danhanson/tree-sitter-django
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

use single stack in scanner

Browse Source
This commit is contained in:
danhanson committed 2026-09-06 04:02:44 -05:00
1 parent 05465cd528
commit 140e099275
1 file changed
+91 -90
Download Patch File Download Diff File Expand all files Collapse all files
src/scanner.c
+91 -90
View File
@@ -4,9 +4,7 @@
#include <string.h> #include <string.h>
#define ERROR_SIZE 64 #define ERROR_SIZE 64
#define STACK_COUNT 3
#define NAME_SEP ' ' #define NAME_SEP ' '
#define STACK_SEP '\n'
enum TokenType { enum TokenType {
/* indicates that an error occurred */ /* indicates that an error occurred */
@@ -50,36 +48,57 @@ static const char *const separated_keywords[] = {
#define KEYWORD_SIZE 6 // "random", the longest of them #define KEYWORD_SIZE 6 // "random", the longest of them
typedef Array(int32_t) Name; typedef Array(int32_t) Name;
typedef Array(Name) Stack;
/* The kinds of tag that open a named region. PushBlock..PushVerbatim and
* PopBlock..PopVerbatim are declared in this order, so a kind is the offset of
* its token within either run. */
typedef enum {
BlockTag,
PartialTag,
VerbatimTag,
} TagKind;
/* one letter per kind, in TagKind order, used to serialize an open tag */
static const char tag_kind_chars[] = "bpv";
typedef struct {
TagKind kind;
Name name;
} OpenTag;
/* The open tags, innermost last. Only one stack is needed because the grammar
* nests these regions: a tag can only be closed by the tag that matches the
* innermost open one, so its pop token is the only one the parser ever asks
* for at a close tag. */
typedef Array(OpenTag) Stack;
struct Scanner { struct Scanner {
bool has_error; bool has_error;
union { union {
Stack stacks [STACK_COUNT]; Stack stack;
char error [ERROR_SIZE]; char error [ERROR_SIZE];
}; };
}; };
static Stack* get_stack_for_token(struct Scanner *scanner, enum TokenType token) { static TagKind kind_for_push(enum TokenType token) {
if (token < PopBlock || token > PushVerbatim) { return (TagKind) (token - PushBlock);
return NULL; }
}
// the pop and push tokens are in the same order, so each pair shares a stack static enum TokenType pop_token_for_kind(TagKind kind) {
return &scanner->stacks[(token - PopBlock) % STACK_COUNT]; return (enum TokenType) (PopBlock + kind);
} }
static void reset_scanner(struct Scanner *const scanner) { static void reset_scanner(struct Scanner *const scanner) {
if (scanner->has_error) { if (scanner->has_error) {
scanner->error[0] = '\0'; scanner->error[0] = '\0';
scanner->has_error = false; scanner->has_error = false;
// the error shares storage with the stack, which is now meaningless
memset(&scanner->stack, 0, sizeof(scanner->stack));
} else { } else {
for (unsigned i = 0; i < STACK_COUNT; ++i) { for (unsigned i = 0; i < scanner->stack.size; ++i) {
Stack *stack = &scanner->stacks[i]; array_delete(&array_get(&scanner->stack, i)->name);
for (unsigned i = 0; i < stack->size; ++i) {
array_delete(array_get(stack, i));
}
array_delete(stack);
} }
array_delete(&scanner->stack);
} }
} }
@@ -162,13 +181,12 @@ write_serialization_error:;
} }
} }
} else { } else {
// Ok scanner requires copying its stack // Ok scanner requires copying its stack, as "<kind><name> " per open tag
for (unsigned i = 0; i < STACK_COUNT; ++i) { for (unsigned i = 0; i < scanner->stack.size; ++i) {
Stack *stack = &scanner->stacks[i]; OpenTag *tag = array_get(&scanner->stack, i);
for (unsigned i = 0; i < stack->size; ++i) { iter += write_code(tag_kind_chars[tag->kind], iter);
Name *name = array_get(stack, i); for (unsigned j = 0; j < tag->name.size; ++j) {
for (unsigned j = 0; j < name->size; ++j) { int32_t code = *array_get(&tag->name, j);
int32_t code = *array_get(name, j);
unsigned write_amt = write_code(code, iter); unsigned write_amt = write_code(code, iter);
if (write_amt == 0) { if (write_amt == 0) {
scanner_error(scanner, "bad code from name"); scanner_error(scanner, "bad code from name");
@@ -178,8 +196,6 @@ write_serialization_error:;
} }
iter += write_code(NAME_SEP, iter); iter += write_code(NAME_SEP, iter);
} }
iter += write_code(STACK_SEP, iter);
}
} }
return iter - buffer; return iter - buffer;
} }
@@ -253,11 +269,10 @@ void tree_sitter_django_external_scanner_deserialize(
case '0': { case '0': {
scanner->has_error = false; scanner->has_error = false;
const char *end = buffer + length - bytes_read; const char *end = buffer + length - bytes_read;
Stack *stack = scanner->stacks; // NULL between tags, and the name being read inside one
Stack *stack_end = stack + STACK_COUNT;
Name *name = NULL; Name *name = NULL;
while (stack < stack_end) { while (buffer < end) {
bytes_read = read_code(buffer, &code); bytes_read = read_code(buffer, &code);
if (bytes_read == 0) { if (bytes_read == 0) {
scanner_error(scanner, "Bad utf8 byte"); scanner_error(scanner, "Bad utf8 byte");
@@ -268,34 +283,30 @@ void tree_sitter_django_external_scanner_deserialize(
scanner_error(scanner, "Scanner truncated by length"); scanner_error(scanner, "Scanner truncated by length");
return; return;
} }
if (code == STACK_SEP) {
++stack;
name = NULL;
continue;
} else if (code == NAME_SEP) {
if (name == NULL) { if (name == NULL) {
// an empty name was pushed (e.g. an anonymous verbatim block) // a tag starts with the letter for its kind
array_push(stack, (Name) array_new()); const char *kind = code > 0 && code < 128 ? strchr(tag_kind_chars, code) : NULL;
} else { if (kind == NULL) {
name = NULL; scanner_error(scanner, "Invalid tag kind");
}
} else {
if (name == NULL) {
if (!check_name_start_char(code)) {
scanner_error(scanner, "Invalid name, must start with letter");
return; return;
} }
array_push(stack, (Name) array_new()); OpenTag tag = { (TagKind) (kind - tag_kind_chars), array_new() };
name = array_back(stack); array_push(&scanner->stack, tag);
} else if (!check_name_char(code)) { name = &array_back(&scanner->stack)->name;
} else if (code == NAME_SEP) {
// an empty name means an unnamed tag (e.g. an anonymous verbatim)
name = NULL;
} else {
bool valid = name->size ? check_name_char(code) : check_name_start_char(code);
if (!valid) {
scanner_error(scanner, "Invalid character in name"); scanner_error(scanner, "Invalid character in name");
return; return;
} }
array_push(name, code); array_push(name, code);
} }
} }
if (buffer != end) { if (name != NULL) {
scanner_error(scanner, "Scanner deserialization finished with left over length"); scanner_error(scanner, "Scanner deserialization finished mid-name");
} }
} }
} }
@@ -337,6 +348,12 @@ static bool check_close_block(TSLexer *const lexer) {
return lexer->lookahead == '}'; return lexer->lookahead == '}';
} }
/* Marks the name just read as the token, then checks that only "%}" follows. */
static bool check_close_block_from(TSLexer *const lexer) {
lexer->mark_end(lexer);
return check_close_block(lexer);
}
const char inline_chars[] = "inline"; const char inline_chars[] = "inline";
static bool check_inline(TSLexer *const lexer) { static bool check_inline(TSLexer *const lexer) {
@@ -396,11 +413,14 @@ static bool check_endverbatim_close(TSLexer *const lexer, const Name *const name
* unrelated (e.g. unnamed) "{% verbatim %}...{% endverbatim %}" pair appear * unrelated (e.g. unnamed) "{% verbatim %}...{% endverbatim %}" pair appear
* unparsed inside a named verbatim block. */ * unparsed inside a named verbatim block. */
static bool scan_verbatim_content(struct Scanner *const scanner, TSLexer *const lexer) { static bool scan_verbatim_content(struct Scanner *const scanner, TSLexer *const lexer) {
Stack *stack = get_stack_for_token(scanner, PopVerbatim); if (scanner->stack.size == 0) {
if (stack->size == 0) {
return false; return false;
} }
Name *name = array_back(stack); OpenTag *tag = array_back(&scanner->stack);
if (tag->kind != VerbatimTag) {
return false;
}
Name *name = &tag->name;
bool consumed_any = false; bool consumed_any = false;
while (true) { while (true) {
if (lexer->eof(lexer)) { if (lexer->eof(lexer)) {
@@ -603,68 +623,49 @@ bool tree_sitter_django_external_scanner_scan(
is_empty = true; is_empty = true;
} }
// save most matched name in case we need to reread characters /* Only the innermost open tag can be closed here, so it is the one and only
Name *most_matched_name; * candidate: there is never a second name to try, and so never any input to
unsigned most_matched_amt = 0; * reread. */
for (unsigned token = PopBlock; token < PushBlock; ++token) { if (scanner->stack.size > 0) {
Stack *stack = get_stack_for_token(scanner, token); OpenTag *tag = array_back(&scanner->stack);
if (valid_symbols[token] && stack->size > 0) { enum TokenType token = pop_token_for_kind(tag->kind);
Name *name = array_back(stack); if (valid_symbols[token]) {
if (is_empty) { if (is_empty) {
// we matched a close block without an id // we matched a close block without an id
array_delete(name); array_delete(&tag->name);
array_pop(stack); array_pop(&scanner->stack);
lexer->result_symbol = token; lexer->result_symbol = token;
return true; return true;
} }
if (most_matched_amt > name->size) { unsigned match_amt = check_name(lexer, &tag->name);
continue; bool has_next_char = tag->name.size ? check_name_char(lexer->lookahead) : check_name_start_char(lexer->lookahead);
} if (match_amt == tag->name.size && !has_next_char && check_close_block_from(lexer)) {
if (most_matched_amt > 0) { array_delete(&tag->name);
for (int i = 0; i < most_matched_amt; ++i) { array_pop(&scanner->stack);
if (*array_get(most_matched_name, i) != *array_get(name, i)) {
continue;
}
}
}
unsigned match_amt = check_name(lexer, name) + most_matched_amt;
bool has_next_char = name->size ? check_name_char(lexer->lookahead) : check_name_start_char(lexer->lookahead);
if (match_amt == name->size && !has_next_char) {
// name matches and there are no remaining name chars from lexer
lexer->mark_end(lexer);
if (check_close_block(lexer)) {
array_delete(name);
array_pop(stack);
lexer->result_symbol = token; lexer->result_symbol = token;
return true; return true;
} else { }
// bad close block means no matches // the close tag names something other than the tag it would close
return false; return false;
} }
} }
most_matched_amt = match_amt;
most_matched_name = name;
}
}
for (unsigned token = PushBlock; token <= PushVerbatim; ++token) { for (unsigned token = PushBlock; token <= PushVerbatim; ++token) {
if (valid_symbols[token]) { if (valid_symbols[token]) {
Stack *stack = get_stack_for_token(scanner, token); TagKind kind = kind_for_push(token);
if (is_empty && token == PushVerbatim) { if (is_empty && token == PushVerbatim) {
// use empty string for name to indicate empty push // use empty string for name to indicate empty push
array_push(stack, (Name) array_new()); OpenTag tag = { kind, array_new() };
array_push(&scanner->stack, tag);
lexer->result_symbol = token; lexer->result_symbol = token;
return true; return true;
} }
// do not push stack until we are done with most_matched_name
Name name = array_new(); Name name = array_new();
if (most_matched_amt > 0) {
array_extend(&name, most_matched_amt, most_matched_name);
}
if (read_name(lexer, &name)) { if (read_name(lexer, &name)) {
// validate tokens after name // validate tokens after name
lexer->mark_end(lexer); lexer->mark_end(lexer);
if (check_close_block(lexer) || token == PushPartial && check_inline(lexer)) { if (check_close_block(lexer) || token == PushPartial && check_inline(lexer)) {
array_push(stack, name); OpenTag tag = { kind, name };
array_push(&scanner->stack, tag);
lexer->result_symbol = token; lexer->result_symbol = token;
return true; return true;
} }