tree-sitter/src/runtime/parser.c

988 lines
32 KiB
C
Raw Normal View History

2015-09-19 13:19:49 -07:00
#include "runtime/parser.h"
#include <assert.h>
#include <stdio.h>
#include <stdbool.h>
#include "tree_sitter/runtime.h"
#include "tree_sitter/parser.h"
#include "runtime/tree.h"
#include "runtime/lexer.h"
#include "runtime/length.h"
2016-02-17 20:41:29 -08:00
#include "runtime/array.h"
#include "runtime/language.h"
2016-01-15 15:08:42 -08:00
#include "runtime/alloc.h"
2014-10-08 16:49:52 -07:00
/*
2014-10-14 09:32:11 -07:00
* Debugging
2014-10-08 16:49:52 -07:00
*/
#define LOG(...) \
2015-10-14 21:52:13 -07:00
if (self->lexer.debugger.debug_fn) { \
snprintf(self->lexer.debug_buffer, TS_DEBUG_BUFFER_SIZE, __VA_ARGS__); \
self->lexer.debugger.debug_fn(self->lexer.debugger.payload, \
TSDebugTypeParse, self->lexer.debug_buffer); \
}
#define LOG_ACTION(...) \
LOG(__VA_ARGS__); \
if (self->print_debugging_graphs) { \
fprintf(stderr, "graph {\nlabel=\""); \
fprintf(stderr, __VA_ARGS__); \
fprintf(stderr, "\"\n}\n\n"); \
}
2016-03-02 09:55:25 -08:00
#define LOG_STACK() \
if (self->print_debugging_graphs) { \
char *graph_string = \
ts_stack_dot_graph(self->stack, self->language->symbol_names); \
fputs(graph_string, stderr); \
fputs("\n\n", stderr); \
ts_free(graph_string); \
}
2016-03-02 09:55:25 -08:00
#define SYM_NAME(symbol) ts_language_symbol_name(self->language, symbol)
2014-10-13 21:20:08 -07:00
#define BOOL_STRING(value) (value ? "true" : "false")
2016-02-17 14:45:00 -08:00
struct LookaheadState {
TSTree *reusable_subtree;
size_t reusable_subtree_pos;
bool is_verifying;
2016-02-17 14:45:00 -08:00
};
typedef struct {
TSSymbol symbol;
TSStateId next_state;
size_t depth;
size_t skipped_subtree_count;
size_t in_progress_state_count;
} ErrorRepair;
typedef enum {
ParseActionFailed,
ParseActionUpdated,
ParseActionRemoved,
} ParseActionResult;
typedef enum {
ReduceFailed,
ReduceSucceeded,
ReduceMerged,
ReduceStoppedAtError,
} ReduceResult;
2014-10-14 09:32:11 -07:00
/*
* Private
*/
2016-03-03 10:16:10 -08:00
static StackSlice ts_parser__pop_one(TSParser *self, int head_index, int count) {
StackSliceArray slices = ts_stack_pop(self->stack, head_index, count, true);
assert(slices.size > 0);
assert(slices.contents[0].head_index == head_index);
for (size_t i = 1; i < slices.size; i++) {
ts_tree_array_clear(&slices.contents[i].trees);
array_delete(&slices.contents[i].trees);
ts_stack_remove_head(self->stack, slices.contents[i].head_index);
}
2016-03-03 10:16:10 -08:00
return slices.contents[0];
}
2016-01-22 22:10:18 -07:00
static ParseActionResult ts_parser__breakdown_top_of_stack(TSParser *self,
int head) {
TSTree *last_child = NULL;
do {
2016-03-03 10:16:10 -08:00
StackSliceArray slices = ts_stack_pop(self->stack, head, 1, false);
if (!slices.size)
return ParseActionFailed;
2016-03-03 10:16:10 -08:00
assert(slices.size > 0);
/*
* Since only one entry (not counting extra trees) is being popped from the
* stack, there should only be one possible array of removed trees.
*/
2016-03-03 10:16:10 -08:00
for (size_t i = 0; i < slices.size; i++) {
StackSlice slice = slices.contents[i];
TreeArray removed_trees = slice.trees;
TSTree *parent = *array_front(&removed_trees);
2016-03-03 10:16:10 -08:00
int head_index = slice.head_index;
LOG("breakdown_pop sym:%s, size:%lu", SYM_NAME(parent->symbol),
ts_tree_total_size(parent).chars);
StackPushResult last_push = StackPushContinued;
TSStateId state = ts_stack_top_state(self->stack, head_index);
for (size_t j = 0; j < parent->child_count; j++) {
last_child = parent->children[j];
2015-12-22 14:20:58 -08:00
if (!last_child->extra) {
TSParseAction action =
ts_language_last_action(self->language, state, last_child->symbol);
assert(action.type == TSParseActionTypeShift);
state = action.data.to_state;
}
LOG("breakdown_push sym:%s, size:%lu", SYM_NAME(last_child->symbol),
ts_tree_total_size(last_child).chars);
2016-02-25 21:46:13 -08:00
last_push = ts_stack_push(self->stack, head_index, last_child, state);
if (last_push == StackPushFailed)
2016-01-21 14:07:38 -07:00
goto error;
}
2016-03-03 10:16:10 -08:00
for (size_t j = 1, count = slice.trees.size; j < count; j++) {
TSTree *tree = slice.trees.contents[j];
2016-02-25 21:46:13 -08:00
last_push = ts_stack_push(self->stack, head_index, tree, state);
if (last_push == StackPushFailed)
2016-01-21 14:07:38 -07:00
goto error;
}
2016-01-21 14:07:38 -07:00
if (i == 0)
assert(last_push != StackPushMerged);
2016-01-21 14:07:38 -07:00
else
assert(last_push == StackPushMerged);
2016-02-02 12:03:11 -08:00
for (size_t j = 0, count = removed_trees.size; j < count; j++)
ts_tree_release(removed_trees.contents[j]);
array_delete(&removed_trees);
}
} while (last_child && last_child->child_count > 0);
return ParseActionUpdated;
2016-01-21 14:07:38 -07:00
error:
return ParseActionFailed;
}
static void ts_parser__pop_reusable_subtree(LookaheadState *state);
/*
2015-10-14 21:52:13 -07:00
* Replace the parser's reusable_subtree with its first non-fragile descendant.
* Return true if a suitable descendant is found, false otherwise.
*/
static void ts_parser__breakdown_reusable_subtree(LookaheadState *state) {
do {
if (state->reusable_subtree->symbol == ts_builtin_sym_error) {
ts_parser__pop_reusable_subtree(state);
return;
}
if (state->reusable_subtree->child_count == 0) {
ts_parser__pop_reusable_subtree(state);
return;
}
state->reusable_subtree = state->reusable_subtree->children[0];
} while (ts_tree_is_fragile(state->reusable_subtree));
}
/*
* Replace the parser's reusable_subtree with its largest right neighbor, or
* NULL if no right neighbor exists.
*/
static void ts_parser__pop_reusable_subtree(LookaheadState *state) {
state->reusable_subtree_pos += ts_tree_total_chars(state->reusable_subtree);
while (state->reusable_subtree) {
TSTree *parent = state->reusable_subtree->context.parent;
size_t next_index = state->reusable_subtree->context.index + 1;
if (parent && parent->child_count > next_index) {
state->reusable_subtree = parent->children[next_index];
return;
}
state->reusable_subtree = parent;
2014-07-20 20:27:33 -07:00
}
}
static bool ts_parser__can_reuse(TSParser *self, int head, TSTree *subtree) {
if (subtree->symbol == ts_builtin_sym_error)
return false;
if (ts_tree_is_fragile(subtree)) {
2015-12-22 14:20:58 -08:00
if (subtree->parse_state != ts_stack_top_state(self->stack, head))
return false;
}
TSStateId state = ts_stack_top_state(self->stack, head);
if (subtree->lex_state != TS_TREE_STATE_INDEPENDENT) {
TSStateId lex_state = self->language->lex_states[state];
2015-12-22 14:20:58 -08:00
if (subtree->lex_state != lex_state)
return false;
}
2015-12-29 21:17:31 -08:00
const TSParseAction action =
ts_language_last_action(self->language, state, subtree->symbol);
2015-12-29 11:20:52 -08:00
if (action.type == TSParseActionTypeError || action.can_hide_split)
return false;
2015-12-29 11:20:52 -08:00
if (subtree->extra && !action.extra)
return false;
return true;
}
/*
* Advance the parser's lookahead subtree. If there is a reusable subtree
* at the correct position in the parser's previous tree, use that. Otherwise,
* run the lexer.
*/
static TSTree *ts_parser__get_next_lookahead(TSParser *self, int head) {
2016-02-17 20:41:29 -08:00
LookaheadState *state = array_get(&self->lookahead_states, head);
TSLength position = ts_stack_top_position(self->stack, head);
2014-10-14 09:32:11 -07:00
while (state->reusable_subtree) {
if (state->reusable_subtree_pos > position.chars) {
break;
}
2014-10-14 09:32:11 -07:00
if (state->reusable_subtree_pos < position.chars) {
LOG("past_reusable sym:%s", SYM_NAME(state->reusable_subtree->symbol));
ts_parser__pop_reusable_subtree(state);
continue;
}
2014-10-14 09:32:11 -07:00
2015-12-22 14:20:58 -08:00
if (state->reusable_subtree->has_changes) {
if (state->is_verifying && state->reusable_subtree->child_count == 0) {
ts_parser__breakdown_top_of_stack(self, head);
state->is_verifying = false;
}
LOG("breakdown_changed sym:%s", SYM_NAME(state->reusable_subtree->symbol));
ts_parser__breakdown_reusable_subtree(state);
continue;
}
if (!ts_parser__can_reuse(self, head, state->reusable_subtree)) {
2015-12-24 22:05:54 -08:00
LOG("breakdown_unreusable sym:%s",
SYM_NAME(state->reusable_subtree->symbol));
ts_parser__breakdown_reusable_subtree(state);
continue;
2014-10-14 09:32:11 -07:00
}
TSTree *result = state->reusable_subtree;
TSLength size = ts_tree_total_size(result);
LOG("reuse sym:%s size:%lu extra:%d", SYM_NAME(result->symbol), size.chars,
2015-12-22 14:20:58 -08:00
result->extra);
ts_parser__pop_reusable_subtree(state);
ts_tree_retain(result);
return result;
2014-10-14 09:32:11 -07:00
}
2014-10-22 12:54:46 -07:00
2015-12-24 10:21:42 -08:00
ts_lexer_reset(&self->lexer, position);
TSStateId parse_state = ts_stack_top_state(self->stack, head);
TSStateId lex_state = self->language->lex_states[parse_state];
LOG("lex state:%d", lex_state);
return self->language->lex_fn(&self->lexer, lex_state, false);
}
static int ts_parser__split(TSParser *self, int head) {
int result = ts_stack_split(self->stack, head);
assert(result == (int)self->lookahead_states.size);
2016-02-17 20:41:29 -08:00
LookaheadState lookahead_state = *array_get(&self->lookahead_states, head);
array_push(&self->lookahead_states, lookahead_state);
return result;
}
static void ts_parser__remove_head(TSParser *self, int head) {
2016-02-17 20:41:29 -08:00
array_erase(&self->lookahead_states, head);
ts_stack_remove_head(self->stack, head);
}
static int ts_parser__select_tree(void *data, TSTree *left, TSTree *right) {
2016-03-02 09:55:25 -08:00
if (!left || left->symbol == ts_builtin_sym_error)
return 1;
2016-03-02 09:55:25 -08:00
if (!right || right->symbol == ts_builtin_sym_error)
return -1;
2015-12-24 22:04:20 -08:00
TSParser *self = data;
int comparison = ts_tree_compare(left, right);
switch (comparison) {
case -1:
LOG_ACTION("select tree:%s, over_tree:%s", SYM_NAME(left->symbol),
SYM_NAME(right->symbol));
break;
case 1:
LOG_ACTION("select tree:%s, over_tree:%s", SYM_NAME(right->symbol),
SYM_NAME(left->symbol));
break;
}
return comparison;
}
/*
* Parse Actions
*/
static ParseActionResult ts_parser__shift(TSParser *self, int head,
2016-01-22 22:10:18 -07:00
TSStateId parse_state,
TSTree *lookahead) {
2016-02-25 21:46:13 -08:00
switch (ts_stack_push(self->stack, head, lookahead, parse_state)) {
case StackPushFailed:
return ParseActionFailed;
case StackPushMerged:
LOG("merge head:%d", head);
2016-02-17 20:41:29 -08:00
array_erase(&self->lookahead_states, head);
return ParseActionRemoved;
2016-02-12 14:07:30 -08:00
default:
return ParseActionUpdated;
}
}
2016-01-22 22:10:18 -07:00
static ParseActionResult ts_parser__shift_extra(TSParser *self, int head,
TSStateId state,
TSTree *lookahead) {
TSSymbolMetadata metadata = self->language->symbol_metadata[lookahead->symbol];
if (metadata.structural && ts_stack_head_count(self->stack) > 1) {
TSTree *copy = ts_tree_make_copy(lookahead);
if (!copy)
return ParseActionFailed;
copy->extra = true;
ParseActionResult result = ts_parser__shift(self, head, state, copy);
ts_tree_release(copy);
return result;
} else {
lookahead->extra = true;
return ts_parser__shift(self, head, state, lookahead);
}
}
static ReduceResult ts_parser__reduce(TSParser *self, int head, TSSymbol symbol,
int child_count, bool extra, bool fragile) {
2016-02-17 20:41:29 -08:00
array_clear(&self->reduce_parents);
2016-03-02 09:55:25 -08:00
TSSymbolMetadata metadata = ts_language_symbol_metadata(self->language, symbol);
2016-03-03 10:16:10 -08:00
StackSliceArray slices =
ts_stack_pop(self->stack, head, child_count, false);
2016-03-03 10:16:10 -08:00
if (!slices.size)
return ReduceFailed;
2016-03-03 10:16:10 -08:00
if (slices.contents[0].trees.size &&
slices.contents[0].trees.contents[0]->symbol == ts_builtin_sym_error) {
if (self->partial_pop.size) {
ts_tree_array_clear(&self->partial_pop);
array_delete(&self->partial_pop);
}
2016-03-03 10:16:10 -08:00
self->partial_pop = slices.contents[0].trees;
for (size_t i = 1; i < slices.size; i++) {
ts_tree_array_clear(&slices.contents[i].trees);
array_delete(&slices.contents[i].trees);
}
return ReduceStoppedAtError;
}
size_t removed_heads = 0;
2016-03-03 10:16:10 -08:00
for (size_t i = 0; i < slices.size; i++) {
StackSlice slice = slices.contents[i];
2015-11-20 00:01:53 -08:00
/*
* If the same set of trees led to a previous stack head, reuse the parent
* tree that was added to that head. Otherwise, create a new parent node
* for this set of trees.
2015-11-20 00:01:53 -08:00
*/
TSTree *parent = NULL;
size_t trailing_extra_count = 0;
2016-03-03 10:16:10 -08:00
for (size_t j = slice.trees.size - 1; j + 1 > 0; j--) {
if (slice.trees.contents[j]->extra)
trailing_extra_count++;
else
2015-11-20 00:01:53 -08:00
break;
}
2016-03-03 10:16:10 -08:00
size_t popped_child_count = slice.trees.size - trailing_extra_count;
2016-02-25 21:46:13 -08:00
parent = ts_tree_make_node(symbol, popped_child_count,
2016-03-03 10:16:10 -08:00
slice.trees.contents, metadata);
if (!parent) {
2016-03-03 10:16:10 -08:00
for (size_t i = 0; i < slice.trees.size; i++)
ts_tree_release(slice.trees.contents[i]);
array_delete(&slice.trees);
goto error;
}
2016-02-17 20:41:29 -08:00
if (!array_push(&self->reduce_parents, parent))
goto error;
2016-03-03 10:16:10 -08:00
int new_head = slice.head_index - removed_heads;
2015-11-20 00:01:53 -08:00
if (i > 0) {
if (symbol == ts_builtin_sym_error) {
removed_heads++;
ts_stack_remove_head(self->stack, new_head);
continue;
}
/*
2016-02-17 14:45:00 -08:00
* If the stack has split in the process of popping, create a duplicate
* of
* the lookahead state for this head, for the new head.
*/
LOG("split_during_reduce new_head:%d", new_head);
2016-02-17 20:41:29 -08:00
LookaheadState lookahead_state = *array_get(&self->lookahead_states, head);
if (!array_push(&self->lookahead_states, lookahead_state))
goto error;
}
2015-11-20 00:01:53 -08:00
/*
* If the parent node is extra, then do not change the state when pushing
* it. Otherwise, proceed to the state given in the parse table for the
* new parent symbol.
*/
TSStateId state;
TSStateId top_state = ts_stack_top_state(self->stack, new_head);
if (parent->parse_state != TS_TREE_STATE_ERROR)
2015-12-22 14:20:58 -08:00
parent->parse_state = top_state;
2015-11-20 00:01:53 -08:00
if (extra) {
2015-12-22 14:20:58 -08:00
parent->extra = true;
2015-11-20 00:01:53 -08:00
state = top_state;
} else {
TSParseAction action =
ts_language_last_action(self->language, top_state, symbol);
2015-11-20 00:01:53 -08:00
if (child_count == -1) {
state = 0;
} else {
2015-11-20 00:01:53 -08:00
assert(action.type == TSParseActionTypeShift);
state = action.data.to_state;
}
2015-11-20 00:01:53 -08:00
}
2015-11-20 00:01:53 -08:00
/*
* If the given state already existed at a different head of the stack,
* then remove the lookahead state for the head.
*/
2016-02-25 21:46:13 -08:00
switch (ts_stack_push(self->stack, new_head, parent, state)) {
case StackPushFailed:
ts_tree_release(parent);
goto error;
case StackPushMerged:
LOG("merge_during_reduce head:%d", new_head);
2016-02-17 20:41:29 -08:00
array_erase(&self->lookahead_states, new_head);
removed_heads++;
continue;
case StackPushContinued:
break;
}
if (trailing_extra_count > 0) {
for (size_t j = 0; j < trailing_extra_count; j++) {
2016-03-03 10:16:10 -08:00
size_t index = slice.trees.size - trailing_extra_count + j;
TSTree *tree = slice.trees.contents[index];
2016-02-25 21:46:13 -08:00
switch (ts_stack_push(self->stack, new_head, tree, state)) {
case StackPushFailed:
return ReduceFailed;
case StackPushMerged:
2016-02-17 20:41:29 -08:00
array_erase(&self->lookahead_states, new_head);
removed_heads++;
2016-02-02 12:03:11 -08:00
break;
case StackPushContinued:
break;
}
2016-02-02 12:03:11 -08:00
ts_tree_release(tree);
}
}
}
for (size_t i = 0; i < self->reduce_parents.size; i++) {
2016-02-17 20:41:29 -08:00
TSTree **parent = array_get(&self->reduce_parents, i);
if (fragile || self->is_split || ts_stack_head_count(self->stack) > 1) {
2015-12-22 14:20:58 -08:00
(*parent)->fragile_left = true;
(*parent)->fragile_right = true;
(*parent)->parse_state = TS_TREE_STATE_ERROR;
}
ts_tree_release(*parent);
}
2016-03-03 10:16:10 -08:00
if (removed_heads < slices.size)
return ReduceSucceeded;
else
return ReduceMerged;
error:
return ReduceFailed;
2015-06-15 15:24:15 -07:00
}
static ParseActionResult ts_parser__handle_error(TSParser *self, int head,
2016-01-22 22:10:18 -07:00
TSTree *lookahead) {
TSTree *next_lookahead = self->language->lex_fn(&self->lexer, 0, true);
lookahead->size = ts_length_add(lookahead->size, next_lookahead->padding);
next_lookahead->padding = ts_length_zero();
if (!ts_parser__shift(self, head, ts_parse_state_error, lookahead))
goto error;
while (ts_language_symbol_metadata(self->language,next_lookahead->symbol).extra) {
if (!ts_parser__shift_extra(self, head, ts_parse_state_error, next_lookahead))
goto error;
ts_tree_release(next_lookahead);
next_lookahead = self->language->lex_fn(&self->lexer, 0, true);
}
if (next_lookahead->symbol == ts_builtin_sym_end) {
2016-03-03 10:16:10 -08:00
StackSlice result = ts_parser__pop_one(self, head, -1);
TSTree *parent = ts_tree_make_node(ts_builtin_sym_start, result.trees.size,
result.trees.contents, ts_language_symbol_metadata(self->language, ts_builtin_sym_start));
ts_stack_push(self->stack, head, parent, 0);
ts_tree_release(parent);
ts_tree_release(next_lookahead);
return ParseActionRemoved;
}
TSParseAction action = ts_language_last_action(
self->language, ts_parse_state_error, next_lookahead->symbol);
assert(action.type == TSParseActionTypeShift);
LOG("resume_without_context state:%d", action.data.to_state);
ParseActionResult result =
ts_parser__shift(self, head, action.data.to_state, next_lookahead);
ts_tree_release(next_lookahead);
return result;
error:
ts_tree_release(next_lookahead);
return ParseActionFailed;
}
static bool ts_parser__error_repair_is_better(ErrorRepair old, ErrorRepair new) {
if (new.in_progress_state_count > old.in_progress_state_count)
return true;
if (new.in_progress_state_count < old.in_progress_state_count)
return false;
return new.skipped_subtree_count < old.skipped_subtree_count;
}
2014-07-20 20:27:33 -07:00
static ParseActionResult ts_parser__repair_error(TSParser *self, int head_index,
TSTree *lookahead,
const TSParseAction *actions,
size_t action_count) {
StackEntry *entry_below = ts_stack_head(self->stack, head_index);
while (entry_below && entry_below->state == ts_parse_state_error)
entry_below = ts_stack_entry_next(entry_below, 0);
bool has_repair = false;
ErrorRepair best_repair;
// Count how many non-extra trees were already popped, not counting the error.
size_t count_above_error = 0;
for (size_t i = 1; i < self->partial_pop.size; i++)
if (!self->partial_pop.contents[i]->extra)
count_above_error++;
// Compare all of the reduce actions that extend back to the error. Find the
// one that best fits the nodes *below* the error on the stack.
for (size_t i = 0; i < action_count; i++) {
TSParseAction action = actions[i];
if (action.type != TSParseActionTypeReduce ||
action.data.child_count <= count_above_error)
continue;
StackEntry *entry = entry_below;
size_t depth = 0;
size_t child_count = 0;
size_t in_progress_state_count = 0;
// Move down the stack until a state is reached in which the reduced symbol
// is expected, and the current lookahead symbol is expected to follow it.
for (;;) {
TSStateId state = entry ? entry->state : 0;
if (ts_language_symbol_is_in_progress(self->language, state, action.data.symbol))
in_progress_state_count++;
if (child_count + count_above_error >= action.data.child_count &&
in_progress_state_count > 0) {
TSParseAction entry_action =
ts_language_last_action(self->language, state, action.data.symbol);
if (entry_action.type == TSParseActionTypeShift) {
TSStateId next_state = entry_action.data.to_state;
TSParseAction next_action = ts_language_last_action(
self->language, next_state, lookahead->symbol);
if (next_action.type != TSParseActionTypeError) {
ErrorRepair repair = {
.symbol = action.data.symbol,
.next_state = next_state,
.depth = depth,
.skipped_subtree_count = child_count + count_above_error - action.data.child_count,
.in_progress_state_count = in_progress_state_count,
};
LOG_ACTION("REPAIR_GOOD sym:%s, in_progress:%lu, skipped:%lu!",
SYM_NAME(action.data.symbol),
in_progress_state_count,
repair.skipped_subtree_count);
if (!has_repair ||
ts_parser__error_repair_is_better(best_repair, repair)) {
has_repair = true;
best_repair = repair;
break;
}
}
}
2014-07-20 20:27:33 -07:00
}
if (entry) {
if (entry->state != ts_parse_state_error)
child_count++;
depth++;
entry = ts_stack_entry_next(entry, 0);
} else {
break;
}
}
}
if (!has_repair) {
ts_tree_array_clear(&self->partial_pop);
array_delete(&self->partial_pop);
ts_parser__remove_head(self, head_index);
return ParseActionRemoved;
}
2014-09-29 10:33:56 -07:00
// Pop any trees that were skipped. Make a new extra error node that contains
// them and the error leaf node.
2016-03-03 10:16:10 -08:00
StackSlice slice = ts_parser__pop_one(self, head_index, best_repair.skipped_subtree_count);
array_push(&slice.trees, *array_front(&self->partial_pop));
TSTree *error = ts_tree_make_node(
ts_builtin_sym_error,
2016-03-03 10:16:10 -08:00
slice.trees.size,
slice.trees.contents,
ts_language_symbol_metadata(self->language, ts_builtin_sym_error));
// Pop any additional trees that are needed for the chosen reduce action. Make
// a new wrapper node of the chosen symbol that contains them, the error node,
// and the trees that were popped above the error node.
2016-03-03 10:16:10 -08:00
slice = ts_parser__pop_one(self, head_index, best_repair.depth - (error->child_count - 1));
array_push(&slice.trees, error);
for (size_t i = 1; i < self->partial_pop.size; i++)
2016-03-03 10:16:10 -08:00
array_push(&slice.trees, self->partial_pop.contents[i]);
array_delete(&self->partial_pop);
TSTree *parent = ts_tree_make_node(
best_repair.symbol,
2016-03-03 10:16:10 -08:00
slice.trees.size,
slice.trees.contents,
ts_language_symbol_metadata(self->language, best_repair.symbol));
StackPushResult push_result = ts_stack_push(self->stack, head_index, parent, best_repair.next_state);
ts_tree_release(parent);
switch (push_result) {
case StackPushFailed:
return ParseActionFailed;
case StackPushMerged:
return ParseActionRemoved;
default:
return ParseActionUpdated;
2014-07-20 20:27:33 -07:00
}
}
static ParseActionResult ts_parser__start(TSParser *self, TSInput input,
2016-01-22 22:10:18 -07:00
TSTree *previous_tree) {
if (previous_tree) {
LOG("parse_after_edit");
} else {
LOG("new_parse");
}
ts_lexer_set_input(&self->lexer, input);
2015-10-14 21:52:13 -07:00
ts_stack_clear(self->stack);
ts_stack_set_tree_selection_callback(self->stack, self,
ts_parser__select_tree);
LookaheadState lookahead_state = {
.reusable_subtree = previous_tree,
.reusable_subtree_pos = 0,
.is_verifying = false,
};
2016-02-17 20:41:29 -08:00
array_clear(&self->lookahead_states);
array_push(&self->lookahead_states, lookahead_state);
2015-12-24 22:04:20 -08:00
self->finished_tree = NULL;
return ParseActionUpdated;
}
static ParseActionResult ts_parser__accept(TSParser *self, int head) {
2016-03-03 10:16:10 -08:00
StackSliceArray slices = ts_stack_pop(self->stack, head, -1, true);
if (!slices.size)
goto error;
2015-12-24 22:04:20 -08:00
2016-03-03 10:16:10 -08:00
for (size_t j = 0; j < slices.size; j++) {
StackSlice slice = slices.contents[j];
TreeArray trees = slice.trees;
for (size_t i = trees.size - 1; i + 1 > 0; i--) {
if (!trees.contents[i]->extra) {
TSTree *root = trees.contents[i];
if (!array_splice(&trees, i, 1, root->child_count, root->children))
goto error;
ts_tree_set_children(root, trees.size, trees.contents);
if (!trees.size)
array_delete(&trees);
2016-03-03 10:16:10 -08:00
ts_parser__remove_head(self, slice.head_index);
int comparison = ts_parser__select_tree(self, self->finished_tree, root);
if (comparison > 0) {
ts_tree_release(self->finished_tree);
self->finished_tree = root;
} else {
ts_tree_release(root);
}
2015-12-24 22:04:20 -08:00
break;
}
}
}
return ParseActionRemoved;
error:
2016-03-03 10:16:10 -08:00
if (slices.size) {
StackSlice slice = *array_front(&slices);
for (size_t i = 0; i < slice.trees.size; i++)
ts_tree_release(slice.trees.contents[i]);
array_delete(&slice.trees);
}
return ParseActionFailed;
2014-08-09 01:03:55 -07:00
}
2015-10-07 12:58:32 -07:00
/*
* Continue performing parse actions for the given head until the current
* lookahead symbol is consumed.
*/
static ParseActionResult ts_parser__consume_lookahead(TSParser *self, int head,
2016-01-22 22:10:18 -07:00
TSTree *lookahead) {
2015-10-07 12:58:32 -07:00
for (;;) {
2015-10-14 21:52:13 -07:00
TSStateId state = ts_stack_top_state(self->stack, head);
2015-12-29 11:20:52 -08:00
size_t action_count;
2015-12-29 21:17:31 -08:00
const TSParseAction *actions = ts_language_actions(
self->language, state, lookahead->symbol, &action_count);
2015-10-07 12:58:32 -07:00
/*
* If there are multiple actions for the current state and lookahead symbol,
* split the stack so that each one can be performed. If there is a `SHIFT`
* action, it will always appear *last* in the list of actions. Perform it
* on the original stack head and return.
*/
bool repaired_error = false;
2015-12-29 11:20:52 -08:00
for (size_t i = 0; i < action_count; i++) {
TSParseAction action = actions[i];
2015-10-07 12:58:32 -07:00
int current_head;
2015-12-29 11:20:52 -08:00
if (i == action_count - 1) {
if (repaired_error && action.type == TSParseActionTypeReduce)
action.type = TSParseActionTypeError;
2015-10-07 12:58:32 -07:00
current_head = head;
} else {
if (repaired_error)
continue;
current_head = ts_parser__split(self, head);
LOG_ACTION("split_action from_head:%d, new_head:%d", head, current_head);
2015-10-07 12:58:32 -07:00
}
LOG_STACK();
LookaheadState *lookahead_state =
2016-02-17 20:41:29 -08:00
array_get(&self->lookahead_states, current_head);
2015-10-07 12:58:32 -07:00
switch (action.type) {
case TSParseActionTypeError:
if (lookahead_state->is_verifying) {
ts_parser__breakdown_top_of_stack(self, current_head);
lookahead_state->is_verifying = false;
return ParseActionRemoved;
}
2015-10-14 21:52:13 -07:00
if (ts_stack_head_count(self->stack) == 1) {
LOG_ACTION("handle_error %s", SYM_NAME(lookahead->symbol));
switch (ts_parser__handle_error(self, current_head, lookahead)) {
case ParseActionFailed:
return ParseActionFailed;
case ParseActionUpdated:
return ParseActionUpdated;
case ParseActionRemoved:
return ts_parser__accept(self, current_head);
2015-12-24 22:04:20 -08:00
}
2015-10-07 12:58:32 -07:00
} else {
LOG_ACTION("bail current_head:%d", current_head);
ts_parser__remove_head(self, current_head);
return ParseActionRemoved;
2015-10-07 12:58:32 -07:00
}
case TSParseActionTypeShift:
if (action.extra) {
LOG_ACTION("shift_extra");
return ts_parser__shift_extra(self, current_head, state, lookahead);
} else {
LOG_ACTION("shift state:%u", action.data.to_state);
lookahead_state->is_verifying = (lookahead->child_count > 0);
TSStateId state = action.data.to_state;
return ts_parser__shift(self, current_head, state, lookahead);
}
2015-10-07 12:58:32 -07:00
case TSParseActionTypeReduce:
lookahead_state->is_verifying = false;
2015-10-07 12:58:32 -07:00
if (action.extra) {
LOG_ACTION("reduce_extra sym:%s", SYM_NAME(action.data.symbol));
ts_parser__reduce(self, current_head, action.data.symbol, 1, true,
false);
} else {
LOG_ACTION("reduce sym:%s, child_count:%u, fragile:%s",
SYM_NAME(action.data.symbol), action.data.child_count,
BOOL_STRING(action.fragile));
switch (ts_parser__reduce(self, current_head, action.data.symbol,
2016-01-22 22:10:18 -07:00
action.data.child_count, false,
action.fragile)) {
case ReduceFailed:
return ParseActionFailed;
case ReduceMerged:
if (current_head == head)
return ParseActionRemoved;
break;
case ReduceSucceeded:
break;
case ReduceStoppedAtError:
repaired_error = true;
LOG_ACTION("repair head:%d", current_head);
switch (ts_parser__repair_error(self, current_head, lookahead,
actions, action_count)) {
case ParseActionFailed:
return ParseActionFailed;
case ParseActionRemoved:
LOG_ACTION("repair_failed");
break;
default:
break;
}
break;
}
}
2015-10-07 12:58:32 -07:00
break;
case TSParseActionTypeAccept:
LOG_ACTION("accept");
return ts_parser__accept(self, current_head);
2015-10-07 12:58:32 -07:00
}
}
}
}
2015-08-16 19:53:34 -07:00
/*
* Public
*/
bool ts_parser_init(TSParser *self) {
ts_lexer_init(&self->lexer);
self->finished_tree = NULL;
self->stack = NULL;
2016-02-17 20:41:29 -08:00
array_init(&self->lookahead_states);
array_init(&self->reduce_parents);
self->stack = ts_stack_new();
if (!self->stack)
goto error;
2016-02-17 20:41:29 -08:00
if (!array_grow(&self->lookahead_states, 4))
goto error;
2016-02-17 20:41:29 -08:00
if (!array_grow(&self->reduce_parents, 4))
goto error;
return true;
error:
if (self->stack) {
ts_stack_delete(self->stack);
self->stack = NULL;
}
if (self->lookahead_states.contents)
2016-02-17 20:41:29 -08:00
array_delete(&self->lookahead_states);
if (self->reduce_parents.contents)
2016-02-17 20:41:29 -08:00
array_delete(&self->reduce_parents);
return false;
2015-08-16 19:53:34 -07:00
}
2015-10-14 21:52:13 -07:00
void ts_parser_destroy(TSParser *self) {
if (self->stack)
ts_stack_delete(self->stack);
if (self->lookahead_states.contents)
2016-02-17 20:41:29 -08:00
array_delete(&self->lookahead_states);
if (self->reduce_parents.contents)
2016-02-17 20:41:29 -08:00
array_delete(&self->reduce_parents);
ts_tree_array_clear(&self->partial_pop);
if (self->partial_pop.contents)
array_delete(&self->partial_pop);
2015-08-16 19:53:34 -07:00
}
2015-10-14 21:52:13 -07:00
TSDebugger ts_parser_debugger(const TSParser *self) {
return self->lexer.debugger;
2015-08-16 19:53:34 -07:00
}
2015-10-14 21:52:13 -07:00
void ts_parser_set_debugger(TSParser *self, TSDebugger debugger) {
self->lexer.debugger = debugger;
2015-08-16 19:53:34 -07:00
}
2015-10-14 21:52:13 -07:00
TSTree *ts_parser_parse(TSParser *self, TSInput input, TSTree *previous_tree) {
ts_parser__start(self, input, previous_tree);
size_t max_position = 0;
for (;;) {
TSTree *lookahead = NULL;
size_t last_position, position = 0;
self->is_split = ts_stack_head_count(self->stack) > 1;
2015-12-24 10:21:42 -08:00
2015-10-14 21:52:13 -07:00
for (int head = 0; head < ts_stack_head_count(self->stack);) {
for (bool removed = false; !removed;) {
last_position = position;
size_t new_position = ts_stack_top_position(self->stack, head).chars;
if (new_position > max_position) {
max_position = new_position;
head++;
break;
} else if (new_position == max_position && head > 0) {
head++;
break;
}
position = new_position;
LOG_ACTION("process head:%d, head_count:%d, state:%d, pos:%lu", head,
ts_stack_head_count(self->stack),
ts_stack_top_state(self->stack, head), position);
if (!lookahead || (position != last_position) ||
!ts_parser__can_reuse(self, head, lookahead)) {
ts_tree_release(lookahead);
lookahead = ts_parser__get_next_lookahead(self, head);
if (!lookahead)
return NULL;
}
LOG_ACTION("lookahead sym:%s, size:%lu", SYM_NAME(lookahead->symbol),
ts_tree_total_chars(lookahead));
switch (ts_parser__consume_lookahead(self, head, lookahead)) {
case ParseActionFailed:
ts_tree_release(lookahead);
goto error;
case ParseActionRemoved:
removed = true;
break;
case ParseActionUpdated:
break;
}
}
2014-08-09 01:03:55 -07:00
}
2015-12-08 13:01:33 -08:00
ts_tree_release(lookahead);
2015-12-24 22:04:20 -08:00
if (ts_stack_head_count(self->stack) == 0) {
ts_stack_clear(self->stack);
2015-12-24 22:04:20 -08:00
ts_tree_assign_parents(self->finished_tree);
return self->finished_tree;
}
2014-07-20 20:27:33 -07:00
}
error:
return NULL;
}