fix load data parser bug

This commit is contained in:
bf0
2021-08-26 21:26:35 +08:00
committed by wangzelin.wzl
parent ca421cf838
commit 7be8c3653f
3 changed files with 29 additions and 33 deletions

View File

@ -1264,29 +1264,29 @@ int ObCSVParser::next_line(bool& yield_line)
bool yield = false;
int with_back_slash = 0;
for (; !yield && cur_pos_ != buf_end_pos_; ++cur_pos_) {
for (; !yield && cur_pos_ != buf_end_pos_; ++cur_pos_, ++cur_field_end_pos_) {
bool line_term_matched = false;
if (*cur_pos_ == formats_.enclose_char_ && !in_enclose_flag_ && cur_pos_ == cur_field_begin_pos_) {
in_enclose_flag_ = true;
}
if (!is_escaped_flag_ && *cur_pos_ == formats_.escape_char_) {
is_escaped_flag_ = true;
last_end_enclosed_ = NULL;
} else if ((*cur_pos_ == formats_.escape_char_ && formats_.escape_char_ != formats_.enclose_char_) ||
(in_enclose_flag_ && formats_.enclose_char_ == *cur_pos_ && cur_pos_ < buf_end_pos_ &&
formats_.enclose_char_ == *(cur_pos_ + 1))) {
if (cur_pos_ < buf_end_pos_) {
cur_pos_++;
if (!is_fast_parse_) {
*cur_field_end_pos_ = escaped_char(*cur_pos_, &with_back_slash);
}
}
} else {
char escaped_res = *cur_pos_;
if (is_escaped_flag_) {
escaped_res = escaped_char(*cur_pos_, &with_back_slash);
if (cur_field_end_pos_ != cur_pos_ && !is_fast_parse_) {
*cur_field_end_pos_ = *cur_pos_;
}
bool has_escaped = cur_field_end_pos_ != cur_pos_;
if (has_escaped && !is_fast_parse_) {
*cur_field_end_pos_ = escaped_res;
}
bool line_term_matched = false;
if (is_terminate_char(*cur_pos_, cur_field_end_pos_, line_term_matched)) {
if (formats_.enclose_char_ == *cur_pos_) {
last_end_enclosed_ = cur_field_end_pos_;
} else if (is_terminate_char(*cur_pos_, cur_field_end_pos_, line_term_matched)) {
if (!line_term_matched || cur_field_begin_pos_ < cur_pos_) {
handle_one_field(cur_field_end_pos_, has_escaped);
handle_one_field(cur_field_end_pos_, cur_field_end_pos_ != cur_pos_);
field_id_++;
}
char* next_pos = cur_pos_ + 1;
@ -1303,12 +1303,6 @@ int ObCSVParser::next_line(bool& yield_line)
cur_line_begin_pos_ = next_pos;
}
}
if (is_escaped_flag_) {
is_escaped_flag_ = false;
}
++cur_field_end_pos_;
}
}