Return inserted ids MySQL #279
Conversation
| [number_of_inserts, []] | ||
| if first_inserted_id | ||
| last_inserted_id = first_inserted_id + number_of_inserts - 1 | ||
| ids = (first_inserted_id..last_inserted_id).to_a |
There was a problem hiding this comment.
@zdennis what do you think about this approach?
There was a problem hiding this comment.
@robertomiranda, thank you for contributing to ar-import! With regard to your question above I have the same concerns as @tonic20 and @jkowens. I'd want to ensure that we however this implemented we can guarantee reliability and avoid unsuspecting issues when the returned IDs are wrong.
|
The database table would need to be InnoDB with |
| else | ||
| value_sets = ::ActiveRecord::Import::ValueSetsBytesParser.parse(values, | ||
| reserved_bytes: sql_size, | ||
| max_bytes: max) | ||
| value_sets.each do |value_set| | ||
| number_of_inserts += 1 | ||
| sql2insert = base_sql + value_set.join( ',' ) + post_sql | ||
| insert( sql2insert, *args ) | ||
| ids << insert( sql2insert, *args ) |
There was a problem hiding this comment.
This will result in missing ids because this is a batch operation.
|
Would be nice to have this feature. It will work fine for all users with |
|
This will also not work when there are duplicates and you use on_duplicate_key_update, right? Than it returns wrong id's probably. |
|
@philipgiuliani you're correct, this would not work with on_duplicate_key_update. |
|
I'm going to close this PR for now because it doesn't seem like MySQL provides a good solution for this problem at the moment. With the need to support |
ref #94