Fix inconsistencies in nodes sizes after edits

This commit is contained in:
Max Brunsfeld 2016-09-19 13:35:08 -07:00
parent ae3b912ddc
commit 3014101104
6 changed files with 70 additions and 67 deletions

View file

@ -10,63 +10,69 @@
#include "helpers/random_helpers.h"
#include <set>
static void expect_the_correct_tree(TSNode node, TSDocument *document, string tree_string) {
const char *node_string = ts_node_string(node, document);
static void assert_correct_tree_shape(const TSDocument *document, string tree_string) {
TSNode root_node = ts_document_root_node(document);
const char *node_string = ts_node_string(root_node, document);
string result(node_string);
ts_free((void *)node_string);
AssertThat(result, Equals(tree_string));
}
static void expect_a_consistent_tree(TSNode node, TSDocument *document) {
static void assert_consistent_sizes(TSNode node) {
size_t child_count = ts_node_child_count(node);
size_t start_byte = ts_node_start_byte(node);
size_t end_byte = ts_node_end_byte(node);
size_t start_char = ts_node_start_char(node);
size_t end_char = ts_node_end_char(node);
TSPoint start_point = ts_node_start_point(node);
TSPoint end_point = ts_node_end_point(node);
bool has_changes = ts_node_has_changes(node);
bool some_child_has_changes = false;
AssertThat(start_byte, !IsGreaterThan(end_byte));
AssertThat(start_char, !IsGreaterThan(end_char));
AssertThat(start_point, !IsGreaterThan(end_point));
size_t last_child_end_byte = start_byte;
size_t last_child_end_char = start_char;
TSPoint last_child_end_point = start_point;
for (size_t i = 0; i < child_count; i++) {
TSNode child = ts_node_child(node, i);
size_t child_start_byte = ts_node_start_byte(child);
size_t child_end_byte = ts_node_end_byte(child);
size_t child_start_char = ts_node_start_char(child);
size_t child_end_char = ts_node_end_char(child);
TSPoint child_start_point = ts_node_start_point(child);
TSPoint child_end_point = ts_node_end_point(child);
AssertThat(child_start_byte, !IsLessThan(last_child_end_byte));
AssertThat(child_start_char, !IsLessThan(last_child_end_char));
AssertThat(child_start_point, !IsLessThan(last_child_end_point));
last_child_end_byte = child_end_byte;
last_child_end_char = child_end_char;
last_child_end_point = child_end_point;
expect_a_consistent_tree(child, document);
assert_consistent_sizes(child);
if (ts_node_has_changes(child))
some_child_has_changes = true;
last_child_end_byte = ts_node_end_byte(child);
last_child_end_point = ts_node_end_point(child);
}
if (child_count > 0) {
AssertThat(end_byte, !IsLessThan(last_child_end_byte));
AssertThat(end_point, !IsLessThan(last_child_end_point));
if (!has_changes)
AssertThat(end_char, !IsLessThan(last_child_end_char));
AssertThat(has_changes, Equals(some_child_has_changes));
}
if (some_child_has_changes) {
AssertThat(ts_node_has_changes(node), IsTrue());
}
}
static void assert_correct_tree_size(TSDocument *document, string content) {
TSNode root_node = ts_document_root_node(document);
size_t expected_size = content.size();
// In the JSON grammar, the start rule (`_value`) is hidden, so the node
// returned from `ts_document_root_node` (e.g. an `object` node), does not
// actually point to the root of the tree. In this weird case, trailing
// whitespace is not included in the root node's size.
//
// TODO: Fix this inconsistency. Maybe disallow the start rule being hidden?
if (ts_document_language(document) == get_test_language("json") &&
string(ts_node_type(root_node, document)) != "ERROR")
expected_size = content.find_last_not_of("\n ") + 1;
AssertThat(ts_node_end_byte(root_node), Equals(expected_size));
assert_consistent_sizes(root_node);
}
START_TEST
@ -105,9 +111,8 @@ describe("The Corpus", []() {
input = new SpyInput(entry.input, 3);
ts_document_set_input(document, input->input());
edit_sequence();
TSNode root_node = ts_document_root_node(document);
expect_the_correct_tree(root_node, document, entry.tree_string);
expect_a_consistent_tree(root_node, document);
assert_correct_tree_shape(document, entry.tree_string);
assert_correct_tree_size(document, input->content);
delete input;
});
};
@ -130,10 +135,10 @@ describe("The Corpus", []() {
it_handles_edit_sequence("repairing an insertion of " + description, [&]() {
ts_document_edit(document, input->replace(edit_position, 0, inserted_text));
ts_document_parse(document);
expect_a_consistent_tree(ts_document_root_node(document), document);
assert_correct_tree_size(document, input->content);
ts_document_edit(document, input->undo());
assert_correct_tree_size(document, input->content);
ts_document_parse(document);
});
}
@ -144,10 +149,10 @@ describe("The Corpus", []() {
it_handles_edit_sequence("repairing a deletion of " + desription, [&]() {
ts_document_edit(document, input->replace(edit_position, deletion_size, ""));
ts_document_parse(document);
expect_a_consistent_tree(ts_document_root_node(document), document);
assert_correct_tree_size(document, input->content);
ts_document_edit(document, input->undo());
assert_correct_tree_size(document, input->content);
ts_document_parse(document);
});
}

View file

@ -109,7 +109,7 @@ describe("Document", [&]() {
assert_node_string_equals(
new_root,
"(object (pair (string) (array (null) (number))))");
AssertThat(spy_input->strings_read, Equals(vector<string>({" [null, 2", ""})));
AssertThat(spy_input->strings_read, Equals(vector<string>({" [null, 2"})));
});
it("reads from the new input correctly when the old input was blank", [&]() {

View file

@ -138,19 +138,8 @@ describe("Node", []() {
it("returns an iterator that yields each of the node's symbols", [&]() {
const TSLanguage *language = ts_document_language(document);
TSSymbolIterator iterator = ts_node_symbols(array_node);
AssertThat(iterator.done, Equals(false));
AssertThat(ts_language_symbol_name(language, iterator.value), Equals("array"));
ts_symbol_iterator_next(&iterator);
AssertThat(iterator.done, Equals(false));
AssertThat(ts_language_symbol_name(language, iterator.value), Equals("_value"));
ts_symbol_iterator_next(&iterator);
AssertThat(iterator.done, Equals(true));
TSNode false_node = ts_node_descendant_for_char_range(array_node, false_index, false_index + 1);
iterator = ts_node_symbols(false_node);
TSSymbolIterator iterator = ts_node_symbols(false_node);
AssertThat(iterator.done, Equals(false));
AssertThat(ts_language_symbol_name(language, iterator.value), Equals("false"));

View file

@ -127,7 +127,6 @@ describe("Parser", [&]() {
TSNode error = ts_node_named_child(root, 1);
AssertThat(ts_node_symbol(error), Equals(ts_builtin_sym_error));
AssertThat(ts_node_type(error, doc), Equals("ERROR"));
AssertThat(get_node_text(error), Equals(", faaaaalse"));
AssertThat(ts_node_child_count(error), Equals<size_t>(2));
TSNode comma = ts_node_child(error, 0);
@ -162,6 +161,15 @@ describe("Parser", [&]() {
});
});
describe("when there is an unexpected string at the end of a token", [&]() {
it("computes the error's size and position correctly", [&]() {
set_text(" [123, \"hi\n, true]");
assert_root_node(
"(array (number) (ERROR (UNEXPECTED '\\n')) (true))");
});
});
describe("when there is an unterminated error", [&]() {
it("maintains a consistent tree", [&]() {
ts_document_set_language(doc, get_test_language("javascript"));
@ -244,7 +252,7 @@ describe("Parser", [&]() {
"(identifier) "
"(math_op (number) (member_access (identifier) (identifier))))))");
AssertThat(input->strings_read, Equals(vector<string>({ " + abc.d)", "" })));
AssertThat(input->strings_read, Equals(vector<string>({ " + abc.d)" })));
});
});
@ -268,7 +276,7 @@ describe("Parser", [&]() {
"(number) "
"(math_op (number) (math_op (number) (identifier)))))))");
AssertThat(input->strings_read, Equals(vector<string>({ "123 || 5 +", "" })));
AssertThat(input->strings_read, Equals(vector<string>({ "123 || 5 +" })));
});
});
@ -517,7 +525,6 @@ describe("Parser", [&]() {
ts_document_free(doc);
doc = nullptr;
AssertThat(record_alloc::outstanding_allocation_indices(), IsEmpty());
}
record_alloc::stop();

View file

@ -82,7 +82,7 @@ void ts_document_edit(TSDocument *self, TSInputEdit edit) {
size_t max_bytes = ts_tree_total_bytes(self->tree);
if (edit.start_byte > max_bytes)
edit.start_byte = max_bytes;
return;
if (edit.bytes_removed > max_bytes - edit.start_byte)
edit.bytes_removed = max_bytes - edit.start_byte;

View file

@ -247,7 +247,6 @@ static TSTree *parser__lex(Parser *self, TSStateId parse_state) {
TSStateId start_state = self->language->lex_states[parse_state];
TSStateId current_state = start_state;
TSLength start_position = self->lexer.current_position;
TSLength position = start_position;
LOG("lex state:%d", start_state);
bool skipped_error = false;
@ -259,34 +258,32 @@ static TSTree *parser__lex(Parser *self, TSStateId parse_state) {
while (!self->language->lex_fn(&self->lexer, current_state)) {
if (current_state != TS_STATE_ERROR) {
LOG("retry_in_error_mode");
ts_lexer_reset(&self->lexer, position);
ts_lexer_start(&self->lexer, start_state);
current_state = TS_STATE_ERROR;
ts_lexer_reset(&self->lexer, start_position);
ts_lexer_start(&self->lexer, current_state);
continue;
}
if (self->lexer.lookahead == 0) {
self->lexer.result_symbol = ts_builtin_sym_error;
break;
if (!skipped_error) {
error_start_position = self->lexer.token_start_position;
first_error_character = self->lexer.lookahead;
}
if (self->lexer.current_position.bytes == position.bytes) {
if (!skipped_error) {
error_start_position = self->lexer.current_position;
first_error_character = self->lexer.lookahead;
if (self->lexer.current_position.bytes == error_end_position.bytes) {
if (self->lexer.lookahead == 0) {
self->lexer.result_symbol = ts_builtin_sym_error;
break;
}
skipped_error = true;
self->lexer.advance(&self->lexer, TS_STATE_ERROR, false);
error_end_position = self->lexer.current_position;
}
position = self->lexer.current_position;
skipped_error = true;
error_end_position = self->lexer.current_position;
}
TSTree *result;
if (skipped_error) {
error_start_position = ts_length_min(error_start_position, self->lexer.token_start_position);
TSLength padding = ts_length_sub(error_start_position, start_position);
TSLength size = ts_length_sub(error_end_position, error_start_position);
ts_lexer_reset(&self->lexer, error_end_position);
@ -804,7 +801,10 @@ static void parser__start(Parser *self, TSInput input, TSTree *previous_tree) {
self->finished_tree = NULL;
}
static bool parser__accept(Parser *self, StackVersion version) {
static bool parser__accept(Parser *self, StackVersion version, TSTree *lookahead) {
lookahead->extra = true;
assert(lookahead->symbol == ts_builtin_sym_end);
CHECK(ts_stack_push(self->stack, version, lookahead, false, 1));
StackPopResult pop = ts_stack_pop_all(self->stack, version);
CHECK(pop.status);
CHECK(pop.slices.size);
@ -821,7 +821,7 @@ static bool parser__accept(Parser *self, StackVersion version) {
for (size_t j = trees.size - 1; j + 1 > 0; j--) {
TSTree *child = trees.contents[j];
if (!child->extra) {
root = ts_tree_make_copy(child);
CHECK(root = ts_tree_make_copy(child));
root->child_count = 0;
for (size_t k = 0; k < child->child_count; k++)
ts_tree_retain(child->children[k]);
@ -848,6 +848,8 @@ static bool parser__accept(Parser *self, StackVersion version) {
return true;
error:
for (size_t i = 0; i < pop.slices.size; i++)
ts_tree_array_delete(&pop.slices.contents[i].trees);
return false;
}
@ -1021,7 +1023,7 @@ static bool parser__recover(Parser *self, StackVersion version, TSStateId state,
TreeArray children = array_new();
TSTree *parent = ts_tree_make_error_node(&children);
CHECK(parser__push(self, version, parent, 1));
return parser__accept(self, version);
return parser__accept(self, version, lookahead);
}
LOG("recover state:%u", state);
@ -1159,7 +1161,7 @@ static bool parser__advance(Parser *self, StackVersion version,
continue;
LOG("accept");
CHECK(parser__accept(self, version));
CHECK(parser__accept(self, version, lookahead));
ts_tree_release(lookahead);
return true;