mirror of https://github.com/python/cpython
Improve readability and style in parser files (GH-20884)
This commit is contained in:
parent
794e7d1ab2
commit
fb61c42361
|
@ -67,10 +67,11 @@ _PyPegen_check_barry_as_flufl(Parser *p) {
|
|||
assert(t->type == NOTEQUAL);
|
||||
|
||||
char* tok_str = PyBytes_AS_STRING(t->bytes);
|
||||
if (p->flags & PyPARSE_BARRY_AS_BDFL && strcmp(tok_str, "<>")){
|
||||
if (p->flags & PyPARSE_BARRY_AS_BDFL && strcmp(tok_str, "<>") != 0) {
|
||||
RAISE_SYNTAX_ERROR("with Barry as BDFL, use '<>' instead of '!='");
|
||||
return -1;
|
||||
} else if (!(p->flags & PyPARSE_BARRY_AS_BDFL)) {
|
||||
}
|
||||
if (!(p->flags & PyPARSE_BARRY_AS_BDFL)) {
|
||||
return strcmp(tok_str, "!=");
|
||||
}
|
||||
return 0;
|
||||
|
@ -245,7 +246,10 @@ raise_decode_error(Parser *p)
|
|||
errtype = "value error";
|
||||
}
|
||||
if (errtype) {
|
||||
PyObject *type, *value, *tback, *errstr;
|
||||
PyObject *type;
|
||||
PyObject *value;
|
||||
PyObject *tback;
|
||||
PyObject *errstr;
|
||||
PyErr_Fetch(&type, &value, &tback);
|
||||
errstr = PyObject_Str(value);
|
||||
if (errstr) {
|
||||
|
@ -274,7 +278,9 @@ raise_tokenizer_init_error(PyObject *filename)
|
|||
}
|
||||
PyObject *errstr = NULL;
|
||||
PyObject *tuple = NULL;
|
||||
PyObject *type, *value, *tback;
|
||||
PyObject *type;
|
||||
PyObject *value;
|
||||
PyObject *tback;
|
||||
PyErr_Fetch(&type, &value, &tback);
|
||||
errstr = PyObject_Str(value);
|
||||
if (!errstr) {
|
||||
|
@ -548,7 +554,8 @@ growable_comment_array_deallocate(growable_comment_array *arr) {
|
|||
int
|
||||
_PyPegen_fill_token(Parser *p)
|
||||
{
|
||||
const char *start, *end;
|
||||
const char *start;
|
||||
const char *end;
|
||||
int type = PyTokenizer_Get(p->tok, &start, &end);
|
||||
|
||||
// Record and skip '# type: ignore' comments
|
||||
|
@ -589,9 +596,8 @@ _PyPegen_fill_token(Parser *p)
|
|||
PyErr_NoMemory();
|
||||
return -1;
|
||||
}
|
||||
else {
|
||||
p->tokens = new_tokens;
|
||||
}
|
||||
p->tokens = new_tokens;
|
||||
|
||||
for (int i = p->size; i < newsize; i++) {
|
||||
p->tokens[i] = PyMem_Malloc(sizeof(Token));
|
||||
if (p->tokens[i] == NULL) {
|
||||
|
@ -615,7 +621,8 @@ _PyPegen_fill_token(Parser *p)
|
|||
int lineno = type == STRING ? p->tok->first_lineno : p->tok->lineno;
|
||||
const char *line_start = type == STRING ? p->tok->multi_line_start : p->tok->line_start;
|
||||
int end_lineno = p->tok->lineno;
|
||||
int col_offset = -1, end_col_offset = -1;
|
||||
int col_offset = -1;
|
||||
int end_col_offset = -1;
|
||||
if (start != NULL && start >= line_start) {
|
||||
col_offset = (int)(start - line_start);
|
||||
}
|
||||
|
@ -634,9 +641,8 @@ _PyPegen_fill_token(Parser *p)
|
|||
if (p->tok->done == E_DECODE) {
|
||||
return raise_decode_error(p);
|
||||
}
|
||||
else {
|
||||
return tokenizer_error(p);
|
||||
}
|
||||
return tokenizer_error(p);
|
||||
|
||||
}
|
||||
|
||||
return 0;
|
||||
|
@ -847,33 +853,36 @@ parsenumber_raw(const char *s)
|
|||
return PyLong_FromString(s, (char **)0, 0);
|
||||
}
|
||||
}
|
||||
else
|
||||
else {
|
||||
x = PyOS_strtol(s, (char **)&end, 0);
|
||||
}
|
||||
if (*end == '\0') {
|
||||
if (errno != 0)
|
||||
if (errno != 0) {
|
||||
return PyLong_FromString(s, (char **)0, 0);
|
||||
}
|
||||
return PyLong_FromLong(x);
|
||||
}
|
||||
/* XXX Huge floats may silently fail */
|
||||
if (imflag) {
|
||||
compl.real = 0.;
|
||||
compl.imag = PyOS_string_to_double(s, (char **)&end, NULL);
|
||||
if (compl.imag == -1.0 && PyErr_Occurred())
|
||||
if (compl.imag == -1.0 && PyErr_Occurred()) {
|
||||
return NULL;
|
||||
}
|
||||
return PyComplex_FromCComplex(compl);
|
||||
}
|
||||
else {
|
||||
dx = PyOS_string_to_double(s, NULL, NULL);
|
||||
if (dx == -1.0 && PyErr_Occurred())
|
||||
return NULL;
|
||||
return PyFloat_FromDouble(dx);
|
||||
dx = PyOS_string_to_double(s, NULL, NULL);
|
||||
if (dx == -1.0 && PyErr_Occurred()) {
|
||||
return NULL;
|
||||
}
|
||||
return PyFloat_FromDouble(dx);
|
||||
}
|
||||
|
||||
static PyObject *
|
||||
parsenumber(const char *s)
|
||||
{
|
||||
char *dup, *end;
|
||||
char *dup;
|
||||
char *end;
|
||||
PyObject *res = NULL;
|
||||
|
||||
assert(s != NULL);
|
||||
|
|
|
@ -42,7 +42,8 @@ warn_invalid_escape_sequence(Parser *p, unsigned char first_invalid_escape_char,
|
|||
static PyObject *
|
||||
decode_utf8(const char **sPtr, const char *end)
|
||||
{
|
||||
const char *s, *t;
|
||||
const char *s;
|
||||
const char *t;
|
||||
t = s = *sPtr;
|
||||
while (s < end && (*s & 0x80)) {
|
||||
s++;
|
||||
|
@ -54,7 +55,8 @@ decode_utf8(const char **sPtr, const char *end)
|
|||
static PyObject *
|
||||
decode_unicode_with_escapes(Parser *parser, const char *s, size_t len, Token *t)
|
||||
{
|
||||
PyObject *v, *u;
|
||||
PyObject *v;
|
||||
PyObject *u;
|
||||
char *buf;
|
||||
char *p;
|
||||
const char *end;
|
||||
|
@ -86,7 +88,8 @@ decode_unicode_with_escapes(Parser *parser, const char *s, size_t len, Token *t)
|
|||
PyObject *w;
|
||||
int kind;
|
||||
void *data;
|
||||
Py_ssize_t len, i;
|
||||
Py_ssize_t w_len;
|
||||
Py_ssize_t i;
|
||||
w = decode_utf8(&s, end);
|
||||
if (w == NULL) {
|
||||
Py_DECREF(u);
|
||||
|
@ -94,8 +97,8 @@ decode_unicode_with_escapes(Parser *parser, const char *s, size_t len, Token *t)
|
|||
}
|
||||
kind = PyUnicode_KIND(w);
|
||||
data = PyUnicode_DATA(w);
|
||||
len = PyUnicode_GET_LENGTH(w);
|
||||
for (i = 0; i < len; i++) {
|
||||
w_len = PyUnicode_GET_LENGTH(w);
|
||||
for (i = 0; i < w_len; i++) {
|
||||
Py_UCS4 chr = PyUnicode_READ(kind, data, i);
|
||||
sprintf(p, "\\U%08x", chr);
|
||||
p += 10;
|
||||
|
@ -169,18 +172,18 @@ _PyPegen_parsestr(Parser *p, int *bytesmode, int *rawmode, PyObject **result,
|
|||
if (Py_ISALPHA(quote)) {
|
||||
while (!*bytesmode || !*rawmode) {
|
||||
if (quote == 'b' || quote == 'B') {
|
||||
quote = *++s;
|
||||
quote =(unsigned char)*++s;
|
||||
*bytesmode = 1;
|
||||
}
|
||||
else if (quote == 'u' || quote == 'U') {
|
||||
quote = *++s;
|
||||
quote = (unsigned char)*++s;
|
||||
}
|
||||
else if (quote == 'r' || quote == 'R') {
|
||||
quote = *++s;
|
||||
quote = (unsigned char)*++s;
|
||||
*rawmode = 1;
|
||||
}
|
||||
else if (quote == 'f' || quote == 'F') {
|
||||
quote = *++s;
|
||||
quote = (unsigned char)*++s;
|
||||
fmode = 1;
|
||||
}
|
||||
else {
|
||||
|
@ -370,112 +373,112 @@ static void fstring_shift_arguments(expr_ty parent, arguments_ty args, int linen
|
|||
fstring_shift_seq_locations(parent, args->defaults, lineno, col_offset);
|
||||
}
|
||||
|
||||
static void fstring_shift_children_locations(expr_ty n, int lineno, int col_offset) {
|
||||
switch (n->kind) {
|
||||
static void fstring_shift_children_locations(expr_ty node, int lineno, int col_offset) {
|
||||
switch (node->kind) {
|
||||
case BoolOp_kind:
|
||||
fstring_shift_seq_locations(n, n->v.BoolOp.values, lineno, col_offset);
|
||||
fstring_shift_seq_locations(node, node->v.BoolOp.values, lineno, col_offset);
|
||||
break;
|
||||
case NamedExpr_kind:
|
||||
shift_expr(n, n->v.NamedExpr.target, lineno, col_offset);
|
||||
shift_expr(n, n->v.NamedExpr.value, lineno, col_offset);
|
||||
shift_expr(node, node->v.NamedExpr.target, lineno, col_offset);
|
||||
shift_expr(node, node->v.NamedExpr.value, lineno, col_offset);
|
||||
break;
|
||||
case BinOp_kind:
|
||||
shift_expr(n, n->v.BinOp.left, lineno, col_offset);
|
||||
shift_expr(n, n->v.BinOp.right, lineno, col_offset);
|
||||
shift_expr(node, node->v.BinOp.left, lineno, col_offset);
|
||||
shift_expr(node, node->v.BinOp.right, lineno, col_offset);
|
||||
break;
|
||||
case UnaryOp_kind:
|
||||
shift_expr(n, n->v.UnaryOp.operand, lineno, col_offset);
|
||||
shift_expr(node, node->v.UnaryOp.operand, lineno, col_offset);
|
||||
break;
|
||||
case Lambda_kind:
|
||||
fstring_shift_arguments(n, n->v.Lambda.args, lineno, col_offset);
|
||||
shift_expr(n, n->v.Lambda.body, lineno, col_offset);
|
||||
fstring_shift_arguments(node, node->v.Lambda.args, lineno, col_offset);
|
||||
shift_expr(node, node->v.Lambda.body, lineno, col_offset);
|
||||
break;
|
||||
case IfExp_kind:
|
||||
shift_expr(n, n->v.IfExp.test, lineno, col_offset);
|
||||
shift_expr(n, n->v.IfExp.body, lineno, col_offset);
|
||||
shift_expr(n, n->v.IfExp.orelse, lineno, col_offset);
|
||||
shift_expr(node, node->v.IfExp.test, lineno, col_offset);
|
||||
shift_expr(node, node->v.IfExp.body, lineno, col_offset);
|
||||
shift_expr(node, node->v.IfExp.orelse, lineno, col_offset);
|
||||
break;
|
||||
case Dict_kind:
|
||||
fstring_shift_seq_locations(n, n->v.Dict.keys, lineno, col_offset);
|
||||
fstring_shift_seq_locations(n, n->v.Dict.values, lineno, col_offset);
|
||||
fstring_shift_seq_locations(node, node->v.Dict.keys, lineno, col_offset);
|
||||
fstring_shift_seq_locations(node, node->v.Dict.values, lineno, col_offset);
|
||||
break;
|
||||
case Set_kind:
|
||||
fstring_shift_seq_locations(n, n->v.Set.elts, lineno, col_offset);
|
||||
fstring_shift_seq_locations(node, node->v.Set.elts, lineno, col_offset);
|
||||
break;
|
||||
case ListComp_kind:
|
||||
shift_expr(n, n->v.ListComp.elt, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(n->v.ListComp.generators); i < l; i++) {
|
||||
comprehension_ty comp = asdl_seq_GET(n->v.ListComp.generators, i);
|
||||
fstring_shift_comprehension(n, comp, lineno, col_offset);
|
||||
shift_expr(node, node->v.ListComp.elt, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(node->v.ListComp.generators); i < l; i++) {
|
||||
comprehension_ty comp = asdl_seq_GET(node->v.ListComp.generators, i);
|
||||
fstring_shift_comprehension(node, comp, lineno, col_offset);
|
||||
}
|
||||
break;
|
||||
case SetComp_kind:
|
||||
shift_expr(n, n->v.SetComp.elt, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(n->v.SetComp.generators); i < l; i++) {
|
||||
comprehension_ty comp = asdl_seq_GET(n->v.SetComp.generators, i);
|
||||
fstring_shift_comprehension(n, comp, lineno, col_offset);
|
||||
shift_expr(node, node->v.SetComp.elt, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(node->v.SetComp.generators); i < l; i++) {
|
||||
comprehension_ty comp = asdl_seq_GET(node->v.SetComp.generators, i);
|
||||
fstring_shift_comprehension(node, comp, lineno, col_offset);
|
||||
}
|
||||
break;
|
||||
case DictComp_kind:
|
||||
shift_expr(n, n->v.DictComp.key, lineno, col_offset);
|
||||
shift_expr(n, n->v.DictComp.value, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(n->v.DictComp.generators); i < l; i++) {
|
||||
comprehension_ty comp = asdl_seq_GET(n->v.DictComp.generators, i);
|
||||
fstring_shift_comprehension(n, comp, lineno, col_offset);
|
||||
shift_expr(node, node->v.DictComp.key, lineno, col_offset);
|
||||
shift_expr(node, node->v.DictComp.value, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(node->v.DictComp.generators); i < l; i++) {
|
||||
comprehension_ty comp = asdl_seq_GET(node->v.DictComp.generators, i);
|
||||
fstring_shift_comprehension(node, comp, lineno, col_offset);
|
||||
}
|
||||
break;
|
||||
case GeneratorExp_kind:
|
||||
shift_expr(n, n->v.GeneratorExp.elt, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(n->v.GeneratorExp.generators); i < l; i++) {
|
||||
comprehension_ty comp = asdl_seq_GET(n->v.GeneratorExp.generators, i);
|
||||
fstring_shift_comprehension(n, comp, lineno, col_offset);
|
||||
shift_expr(node, node->v.GeneratorExp.elt, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(node->v.GeneratorExp.generators); i < l; i++) {
|
||||
comprehension_ty comp = asdl_seq_GET(node->v.GeneratorExp.generators, i);
|
||||
fstring_shift_comprehension(node, comp, lineno, col_offset);
|
||||
}
|
||||
break;
|
||||
case Await_kind:
|
||||
shift_expr(n, n->v.Await.value, lineno, col_offset);
|
||||
shift_expr(node, node->v.Await.value, lineno, col_offset);
|
||||
break;
|
||||
case Yield_kind:
|
||||
shift_expr(n, n->v.Yield.value, lineno, col_offset);
|
||||
shift_expr(node, node->v.Yield.value, lineno, col_offset);
|
||||
break;
|
||||
case YieldFrom_kind:
|
||||
shift_expr(n, n->v.YieldFrom.value, lineno, col_offset);
|
||||
shift_expr(node, node->v.YieldFrom.value, lineno, col_offset);
|
||||
break;
|
||||
case Compare_kind:
|
||||
shift_expr(n, n->v.Compare.left, lineno, col_offset);
|
||||
fstring_shift_seq_locations(n, n->v.Compare.comparators, lineno, col_offset);
|
||||
shift_expr(node, node->v.Compare.left, lineno, col_offset);
|
||||
fstring_shift_seq_locations(node, node->v.Compare.comparators, lineno, col_offset);
|
||||
break;
|
||||
case Call_kind:
|
||||
shift_expr(n, n->v.Call.func, lineno, col_offset);
|
||||
fstring_shift_seq_locations(n, n->v.Call.args, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(n->v.Call.keywords); i < l; i++) {
|
||||
keyword_ty keyword = asdl_seq_GET(n->v.Call.keywords, i);
|
||||
shift_expr(n, keyword->value, lineno, col_offset);
|
||||
shift_expr(node, node->v.Call.func, lineno, col_offset);
|
||||
fstring_shift_seq_locations(node, node->v.Call.args, lineno, col_offset);
|
||||
for (Py_ssize_t i = 0, l = asdl_seq_LEN(node->v.Call.keywords); i < l; i++) {
|
||||
keyword_ty keyword = asdl_seq_GET(node->v.Call.keywords, i);
|
||||
shift_expr(node, keyword->value, lineno, col_offset);
|
||||
}
|
||||
break;
|
||||
case Attribute_kind:
|
||||
shift_expr(n, n->v.Attribute.value, lineno, col_offset);
|
||||
shift_expr(node, node->v.Attribute.value, lineno, col_offset);
|
||||
break;
|
||||
case Subscript_kind:
|
||||
shift_expr(n, n->v.Subscript.value, lineno, col_offset);
|
||||
fstring_shift_slice_locations(n, n->v.Subscript.slice, lineno, col_offset);
|
||||
shift_expr(n, n->v.Subscript.slice, lineno, col_offset);
|
||||
shift_expr(node, node->v.Subscript.value, lineno, col_offset);
|
||||
fstring_shift_slice_locations(node, node->v.Subscript.slice, lineno, col_offset);
|
||||
shift_expr(node, node->v.Subscript.slice, lineno, col_offset);
|
||||
break;
|
||||
case Starred_kind:
|
||||
shift_expr(n, n->v.Starred.value, lineno, col_offset);
|
||||
shift_expr(node, node->v.Starred.value, lineno, col_offset);
|
||||
break;
|
||||
case List_kind:
|
||||
fstring_shift_seq_locations(n, n->v.List.elts, lineno, col_offset);
|
||||
fstring_shift_seq_locations(node, node->v.List.elts, lineno, col_offset);
|
||||
break;
|
||||
case Tuple_kind:
|
||||
fstring_shift_seq_locations(n, n->v.Tuple.elts, lineno, col_offset);
|
||||
fstring_shift_seq_locations(node, node->v.Tuple.elts, lineno, col_offset);
|
||||
break;
|
||||
case JoinedStr_kind:
|
||||
fstring_shift_seq_locations(n, n->v.JoinedStr.values, lineno, col_offset);
|
||||
fstring_shift_seq_locations(node, node->v.JoinedStr.values, lineno, col_offset);
|
||||
break;
|
||||
case FormattedValue_kind:
|
||||
shift_expr(n, n->v.FormattedValue.value, lineno, col_offset);
|
||||
if (n->v.FormattedValue.format_spec) {
|
||||
shift_expr(n, n->v.FormattedValue.format_spec, lineno, col_offset);
|
||||
shift_expr(node, node->v.FormattedValue.value, lineno, col_offset);
|
||||
if (node->v.FormattedValue.format_spec) {
|
||||
shift_expr(node, node->v.FormattedValue.format_spec, lineno, col_offset);
|
||||
}
|
||||
break;
|
||||
default:
|
||||
|
@ -710,15 +713,17 @@ fstring_find_literal(Parser *p, const char **str, const char *end, int raw,
|
|||
assert(s == end || *s == '{' || *s == '}');
|
||||
done:
|
||||
if (literal_start != s) {
|
||||
if (raw)
|
||||
if (raw) {
|
||||
*literal = PyUnicode_DecodeUTF8Stateful(literal_start,
|
||||
s - literal_start,
|
||||
NULL, NULL);
|
||||
else
|
||||
} else {
|
||||
*literal = decode_unicode_with_escapes(p, literal_start,
|
||||
s - literal_start, t);
|
||||
if (!*literal)
|
||||
}
|
||||
if (!*literal) {
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
return result;
|
||||
}
|
||||
|
@ -790,10 +795,11 @@ fstring_find_expr(Parser *p, const char **str, const char *end, int raw, int rec
|
|||
/* Loop invariants. */
|
||||
assert(nested_depth >= 0);
|
||||
assert(*str >= expr_start && *str < end);
|
||||
if (quote_char)
|
||||
if (quote_char) {
|
||||
assert(string_type == 1 || string_type == 3);
|
||||
else
|
||||
} else {
|
||||
assert(string_type == 0);
|
||||
}
|
||||
|
||||
ch = **str;
|
||||
/* Nowhere inside an expression is a backslash allowed. */
|
||||
|
@ -890,7 +896,7 @@ fstring_find_expr(Parser *p, const char **str, const char *end, int raw, int rec
|
|||
goto error;
|
||||
}
|
||||
nested_depth--;
|
||||
int opening = parenstack[nested_depth];
|
||||
int opening = (unsigned char)parenstack[nested_depth];
|
||||
if (!((opening == '(' && ch == ')') ||
|
||||
(opening == '[' && ch == ']') ||
|
||||
(opening == '{' && ch == '}')))
|
||||
|
@ -915,20 +921,22 @@ fstring_find_expr(Parser *p, const char **str, const char *end, int raw, int rec
|
|||
goto error;
|
||||
}
|
||||
if (nested_depth) {
|
||||
int opening = parenstack[nested_depth - 1];
|
||||
int opening = (unsigned char)parenstack[nested_depth - 1];
|
||||
RAISE_SYNTAX_ERROR("f-string: unmatched '%c'", opening);
|
||||
goto error;
|
||||
}
|
||||
|
||||
if (*str >= end)
|
||||
if (*str >= end) {
|
||||
goto unexpected_end_of_string;
|
||||
}
|
||||
|
||||
/* Compile the expression as soon as possible, so we show errors
|
||||
related to the expression before errors related to the
|
||||
conversion or format_spec. */
|
||||
simple_expression = fstring_compile_expr(p, expr_start, expr_end, t);
|
||||
if (!simple_expression)
|
||||
if (!simple_expression) {
|
||||
goto error;
|
||||
}
|
||||
|
||||
/* Check for =, which puts the text value of the expression in
|
||||
expr_text. */
|
||||
|
@ -957,10 +965,11 @@ fstring_find_expr(Parser *p, const char **str, const char *end, int raw, int rec
|
|||
/* Check for a conversion char, if present. */
|
||||
if (**str == '!') {
|
||||
*str += 1;
|
||||
if (*str >= end)
|
||||
if (*str >= end) {
|
||||
goto unexpected_end_of_string;
|
||||
}
|
||||
|
||||
conversion = **str;
|
||||
conversion = (unsigned char)**str;
|
||||
*str += 1;
|
||||
|
||||
/* Validate the conversion. */
|
||||
|
@ -974,22 +983,26 @@ fstring_find_expr(Parser *p, const char **str, const char *end, int raw, int rec
|
|||
}
|
||||
|
||||
/* Check for the format spec, if present. */
|
||||
if (*str >= end)
|
||||
if (*str >= end) {
|
||||
goto unexpected_end_of_string;
|
||||
}
|
||||
if (**str == ':') {
|
||||
*str += 1;
|
||||
if (*str >= end)
|
||||
if (*str >= end) {
|
||||
goto unexpected_end_of_string;
|
||||
}
|
||||
|
||||
/* Parse the format spec. */
|
||||
format_spec = fstring_parse(p, str, end, raw, recurse_lvl+1,
|
||||
first_token, t, last_token);
|
||||
if (!format_spec)
|
||||
if (!format_spec) {
|
||||
goto error;
|
||||
}
|
||||
}
|
||||
|
||||
if (*str >= end || **str != '}')
|
||||
if (*str >= end || **str != '}') {
|
||||
goto unexpected_end_of_string;
|
||||
}
|
||||
|
||||
/* We're at a right brace. Consume it. */
|
||||
assert(*str < end);
|
||||
|
@ -1009,8 +1022,9 @@ fstring_find_expr(Parser *p, const char **str, const char *end, int raw, int rec
|
|||
format_spec, first_token->lineno,
|
||||
first_token->col_offset, last_token->end_lineno,
|
||||
last_token->end_col_offset, p->arena);
|
||||
if (!*expression)
|
||||
if (!*expression) {
|
||||
goto error;
|
||||
}
|
||||
|
||||
return 0;
|
||||
|
||||
|
@ -1059,28 +1073,32 @@ fstring_find_literal_and_expr(Parser *p, const char **str, const char *end, int
|
|||
|
||||
/* Get any literal string. */
|
||||
result = fstring_find_literal(p, str, end, raw, literal, recurse_lvl, t);
|
||||
if (result < 0)
|
||||
if (result < 0) {
|
||||
goto error;
|
||||
}
|
||||
|
||||
assert(result == 0 || result == 1);
|
||||
|
||||
if (result == 1)
|
||||
if (result == 1) {
|
||||
/* We have a literal, but don't look at the expression. */
|
||||
return 1;
|
||||
}
|
||||
|
||||
if (*str >= end || **str == '}')
|
||||
if (*str >= end || **str == '}') {
|
||||
/* We're at the end of the string or the end of a nested
|
||||
f-string: no expression. The top-level error case where we
|
||||
expect to be at the end of the string but we're at a '}' is
|
||||
handled later. */
|
||||
return 0;
|
||||
}
|
||||
|
||||
/* We must now be the start of an expression, on a '{'. */
|
||||
assert(**str == '{');
|
||||
|
||||
if (fstring_find_expr(p, str, end, raw, recurse_lvl, expr_text,
|
||||
expression, first_token, t, last_token) < 0)
|
||||
expression, first_token, t, last_token) < 0) {
|
||||
goto error;
|
||||
}
|
||||
|
||||
return 0;
|
||||
|
||||
|
@ -1099,8 +1117,9 @@ ExprList_check_invariants(ExprList *l)
|
|||
hasn't been deallocated. */
|
||||
assert(l->size >= 0);
|
||||
assert(l->p != NULL);
|
||||
if (l->size <= EXPRLIST_N_CACHED)
|
||||
if (l->size <= EXPRLIST_N_CACHED) {
|
||||
assert(l->data == l->p);
|
||||
}
|
||||
}
|
||||
#endif
|
||||
|
||||
|
@ -1130,11 +1149,13 @@ ExprList_Append(ExprList *l, expr_ty exp)
|
|||
/* We're still using the cached data. Switch to
|
||||
alloc-ing. */
|
||||
l->p = PyMem_RawMalloc(sizeof(expr_ty) * new_size);
|
||||
if (!l->p)
|
||||
if (!l->p) {
|
||||
return -1;
|
||||
}
|
||||
/* Copy the cached data into the new buffer. */
|
||||
for (i = 0; i < l->size; i++)
|
||||
for (i = 0; i < l->size; i++) {
|
||||
l->p[i] = l->data[i];
|
||||
}
|
||||
} else {
|
||||
/* Just realloc. */
|
||||
expr_ty *tmp = PyMem_RawRealloc(l->p, sizeof(expr_ty) * new_size);
|
||||
|
@ -1184,8 +1205,9 @@ ExprList_Finish(ExprList *l, PyArena *arena)
|
|||
seq = _Py_asdl_seq_new(l->size, arena);
|
||||
if (seq) {
|
||||
Py_ssize_t i;
|
||||
for (i = 0; i < l->size; i++)
|
||||
for (i = 0; i < l->size; i++) {
|
||||
asdl_seq_SET(seq, i, l->p[i]);
|
||||
}
|
||||
}
|
||||
ExprList_Dealloc(l);
|
||||
return seq;
|
||||
|
@ -1197,8 +1219,9 @@ ExprList_Finish(ExprList *l, PyArena *arena)
|
|||
static void
|
||||
FstringParser_check_invariants(FstringParser *state)
|
||||
{
|
||||
if (state->last_str)
|
||||
if (state->last_str) {
|
||||
assert(PyUnicode_CheckExact(state->last_str));
|
||||
}
|
||||
ExprList_check_invariants(&state->expr_list);
|
||||
}
|
||||
#endif
|
||||
|
@ -1268,8 +1291,9 @@ _PyPegen_FstringParser_ConcatAndDel(FstringParser *state, PyObject *str)
|
|||
} else {
|
||||
/* Concatenate this with the previous string. */
|
||||
PyUnicode_AppendAndDel(&state->last_str, str);
|
||||
if (!state->last_str)
|
||||
if (!state->last_str) {
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
FstringParser_check_invariants(state);
|
||||
return 0;
|
||||
|
@ -1298,8 +1322,9 @@ _PyPegen_FstringParser_ConcatFstring(Parser *p, FstringParser *state, const char
|
|||
int result = fstring_find_literal_and_expr(p, str, end, raw, recurse_lvl,
|
||||
&literal, &expr_text,
|
||||
&expression, first_token, t, last_token);
|
||||
if (result < 0)
|
||||
if (result < 0) {
|
||||
return -1;
|
||||
}
|
||||
|
||||
/* Add the literal, if any. */
|
||||
if (literal && _PyPegen_FstringParser_ConcatAndDel(state, literal) < 0) {
|
||||
|
@ -1318,12 +1343,14 @@ _PyPegen_FstringParser_ConcatFstring(Parser *p, FstringParser *state, const char
|
|||
and expression, while ignoring the expression this
|
||||
time. This is used for un-doubling braces, as an
|
||||
optimization. */
|
||||
if (result == 1)
|
||||
if (result == 1) {
|
||||
continue;
|
||||
}
|
||||
|
||||
if (!expression)
|
||||
if (!expression) {
|
||||
/* We're done with this f-string. */
|
||||
break;
|
||||
}
|
||||
|
||||
/* We know we have an expression. Convert any existing string
|
||||
to a Constant node. */
|
||||
|
@ -1331,13 +1358,15 @@ _PyPegen_FstringParser_ConcatFstring(Parser *p, FstringParser *state, const char
|
|||
/* Do nothing. No previous literal. */
|
||||
} else {
|
||||
/* Convert the existing last_str literal to a Constant node. */
|
||||
expr_ty str = make_str_node_and_del(p, &state->last_str, first_token, last_token);
|
||||
if (!str || ExprList_Append(&state->expr_list, str) < 0)
|
||||
expr_ty last_str = make_str_node_and_del(p, &state->last_str, first_token, last_token);
|
||||
if (!last_str || ExprList_Append(&state->expr_list, last_str) < 0) {
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
if (ExprList_Append(&state->expr_list, expression) < 0)
|
||||
if (ExprList_Append(&state->expr_list, expression) < 0) {
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
/* If recurse_lvl is zero, then we must be at the end of the
|
||||
|
@ -1373,8 +1402,9 @@ _PyPegen_FstringParser_Finish(Parser *p, FstringParser *state, Token* first_toke
|
|||
if (!state->last_str) {
|
||||
/* Create a zero length string. */
|
||||
state->last_str = PyUnicode_FromStringAndSize(NULL, 0);
|
||||
if (!state->last_str)
|
||||
if (!state->last_str) {
|
||||
goto error;
|
||||
}
|
||||
}
|
||||
return make_str_node_and_del(p, &state->last_str, first_token, last_token);
|
||||
}
|
||||
|
@ -1383,15 +1413,17 @@ _PyPegen_FstringParser_Finish(Parser *p, FstringParser *state, Token* first_toke
|
|||
last node in our expression list. */
|
||||
if (state->last_str) {
|
||||
expr_ty str = make_str_node_and_del(p, &state->last_str, first_token, last_token);
|
||||
if (!str || ExprList_Append(&state->expr_list, str) < 0)
|
||||
if (!str || ExprList_Append(&state->expr_list, str) < 0) {
|
||||
goto error;
|
||||
}
|
||||
}
|
||||
/* This has already been freed. */
|
||||
assert(state->last_str == NULL);
|
||||
|
||||
seq = ExprList_Finish(&state->expr_list, p->arena);
|
||||
if (!seq)
|
||||
if (!seq) {
|
||||
goto error;
|
||||
}
|
||||
|
||||
return _Py_JoinedStr(seq, first_token->lineno, first_token->col_offset,
|
||||
last_token->end_lineno, last_token->end_col_offset, p->arena);
|
||||
|
|
Loading…
Reference in New Issue