spider_string::q_append and other methods are modelled after String methods in sql/sql_string.h.

I think it'd be very confusing if methods with identical names (e.g. q_append) will start behaving differently in spider_string and String)

I think that we can avoid relying on the implicit assumption by using spider_string::replace() instead of spider_string::length() + spider_string::q_append(). However, I wouldn't like to make major change in GA older versions. So, I will limit the scope of the issue to 10.9+.

I noticed that we cannot simply replace spider_string::q_append() with spider_string::append() because Spider relies on the fact that q_append() never call String::realloc_with_efraxtra_if_needed(). realloc_with_efraxtra_if_needed() might wipes the data, in String::Ptr, which is placed beyond String::str_length.

For example, the test spider.ha fails, if we do the simple replacement. That is because spider_mbase_share::append_table_name_with_adjusting() behaves wrongly.

Thank you for your comment.

I would like to replace all q_appends with append, but that would be too big of a change. So, I will make spider_string::q_append() behave exactly same as spider_string::append(). Then, I'd gradually remove q_apppend(). I will do this on 10.3+.

that'd be str->append(). q_append() means "quick append" that doesn't check if there's enough space and assumes the caller knows what he's doing. Invoking reserve() explicitly every time before q_append() makes little sense, one can just as well use append().