |
} |
| 44 |
|
| 45 |
// Calculate the lineno and the col_offset of the invalid escape sequence |
| 46 |
const char *start = buffer; |
| 47 |
const char *end = first_invalid_escape; |
| 48 |
int lineno = t->lineno; |
| 49 |
int col_offset = t->col_offset; |
| 50 |
while (start < end) { |
| 51 |
if (*start == '\n') { |
| 52 |
lineno++; |
| 53 |
col_offset = 0; |
| 54 |
} |
| 55 |
else { |
| 56 |
col_offset++; |
| 57 |
} |
| 58 |
start++; |
| 59 |
} |
| 60 |
|
| 61 |
// Count the number of quotes in the token |
| 62 |
char first_quote = 0; |
| 63 |
if (lineno == t->lineno) { |
| 64 |
int quote_count = 0; |
| 65 |
char* tok = PyBytes_AsString(t->bytes); |
| 66 |
for (int i = 0; i < PyBytes_Size(t->bytes); i++) { |
| 67 |
if (tok[i] == '\'' || tok[i] == '\"') { |
| 68 |
if (quote_count == 0) { |
| 69 |
first_quote = tok[i]; |
| 70 |
} |
| 71 |
if (tok[i] == first_quote) { |
| 72 |
quote_count++; |
| 73 |
} |
| 74 |
} else { |
| 75 |
break; |
| 76 |
} |
| 77 |
} |
| 78 |
|
| 79 |
col_offset += quote_count; |
| 80 |
} |
| 81 |
|
| 82 |
if (PyErr_WarnExplicitObject(category, msg, p->tok->filename, |
| 83 |
lineno, NULL, NULL) < 0) { |
| 84 |
if (PyErr_ExceptionMatches(category)) { |
| 85 |
/* Replace the Syntax/DeprecationWarning exception with a SyntaxError |
| 86 |
to get a more accurate error report */ |
| 87 |
PyErr_Clear(); |
| 88 |
|
| 89 |
/* This is needed, in order for the SyntaxError to point to the token t, |
| 90 |
since _PyPegen_raise_error uses p->tokens[p->fill - 1] for the |
| 91 |
error location, if p->known_err_token is not set. */ |
| 92 |
p->known_err_token = t; |
| 93 |
if (octal) { |
| 94 |
RAISE_ERROR_KNOWN_LOCATION(p, PyExc_SyntaxError, lineno, col_offset-1, lineno, col_offset+1, |
| 95 |
"invalid octal escape sequence '\\%.3s'", first_invalid_escape); |
| 96 |
} |
| 97 |
else { |
| 98 |
RAISE_ERROR_KNOWN_LOCATION(p, PyExc_SyntaxError, lineno, col_offset-1, lineno, col_offset+1, |
| 99 |
"invalid escape sequence '\\%c'", c); |
| 100 |
} |
| 101 |
} |
| 102 |
Py_DECREF(msg); |
| 103 |
return -1; |
| 104 |
} |
| 105 |
Py_DECREF(msg); |
| 106 |
return 0; |
| 107 |
} |
| 108 |
|
| 109 |
static PyObject * |
| 110 |
decode_utf8(const char **sPtr, const char *end) |
| 111 |
{ |
| 112 |
const char *s; |
| 113 |
const char *t; |
| 114 |
t = s = *sPtr; |
| 115 |
while (s < end && (*s & 0x80)) { |
| 116 |
s++; |
| 117 |
} |
| 118 |
*sPtr = s; |
| 119 |
return PyUnicode_DecodeUTF8(t, s - t, NULL); |
| 120 |
} |
| 121 |
|
| 122 |
static PyObject * |
| 123 |
decode_unicode_with_escapes(Parser *parser, const char *s, size_t len, Token *t) |
| 124 |
{ |
| 125 |
PyObject *v; |
| 126 |
PyObject *u; |
| 127 |
char *buf; |
| 128 |
char *p; |
| 129 |
const char *end; |
| 130 |
|
| 131 |
/* check for integer overflow */ |
| 132 |
if (len > (size_t)PY_SSIZE_T_MAX / 6) { |
| 133 |
return NULL; |
| 134 |
} |
| 135 |
/* "ä" (2 bytes) may become "\U000000E4" (10 bytes), or 1:5 |
| 136 |
"\ä" (3 bytes) may become "\u005c\U000000E4" (16 bytes), or ~1:6 */ |
| 137 |
u = PyBytes_FromStringAndSize((char *)NULL, (Py_ssize_t)len * 6); |
| 138 |
if (u == NULL) { |
| 139 |
return NULL; |
| 140 |
} |
| 141 |
p = buf = PyBytes_AsString(u); |
| 142 |
if (p == NULL) { |
| 143 |
return NULL; |
| 144 |
} |
| 145 |
end = s + len; |
| 146 |
while (s < end) { |
| 147 |
if (*s == '\\') { |
| 148 |
*p++ = *s++; |
| 149 |
if (s >= end || *s & 0x80) { |
| 150 |
strcpy(p, "u005c"); |
| 151 |
p += 5; |
| 152 |
if (s >= end) { |
| 153 |
break; |
| 154 |
} |
| 155 |
} |
| 156 |
} |
| 157 |
if (*s & 0x80) { |
| 158 |
PyObject *w; |
| 159 |
int kind; |
| 160 |
const void *data; |
| 161 |
Py_ssize_t w_len; |
| 162 |
Py_ssize_t i; |
| 163 |
w = decode_utf8(&s, end); |
| 164 |
if (w == NULL) { |
| 165 |
Py_DECREF(u); |
| 166 |
return NULL; |
| 167 |
} |
| 168 |
kind = PyUnicode_KIND(w); |
| 169 |
data = PyUnicode_DATA(w); |
| 170 |
w_len = PyUnicode_GET_LENGTH(w); |
| 171 |
for (i = 0; i < w_len; i++) { |
| 172 |
Py_UCS4 chr = PyUnicode_READ(kind, data, i); |
| 173 |
sprintf(p, "\\U%08x", chr); |
| 174 |
p += 10; |
| 175 |
} |
| 176 |
/* Should be impossible to overflow */ |
| 177 |
assert(p - buf <= PyBytes_GET_SIZE(u)); |
| 178 |
Py_DECREF(w); |
| 179 |
} |
| 180 |
else { |
| 181 |
*p++ = *s++; |
| 182 |
} |
| 183 |
} |
| 184 |
len = (size_t)(p - buf); |
| 185 |
s = buf; |
| 186 |
|
| 187 |
int first_invalid_escape_char; |
| 188 |
const char *first_invalid_escape_ptr; |
| 189 |
v = _PyUnicode_DecodeUnicodeEscapeInternal2(s, (Py_ssize_t)len, NULL, NULL, |
| 190 |
&first_invalid_escape_char, |
| 191 |
&first_invalid_escape_ptr); |
| 192 |
|
| 193 |
// HACK: later we can simply pass the line no, since we don't preserve the tokens |
| 194 |
// when we are decoding the string but we preserve the line numbers. |
| 195 |
if (v != NULL && first_invalid_escape_ptr != NULL && t != NULL) { |
| 196 |
if (warn_invalid_escape_sequence(parser, s, first_invalid_escape_ptr, t) < 0) { |
| 197 |
/* We have not decref u before because first_invalid_escape_ptr |
| 198 |
points inside u. */ |
| 199 |
Py_XDECREF(u); |
| 200 |
Py_DECREF(v); |
| 201 |
return NULL; |
| 202 |
} |
| 203 |
} |
| 204 |
Py_XDECREF(u); |
| 205 |
return v; |
| 206 |
} |
| 207 |
|
| 208 |
static PyObject * |
| 209 |
decode_bytes_with_escapes(Parser *p, const char *s, Py_ssize_t len, Token *t) |
| 210 |
{ |
| 211 |
int first_invalid_escape_char; |
| 212 |
const char *first_invalid_escape_ptr; |
| 213 |
PyObject *result = _PyBytes_DecodeEscape2(s, len, NULL, |
| 214 |
&first_invalid_escape_char, |
| 215 |
&first_invalid_escape_ptr); |
| 216 |
if (result == NULL) { |
| 217 |
return NULL; |
| 218 |
} |
| 219 |
|
| 220 |
if (first_invalid_escape_ptr != NULL) { |
| 221 |
if (warn_invalid_escape_sequence(p, s, first_invalid_escape_ptr, t) < 0) { |
| 222 |
Py_DECREF(result); |
| 223 |
return NULL; |
| 224 |
} |
| 225 |
} |
| 226 |
return result; |
| 227 |
} |
| 228 |
|
| 229 |
PyObject * |
| 230 |
_PyPegen_decode_string(Parser *p, int raw, const char *s, size_t len, Token *t) |
| 231 |
{ |
| 232 |
if (raw) { |
| 233 |
return PyUnicode_DecodeUTF8Stateful(s, (Py_ssize_t)len, NULL, NULL); |
| 234 |
} |
| 235 |
return decode_unicode_with_escapes(p, s, len, t); |
| 236 |
} |
| 237 |
|
| 238 |
/* s must include the bracketing quote characters, and r, b &/or f prefixes |
| 239 |
(if any), and embedded escape sequences (if any). (f-strings are handled by the parser) |
| 240 |
_PyPegen_parse_string parses it, and returns the decoded Python string object. */ |
| 241 |
PyObject * |
| 242 |
_PyPegen_parse_string(Parser *p, Token *t) |
| 243 |
{ |
| 244 |
const char *s = PyBytes_AsString(t->bytes); |
| 245 |
if (s == NULL) { |
| 246 |
return NULL; |
| 247 |
} |
| 248 |
|
| 249 |
size_t len; |
| 250 |
int quote = Py_CHARMASK(*s); |
| 251 |
int bytesmode = 0; |
| 252 |
int rawmode = 0; |
| 253 |
|
| 254 |
if (Py_ISALPHA(quote)) { |
| 255 |
while (!bytesmode || !rawmode) { |
| 256 |
if (quote == 'b' || quote == 'B') { |
| 257 |
quote =(unsigned char)*++s; |
| 258 |
bytesmode = 1; |
| 259 |
} |
| 260 |
else if (quote == 'u' || quote == 'U') { |
| 261 |
quote = (unsigned char)*++s; |
| 262 |
} |
| 263 |
else if (quote == 'r' || quote == 'R') { |
| 264 |
quote = (unsigned char)*++s; |
| 265 |
rawmode = 1; |
| 266 |
} |
| 267 |
else { |
| 268 |
break; |
| 269 |
} |
| 270 |
} |
| 271 |
} |
| 272 |
|
| 273 |
if (quote != '\'' && quote != '\"') { |
| 274 |
PyErr_BadInternalCall(); |
| 275 |
return NULL; |
| 276 |
} |
| 277 |
|
| 278 |
/* Skip the leading quote char. */ |
| 279 |
s++; |
| 280 |
len = strlen(s); |
| 281 |
// gh-120155: 's' contains at least the trailing quote, |
| 282 |
// so the code '--len' below is safe. |
| 283 |
assert(len >= 1); |
| 284 |
|
| 285 |
if (len > INT_MAX) { |
| 286 |
PyErr_SetString(PyExc_OverflowError, "string to parse is too long"); |
| 287 |
return NULL; |
| 288 |
} |
| 289 |
if (s[--len] != quote) { |
| 290 |
/* Last quote char must match the first. */ |
| 291 |
PyErr_BadInternalCall(); |
| 292 |
return NULL; |
| 293 |
} |
| 294 |
if (len >= 4 && s[0] == quote && s[1] == quote) { |
| 295 |
/* A triple quoted string. We've already skipped one quote at |
| 296 |
the start and one at the end of the string. Now skip the |
| 297 |
two at the start. */ |
| 298 |
s += 2; |
| 299 |
len -= 2; |
| 300 |
/* And check that the last two match. */ |
| 301 |
if (s[--len] != quote || s[--len] != quote) { |
| 302 |
PyErr_BadInternalCall(); |
| 303 |
return NULL; |
| 304 |
} |
| 305 |
} |
| 306 |
|
| 307 |
/* Avoid invoking escape decoding routines if possible. */ |
| 308 |
rawmode = rawmode || strchr(s, '\\') == NULL; |
| 309 |
if (bytesmode) { |
| 310 |
/* Disallow non-ASCII characters. */ |
| 311 |
const char *ch; |
| 312 |
for (ch = s; *ch; ch++) { |
| 313 |
if (Py_CHARMASK(*ch) >= 0x80) { |
| 314 |
RAISE_SYNTAX_ERROR_KNOWN_LOCATION( |
| 315 |
t, |
| 316 |
"bytes can only contain ASCII " |
| 317 |
"literal characters"); |
| 318 |
return NULL; |
| 319 |
} |
| 320 |
} |
| 321 |
if (rawmode) { |
| 322 |
return PyBytes_FromStringAndSize(s, (Py_ssize_t)len); |
| 323 |
} |
| 324 |
return decode_bytes_with_escapes(p, s, (Py_ssize_t)len, t); |
| 325 |
} |
| 326 |
return _PyPegen_decode_string(p, rawmode, s, len, t); |
| 327 |
} |